Bo sizing guide

20 February 2014
SAP BusinessObjects BI 4
Sizing Guide

2 © SAP AG – Business Intelligence 4 Sizing Guide
Contents
Who should use this document? ..................................................................................................................6
What You Need to Know...............................................................................................................................6
Get the Latest Version of this Document......................................................................................................6
Pre-Sizing Checklist .......................................................................................................................................7
Post-Sizing Checklist......................................................................................................................................8
BI4 Machine Requirements...........................................................................................................................9
Introduction ..................................................................................................................................................9
Components and Platforms in Scope............................................................................................................9
Disclaimer....................................................................................................................................................10
Resizing .......................................................................................................................................................10
Sizing Tools..................................................................................................................................................11
BI4 Resource Usage Estimator................................................................................................................11
BI4.1 Configuration Wizard.....................................................................................................................11
BI 4 Architecture .........................................................................................................................................12
User Class Definitions .................................................................................................................................13
Users, Active Users, Active Concurrent Users ............................................................................................13
Conceptual tiers and services .....................................................................................................................14
Deployment and Sizing Methodology.........................................................................................................15
Sizing Steps and Methodology................................................................................................................15
Processing and Memory Requirements Calculations .............................................................................15
Prerequisites ...........................................................................................................................................16
Step 1: Resource Usage Estimator Setup................................................................................................17
Step 2: Intelligence DB Tier.....................................................................................................................18
Step 3: Intelligence Tier ..........................................................................................................................18
Step 4: Application Tier...........................................................................................................................18
Step 5: Processing Tier............................................................................................................................19
Analysis OLAP......................................................................................................................................19
Crystal Reports 2011...........................................................................................................................20
Crystal Reports for Enterprise.............................................................................................................20
Dashboard Design...............................................................................................................................22
Web Intelligence .................................................................................................................................23
Lifecycle Management Services..........................................................................................................25
Platform Search Services.....................................................................................................................26
Data Federation Service......................................................................................................................26

Adaptive Connectivity Service ............................................................................................................26
Step 6: Sizing Analysis and Scale-out......................................................................................................27
Step 7: APS Configuration.......................................................................................................................28
APS Service Groupings ........................................................................................................................28
Basic APS Configuration Instructions..................................................................................................29
Adaptive Job Server ............................................................................................................................30
Step 8: Deployment and Monitoring ......................................................................................................31
Scaling-out BI4 Servers........................................................................................................................31
Building the System ............................................................................................................................31
Monitoring BI4 ....................................................................................................................................32
Using Audit Reports ............................................................................................................................35
Monitoring CPU and Memory Usage..................................................................................................36
BI Tool Simulation Workflows.............................................................................................................36
Suggested Service Settings and Limits....................................................................................................37
Central Management Service (CMS)...................................................................................................37
Crystal Reports Cache Service.............................................................................................................37
Dashboard Cache Service, Dashboard Processing Service..................................................................38
Crystal Reports Processing Service .....................................................................................................38
File Repository Service (FRS)...............................................................................................................39
Web Intelligence (WebI) Processing Service.......................................................................................39
DSL Bridge Service...............................................................................................................................41
Connection Server...............................................................................................................................41
Web Application Server ......................................................................................................................42
Scale-out Memory Expectations.........................................................................................................42
Sizing Example.............................................................................................................................................43
Scheduling and Publishing ..........................................................................................................................51
Scheduling and Publishing Best Practices for Performance ...................................................................52
Use Dedicated Machines ....................................................................................................................52
Scaling .................................................................................................................................................52
I/O .......................................................................................................................................................52
Service Tuning.............................................................................................................................................53
Java Garbage Collection..........................................................................................................................53
Setting an APS Garbage Collector.......................................................................................................53
Setting the Tomcat Garbage Collector on Windows...........................................................................54
Setting the Tomcat Garbage Collector on UNIX and Linux .................................................................55

Minimum Memory Settings....................................................................................................................55
Web Intelligence Tuning .........................................................................................................................56
Visualization Service ...............................................................................................................................56
Web Application Server ..........................................................................................................................56
Crystal Reports for Enterprise Tuning.....................................................................................................57
Dashboards Tuning .................................................................................................................................59
Analysis OLAP Tuning..............................................................................................................................60
SAP BW Considerations and Recommendations ........................................................................................61
BW Configuration....................................................................................................................................61
Limiting Query Result Size.......................................................................................................................61
BEx Query Performance..........................................................................................................................61
Navigational Attributes...........................................................................................................................61
BW Specific Tuning and Configuration ...................................................................................................62
Set Specific Properties ........................................................................................................................62
Activate OLAP Cache and Delta Cache................................................................................................63
Additional Content..................................................................................................................................63
CMS Database Tuning.................................................................................................................................64
Virtualization: Service Level and Performance...........................................................................................65
Desktop Virtualization.................................................................................................................................66
Sizing SAP BusinessObjects Live Office .......................................................................................................67
Sizing SAP BusinessObjects Explorer...........................................................................................................68
Sizing SAP BusinessObjects Mobile.............................................................................................................69
Sizing SAP BusinessObjects Design Studio..................................................................................................69
Appendix .....................................................................................................................................................70
BI4 P&R Benchmark Testing Details........................................................................................................70
Test report documents and data characteristics................................................................................70
Report data description ......................................................................................................................73
CMS repository data description ........................................................................................................74
Test hardware specifications ..............................................................................................................74
Document Update History..........................................................................................................................75

© Copyright 2013, 2014 SAP AG. All rights reserved.
No part of this publication may be reproduced or
transmitted in any form or for any purpose without the
express permission of SAP AG. The information
contained herein may be changed without prior notice.
Some software products marketed by SAP AG and its
distributors contain proprietary software components of
other software vendors.
Microsoft, Windows, Outlook, and PowerPoint are
registered trademarks of Microsoft Corporation.
IBM, DB2, DB2 Universal Database, OS/2, Parallel
Sysplex, MVS/ESA, AIX, S/390, AS/400, OS/390,
OS/400, iSeries, pSeries, xSeries, zSeries, z/OS,
AFP,Intelligent Miner, WebSphere, Netfinity, Tivoli, and
Informix are trademarks or registered trademarks of IBM
Corporation in the United States and/or other countries.
Oracle is a registered trademark of Oracle Corporation.
UNIX, X/Open, OSF/1, and Motif are registered
trademarks of the Open Group.
Citrix, ICA, Program Neighborhood, MetaFrame,
WinFrame, VideoFrame, and MultiWin are trademarks or
registered trademarks of Citrix Systems, Inc.
HTML, XML, XHTML and W3C are trademarks or
registered trademarks of W3C®, World Wide Web
Consortium, Massachusetts Institute of Technology.
Java is a registered trademark of Sun Microsystems, Inc.
JavaScript is a registered trademark of Sun
Microsystems, Inc., used under license for technology
invented and implemented by Netscape.
MaxDB is a trademark of MySQL AB, Sweden.
SAP, R/3, mySAP, mySAP.com, xApps, xApp, SAP
NetWeaver, SAP BusinessObjects and other SAP
products and services mentioned herein as well as their
respective logos are trademarks or registered trademarks of
SAP AG in Germany and in several other countries all over
the world. All other product and service names mentioned
are the trademarks of their respective companies. Data
contained in this document serves informational purposes
only. National product specifications may vary.
These materials are subject to change without notice.
These materials are provided by SAP AG and its affiliated
companies ("SAP Group") for informational purposes only,
without representation or warranty of any kind, and SAP
Group shall not be liable for errors or omissions with respect
to the materials. The only warranties for SAP Group
products and services are those that are set forth in the
express warranty statements accompanying such products
and services, if any.
Nothing herein should be construed as constituting an
additional warranty.
Disclaimer
Some components of this product are based on Java™.
Any code change in these components may cause
unpredictable and severe malfunctions and is therefore
expressively prohibited, as is any decompilation of these
components.
SAP Library document classification: CUSTOMERS &
PARTNERS
Documentation in the SAP Service Marketplace
You can find this documentation at the following address:
http://service.sap.com/sizing

Who should use this document?
This document provides recommendations and best practices to help with deploying and scaling the
various services in SAP BusinessObjects BI 4.
We recommend consulting this guide if you are:
 Working on a Sizing strategy, especially in conjunction with the BI4 Resource Usage Estimator.
Be sure to use the most recent version of the Resource Usage Estimator.
 Planning to deploy SAP BusinessObjects BI 4 software
 Optimizing or tuning an existing SAP BusinessObjects BI 4 deployment
For all deployments, it is necessary to consult a Sizing Expert in order to validate sizing exercises after
initial planning using the Resource Usage Estimator and this Sizing Guide.
Sometimes customers, partners, or consultants develop their own applications, strongly modify SAP’s
out-of-the-box solutions, or implement complex integrations involving multiple systems. In these cases,
when sizing for the largest deployments or when your sizing requirements need to be extremely precise,
you should consider an “Expert Sizing”. An Expert Sizing is a hands-on exercise performed by or in
collaboration with an SAP Professional, where customer-specific data is analyzed and used to achieve a
sizing result with greater precision. The main objective is to determine the resource consumption of
customized content, applications, and usage patterns by taking comprehensive measurements. For
more information, see http://service.sap.com/sizing → Sizing Guidelines → General Sizing Procedures →
Expert Sizing.
What You Need to Know
To Size your deployment effectively and accurately, you need to know the following, which are
explained further in this document:
 The SAPS rating of your hardware. This is the performance rating of the hardware you are going
to use for your deployment. You also need to determine the number of actual* cores in the
system and from that determine the SAPS-per-core value.
 If you are using virtualized hardware, you need to know if the IT department will allow CPU and
memory reservations. Without reservations, the SAPS number will not be accurate and your
sizing calculations will be incorrect.
 The number of users of each BI Tool that you intend to use. Within that context, the types of
each user (consumer, business user, expert). This is explained further in this document.
 The relative document sizes for each BI Tool.
 The data sources the BI Tools will access
** ‘Actual’ cores refer to ‘full’ cores in a CPU. Hyper-threaded cores should not be counted for the
purposes of sizing.
Get the Latest Version of this Document
You can get the latest version of this document an www.sap.com/bisizing.

Pre-Sizing Checklist
These are the things you need to do before you start your sizing exercise:
 How many Active Concurrent users will you have for each BI tool? See “Users, Active Users,
Active Concurrent Users” in this document for more information.
 What types of users will be using each tool? Information Consumers (low workload), Business
Users (medium workload) or Experts (high workload)? See “User Class Definitions” in this document for
more information.
 Do you know how users will use each BI tool? Will they refresh documents rapidly? Will they
schedule them at off-peak time? Will they open three documents at once? Will they build dashboards
that consume multiple reports? How users use the tools affects the workload they produce.
 What types of machines will you be deploying to? What is the SAPS rating of each machine? See
“BI4 Machine Requirements” in this document for more information.
 What types of data sources will your users access? See “Data Sources” in this document for
more information.
 Will you be deploying to a virtualized environment? See “Virtualization: Service Level and
Performance” in this document and www.sap.com/bivirtualization for more information.
 Do not assume you are finished sizing if you only fill in the Resource Usage Estimator or the
Quicksizer. These tools do not cover all deployment scenarios nor all data sources. You must follow the
Deployment and Sizing Methodology section below to accurately size your deployment.

Post-Sizing Checklist
Here are the things you should consider once you have followed the sizing steps in this guide:
 Do a sanity check on your deployment landscape. Are there too many I/O-intensive services
running on the same machine? Is it configured to allow for peak usage? I.e., do you think it will run
comfortably (around 65% average utilization) so that peak times don’t cause issues?
 Are the other parts of your IT infrastructure ready and sized for the load the BI system will place
on them? Has your BW system been resized and patched? Are the relational databases tuned and ready
for the load?
 Check the BI Pattern Books for best practices in how to configure your deployment.
 Check www.sapbusinessobjectsbi.com for the latest information on BI4, especially the Upgrade
area if that applies to you.
 To simulate user load, check out the Performance Testing SAP BusinessObjects BI Platform Web
site.
 To learn more about getting the most out of your infrastructure, see the resources available at
www.sap.com/bisizing.
 To ensure good performance in a virtualized environment, see “Virtualization: Service Level and
Performance” in this document and www.sap.com/bivirtualization for more information.
 Install auditing reports so you can monitor your system. A set of auditing reports is available for
download at http://scn.sap.com/docs/DOC-6175
 Schedule a re-sizing in six months. See the Resizing section in this document for more
information.

BI4 Machine Requirements
For BI4, the minimum hardware for evaluation purposes is 8000 SAPS of processing power and 16 GB of
memory. This is defined in the BI 4.1 PAM, found here: http://scn.sap.com/docs/DOC-41355
Hardware requirements for deployments of any kind are determined by following the remainder of this
guide.
Introduction
This document discusses the Sizing of the Business Intelligence 4 suite of services, hereafter referred to
as BI4. The BI4 suite is designed to serve customers from small business to SAP’s largest customers. The
number of users, types of users, usage patterns, number of BI tools included in the suite, number of
data sources supported by the suite as well as the deployment options supported by the suite all factor
into a series of variables that affect the successful deployment of the BI suite. No configuration fits all
customers. The purpose of this document is to help guide you through the Sizing Exercise for BI4.
Sizing BI is very different compared to sizing of other types of Enterprise software. BI is a very resource
intensive task. The act of extracting information from a potentially large amount of data requires
adequate amounts of processing power and exercises all the important subsystems of a computer: CPU,
memory, disk and network. Having the right amount of capacity of these four aspects of your system is
key to success. BI can also be very bursty, since the load can rely a lot on the schedule of its users.
A BI system is used much like Google is: searching for information interactively. If Google required a
number of seconds to respond, most users would consider “the Internet broken”. They have an
expectation and often a business need to have a responsive system available to them. The purpose of
this document is to help ensure your users can access BI decision-making information in a timely
manner.
No tool or document can replace human judgment. So while this document attempts to cover as many
aspects of the Sizing Exercise as possible, Sizing Experts at SAP should always be consulted.
Make sure you have the latest version of this document and the other BI4 Sizing resources by visiting
www.sap.com/bisizing
Components and Platforms in Scope
The guidance and estimates referenced in this document are based on the following:
 Analysis OLAP
 Crystal Reports 2011
 Crystal Reports for Enterprise
 Dashboard Designer
 Web Intelligence

General sizing tips are also provided for:
 Explorer
 Mobile BI
 Live Office
 BI4 Platform services: Life Cycle Management, Search, Data Federation
 SAP BW
It is assumed that readers of this document are already familiar with core concepts from the Business
Intelligence Platform Administrator Guide (Admin Guide). There you can find conceptual information
and technical details on a wide number of topics which are intentionally not covered in this guide.
Disclaimer
This document demonstrates how someone might perform sizing of a BI4 system. The methodology and
sample walkthroughs offered here are examples of the tasks and thinking involved. The performance
and functioning of an actual system may vary for many reasons. The examples offered here should not
be considered a guarantee of success of a particular deployment.
You should work with your SAP account representative and ensure you get the advice of a Sizing Expert
when making final decisions with regards to a deployment of BI4.
Resizing
It is recommended that you resize your environment every six months or sooner if you have had
significant changes in the scope of your deployment. It can be difficult to anticipate how your users will
make use of the BI environment at the outset of a project. Resizing is recommended to ensure your
systems are configured for how your users are using the system.
In order to be informed accurately about the usage of the system, it is recommended that you take
advantage of the auditing features and auditing reports available for BusinessObjects BI4. A set of
auditing reports is available for download at http://scn.sap.com/docs/DOC-6175 . These reports will
help you determine how your BI services are being used.
Monitoring and probe reports are also recommended. Probe reports can give you alerts when time to
run expectations are longer than expected. See the BI4 Admin Guide for more information on creating
probes. CMS database performance should also be monitored.
Monitoring and recording of the basic CPU and memory usage of the machines in your deployment is
also recommended. The goal is to have an average CPU usage of 60% to 65% in order to handle peaks in
the range of 80%. Analyzing the historical usage of your deployment can help you determine if more
resources are needed for a particular node. Similarly, if memory is frequently being fully used, you may
be experiencing reduced performance and need to add more.

Sizing Tools
BI4 Resource Usage Estimator
The sizing tool reference in this document is the BI4 Resource Usage Estimator. It was previously named
the BI4 Sizing Estimator. The estimates it provides are based on Performance and Reliability (P&R) sizing
benchmark tests performed in SAP’s test labs. The results and this tool help to estimate load based on
performance results in these lab tests. They are only estimates and should not be used as simple
predictions of performance or deployment guidance for your project. The Resource Usage Estimator is
just a tool. It and this guide must be used together to provide effective sizing guidance.
BI4.1 Configuration Wizard
Starting with SAP BusinessObjects BI4.1, the Configuration Wizard helps the sizing exercise by defining a
number of APSes based on the deployment size chosen. This document explains how to perform a
custom sizing exercise. The Configuration Wizard’s APSes can be used as a time saver when it comes
time to define the APSes for your deployment. For more information, see the BI4.1 Admin Guide.

BI 4 Architecture
The BI4 architecture can be a complex set of services and technologies. The following is a representation
of the system from a technology perspective:
For the purposes of sizing BI4, the essential aspects of the BI4 architecture are this:
The BI4 Platform: This is a set of base services that enable BI services to run in a distributed deployment.
Each node in a BI4 cluster runs a Server Intelligence Agent (SIA) which runs selected services on that
node. Nodes in a cluster communicate over a service bus.
The BI4 Platform and suite of services are designed using a 64-bit architecture to ensure scalability to
the largest customers.
The CMS: This is where all the objects and configuration information that control and secure the
platform are located. The CMS is a combination of service plus database for storing and retrieving this
information.
Basic Services: Core services such as the File Repository (FRS) need to run somewhere in the cluster.
APS: The Application Processing Service (APS) is a service host. A large number of the BI services run
inside APSes. There can be more than one APS running on a given node, and there can be many nodes
running APSes in a BI landscape. Services running in the APS are Java-based and the APS controls the
JVM that the services run in. There is one JVM per APS.

Non-APS Services: Not all BI services run inside an APS. Those that don’t are run separately by the SIA.
BI Servcies: The BI services in the system have specific architectural considerations:
Crystal Reports: Crystal Reports services are available in two versions: Crystal Reports for Enterprise is
designed to have great Universe and SAP connectivity. It is a newer version of Crystal Reports. It uses
data access technology common to the other tools in the suite.
Crystal Reports 2011 is the original Crystal Reports product line, maintained and available for backward
compatibility.
Web Intelligence: WebI provides analysis and reporting capabilities to the suite. It provides compatibility
with old Universes (UNV), new Universes (UNX) as well as SAP BW. The Visualization Service (aka CVOM)
provides data visualizations (charting, graphics) services to WebI and the DSL and DSL Bridge house the
data services for WebI.
User Class Definitions
Three user classes are defined below. Class definitions are used to identify how much the system will be
used and how much load will be placed on it. Information Consumers use the system a little and Expert
Users use the system a lot. Knowing the mix of user classes is important since it directly affects the
performance and resources required by the system.
Be aware that customers often underestimate the amount of use that a user will make of BI system.
Information Consumers - The least active of all the user types. Information consumers spend an average
of 300 seconds (5 minutes) idle in between navigation steps. These users typically view predefined and
static content and perform relatively little drilling and filtering on their own.
Business Users - These users perform some moderate amount of drilling and filtering on their own.
Business users spend an average of 30 seconds idle in between navigation steps.
Expert Users - The most active of all the user types. Expert users spend an average of 10 seconds idle in
between navigation steps. These users are much more likely to perform resource-intensive operations in
the system including ad-hoc analysis and customization of reports, retrieving a large number of rows,
and heavy client-side filtering.
Users, Active Users, Active Concurrent Users
Many sizing exercises begin with a total number of users, for example: the customer needs a system
sized for 10,000 users. The next step is to determine how many of those users will be logged-in to the
system at the same time. The best way is to work with the customer to understand their usage
patterns. If a customer can’t provide guidance on this number, it’s common to estimate 10% as a
minimum. We call these active users. In this example, there would be 1,000 active users. Because
users can spend much of their time being idle after logging-in, you also need to determine how many of
the active users in the system will be concurrently generating load. This is another opportunity to take
customer usage patterns into account, but if there is no information available on this, the most common

estimate is 10% as a minimum. We call these users who are logged-in and concurrently generating load,
active concurrent users. In our example, there would be 100 active concurrent users.
When you use the BI 4 sizing tools, specify your inputs in terms of Active Concurrent Users.
In the example above, we went from 10,000 users to 100 active concurrent users, which translate to 1%
concurrency. It’s important to understand that the typical concurrency recommendations vary
depending on the size of deployment and other factors including the kind of work users perform.
If you expect or experience typical concurrency higher than these nominal ratios, you should expect a
heavier load and should compensate accordingly. For guidance in such a case, you may want to use the
Resource Usage Estimator and specify more users as ‘expert users’ to account for the increased use of
the system.
Conceptual tiers and services
The SAP BI platform can be thought of as a set of conceptual tiers. A subset of each list of services is
provided here for convenience. Refer to the BI4 Admin Guide for the complete list.
Application Tier
 Web Application Service (typically Tomcat)
o CMC – Central Management Console
o BI launch pad
Intelligence Tier (Referred to in the Admin Guide as “Management Tier”)
 CMS*1
 Crystal Reports Cache Service
 Dashboard Cache Service
 FRS
Processing Tier
 Adaptive Job Service
 Adaptive Processing Service (APS)
o Analysis OLAP and Multi-Dimensional Analysis Service (MDAS)
o DSL Bridge Service
o Visualization Service (CVOM)
 Crystal Reports Processing Service
 Web Intelligence Processing Service
*1
The database which the CMS uses for its repository may sometimes be referred to as the “Intelligence Tier DB”

Deployment and Sizing Methodology
Sizing a BI4 deployment requires a reasonable degree of planning so that calculations and predictions
can be made about the needs of the system. The number of users and the needs of those users can be
used to predict load on the system. The types of data sources used have an effect on the load and needs
of the system. The needs of the users include the BI services that they need to use. Some users use the
services a lot, some use them a little. Some reports are scheduled to be processed at night and viewed
during the day. Some reports need to be refreshed when viewed, which causes more load as the
number of users increases.
Once the user requirements are defined, the system can be defined that will achieve the required
amount of processing. The final step in the sizing is to apply it to the hardware landscape. The
deployment hardware can range from many small machines to one large machine. The Sizing Exercise
includes the allocation of BI services to the nodes in the system, taking into account the CPU, memory,
disk and network capabilities of nodes to be used in the construction of the system landscape.
Sizing Steps and Methodology
The basic approach to sizing is working through the requirements of all the tiers in the deployment,
accounting for the processing and memory requirements of each subsystem. Most subsystems and
services have special considerations that need to be taken into account.
Once the processing and memory requirements are determined, the next major step is to fit the services
to the hardware landscape. This requires knowing the SAPS per core rating of the computer to be used
as well as the amount of memory available to the machine or machines.
You must determine the SAPS rating of the machines that will make up your deployment. You
cannot assume any SAPS rating from this or any other document.
You can determine the SAPS rating of machines by visiting the Web site: SAP SD Standard Application
Benchmark Results
The BI4 Resource Usage Estimator is used as a tool to help calculate the processing load to be handled
by each tier in your system. The Estimator is used to get initial numbers for processing and memory
needs. Depending on the BI tool under consideration, additional calculations may need to be made with
corresponding changes and/or additions to the sizing.
Processing and Memory Requirements Calculations
For each tier, especially the Processing Tier, the processing power and memory requirements need to be
determined. This is the core of the Sizing Exercise. Once these numbers are determined, scale-out to
deployment hardware can be done.
The processing power is calculated and specified in a processing unit called SAPS. This is a processor and
computer independent unit that describes the processing power of a CPU in a given computer. SAPS is
an SAP-derived processing rating that allows you to calculate the power required by a deployment. It

takes into account the I/O processing capabilities of the computer in addition to its raw computation
abilities. Computers can have different SAPS processing capabilities even with the same type of CPU.
Prerequisites
The goal of the sizing exercise is to calculate the peak load that will be placed on the system. In order to
proceed with the sizing steps, you need to know the following information listed here. Gathering this
data accurately is the most important part of the sizing exercise since all the sizing calculations derive
from this information.
Users: How many Information Consumers, Business Users and Expert Users of each type of BI tool. An
average user workflow is also important to know. If you expect users to open five BI documents and
refresh them all at the same time, that’s five times the load of one user. That needs to be accounted for
since the system needs to handle that load.
It is very important to know if the common workflow of the users is going to include refreshing reports
and if so, how frequently. Will the reports be scheduled to run at night and only be viewed during the
day or will they be refreshed by users every hour? That’s an important part of the load prediction and
thus the sizing estimate.
Data Sources: What types of data sources will be used: direct-access SQL databases, UNV Universes,
UNX Universes, SAP BW, BW on HANA or HANA direct? It can be a mix. It’s important to know which
types of data sources will be used for the majority of BI processing so that peak load can be predicted.
Some customers expect their mix of data sources to change over time. This is an important thing to
consider.
Document Size: The relative size of each document is important to know. Will most documents be small
and cause little impact on the system or will most be large and require a lot of processing? This should
be determined for each BI tool expected to be used.
SAPS Rating: In order to know how much computing resources are needed for a deployment, you need
to know the computer’s performance rating. This is measured in SAPS. You may be building the system
using one computer or many. You may have the computer(s) specified in advance or not until the initial
sizing has been calculated.
The SAPS rating of the system’s intended hardware allows you to determine the required amount of
hardware to meet the system’s processing requirements. The SAPS rating takes into account CPU
processing power as well as the computer’s I/O capabilities. To learn the exact amount of SAPS
produced by your hardware, check the SAP SD Standard Application Benchmark Results Web site.
For the sizing exercise, the amount of SAPS per core needs to be determined. The benchmark shows the
SAPS rating for the computer. You then need to divide by the number of actual cores of the CPU.
Note: Hyper-threaded cores should not be used for this calculation.

Step 1: Resource Usage Estimator Setup
The Resource Usage Estimator should be initialized with the number of users of each type for each type
of BI tool. The report size sliders should also be set to reflect the sizes and types of documents that will
be processed by the system.
It might look like this:

Step 2: Intelligence DB Tier
The Intelligence database is the CMS database. For larger deployments, the Intelligence database should
be installed on a dedicated machine.
Be sure to follow your database vendor’s guidance regarding how to size and scale the deployment of
the database. Be aware that your database vendor may have strict recommendations regarding
deployment to a virtual machine.
If you intend to use an existing database that will be shared with other workloads, you must account for
the processing and memory requirements shown in the Resource Usage Estimator. I.e., it should have
free headroom to handle the given amount of processing (SAPS) and have the given amount of memory
free to do that processing.
In your sizing document, add in the SAPS and memory required for the Intelligence DB tier.
Step 3: Intelligence Tier
The workload of the Intelligence Tier is primarily the CMS. There are other services included in the
Intelligence Tier such as the caching services, FRS and LCM.
Note: the Resource Usage Estimator allocates the Crystal Reports Cache Server to the Intelligence Tier. It
is recommended that the Crystal Reports Cache Server be run on the same machine as the Crystal
Reports Processing Servers. In a scale-out scenario, a Crystal Reports Cache Server instance should be
allocated to run on each of the machines running Crystal Reports Processing Servers. It can be disabled
on nodes that are not running Crystal Reports Processing Servers.
In your sizing document, add in the SAPS and memory for the Intelligence Tier.
Step 4: Application Tier
The Application Tier is the Web server processing. Almost all interaction with the system is through the
Web server, including mobile clients.
It is recommended that each Web server be allocated a minimum of 8 GB of RAM for processing
requests as well be set to allow 900 maximum threads. For every 500 active concurrent users, it is
recommended that an additional Web server instance be created.

In your sizing document, add in the SAPS and memory for the Intelligence Tier.
Step 5: Processing Tier
In order to do tool-specific sizing using the Resource Usage Estimator, the numbers entered into the
Estimator are changed as each BI tool is analyzed. To get the starting sizing estimates for a certain BI
tool, set the user numbers for the other tools to zero.
Analysis OLAP
To size Analysis OLAP (aka MDAS), set the user numbers in the Resource Usage Estimator appropriately
and set the user numbers for all other tools to zero.
It is recommended that an APS be created for the purpose of running the Analysis OLAP service. This
allows the Analysis services to be monitored and configured easily.
Note: the default limit for the number of concurrent users of the Analysis OLAP service is 15. This should
be changed as you deploy the Analysis OLAP service to one or more APS instances.
Note: Analysis OLAP is also referred to as the ‘Multi Dimensional Analysis Service’ or MDAS.
See the BI4 Admin Guide for more information on changing Analysis OLAP settings.
In your sizing document, add in the SAPS and memory required for Analysis OLAP.

Crystal Reports 2011
To size Crystal Reports 2011, set the user numbers appropriately and set the user numbers for all other
tools to zero.
If your deployment will be to one large machine, you can record these requirements in your sizing
document and move to the next tool. Proceed here if you will be scaling out your deployment.
The Crystal Reports 2011 services required resources are determined by the processing requirements.
The required memory is calculated as follows:
Number of CPU cores x 1.25 GB/core
The number of cores is determined by the SAPS shown in the Resource Usage Estimator for the
Processing Tier (while having set only the Crystal Reports 2011 user numbers). To determine the
required CPU cores, divide the SAPS needed by your SAPS per core calculation (see the Prerequisites
section above for more information). You should round up when you encounter fractional cores.
Once you know the number of cores needed, you can calculate the memory requirement by multiplying
by 1.25 GB per core.
In your sizing document, add in the SAPS and memory for Crystal Reports 2011.
Crystal Reports for Enterprise
To size Crystal Reports for Enterprise, set the user numbers appropriately and set the user numbers for
all other tools to zero.
The Crystal Reports for Enterprise services required resources are determined by the processing
requirements.
The required memory is calculated as follows:
Number of CPU cores x 1.5 GB/core
The number of cores is determined by the SAPS shown in the Resource Usage Estimator for the
Processing Tier (while having set only the Crystal Reports for Enterprise user numbers). To determine
the required CPU cores, divide the SAPS needed by your SAPS per core calculation (see the Prerequisites
section above for more information). You should round up when you encounter fractional cores.

Once you know the number of cores needed, you can calculate the memory requirement.
For example, 20 cores of processing power requires 20 * 1.5 = 30 GB of memory.
In your sizing document, add in the SAPS and memory for Crystal Reports for Enterprise.
Data Sources
Crystal Reports for Enterprise is sized differently depending on the data sources expected to be used.
Reporting directly on relational data sources or reporting on Universes requires no additional sizing.
When reporting on SAP BW including SAP BW on HANA, additional sizing is required.
BW Data Source Sizing
Querying report data delivered via BW requires additional processing and memory resources. Sizing is
done relative to the amount of data you expect to be retrieved by an average query and also by the
frequency of queries submitted to BW.
Note: it is very important to have a fully patched BW system and also for your BW system to be sized
and tuned to your BI query needs. It is highly recommended that your BW system be checked and tuned
by a BW expert before connecting a BI system to it.
An example query is 100,000 rows with 10 fields, equivalent to 1 million cells. Your data set(s) may vary
in shape and size.
With a 2000 SAPS per core CPU, 1 million cells take approximately 10 seconds to process and require
1GB of memory for the processing. You need to judge the processing and memory requirements of your
queries accordingly, taking into account the SAPS per core rating of the computer. The processing time is
linear, so 500,000 cells should take about 5 seconds to process, for example. Similarly, if your CPU is
rated at 1000 SAPS, then 1 million cells will take about 20 seconds to process.
Once you know the amount of processing to be done for each query, you need to determine how often
queries will be submitted. This can be new queries or report refreshes. The refresh frequency may be
quite low in the case of scheduled reports, or may be quite high if a sales team needs to refresh their
reports all at the same time of day. It is important to get an accurate prediction of the peak load, since
under-rating this requirement can result in significant degradation of performance as the system
attempts to process multiple processing and memory-intensive operations at the same time.

Divide the number of queries expected per minute by the time is takes to process a query to get the
number of cores needed to process BW queries.
𝑞𝑢𝑒𝑟𝑖𝑒𝑠 𝑝𝑒𝑟 𝑐𝑜𝑟𝑒 𝑝𝑒𝑟 𝑚𝑖𝑛𝑢𝑡𝑒 =
60𝑠 𝑝𝑒𝑟 𝑚𝑖𝑛𝑢𝑡𝑒
single query process time (s)
𝑐𝑜𝑟𝑒𝑠 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑 =
required queries per minute
queries per core per minute
For example, to support 30 queries per minute where the query returns 1 million cells:
cores required =
30
(
60
10
)
=
30
6
= 5 cores
To size the memory requirement, multiply the expected query size by 1000 and round up to the nearest
GB. This amount of memory will be needed for each core to process the BW queries.
For example, 1,000,000 cells per query x 1000 = approximately 1GB per core.
 In your sizing document, add in the SAPS and memory required for BW query processing.
JDBC Considerations
For medium to high usage Universe scenarios that make use of JDBC connectivity, including HANA JDBC
connectivity, additional memory should be allocated for the Java environment created by the
Connection Server. You should add 2GB of memory to the node that hosts the Connection Server and
change the configuration of the Connection Server to allocate that additional memory for the JVM. See
the Connection Server tuning section Connection Server on page 41 for details.
Dashboard Design
To size Dashboard Design, set the user numbers appropriately and set the user numbers for all other
tools to zero.
In your sizing document, add in the SAPS and memory for Dashboard Design.

Web Intelligence
To size Web Intelligence, set the user numbers appropriately and set the user numbers for all other
tools to zero.
If less than 6GB of memory is shown, 6GB is recommended.
In your sizing document, add in the SAPS and memory for the WebI Processing Server.
Data Visualization
The Data Visualization (CVOM) service is used by WebI. This service is recommended to have a memory
allocation of 2GB.
 In your sizing document, add in 2GB of memory for the Visualization service.
DSL Bridge
The DSL Bridge service contains the data connectivity for UNX Universes as well as SAP BW. When WebI
uses either of these types of data sources, additional sizing for the DSL Bridge is necessary. See the two
following sections depending on the type of data sources that are expected to be used.
The DSL Bridge is an APS-hosted service.
Data Sources
WebI is sized differently depending on the data sources expected to be used. Reporting on UNV
Universes requires no additional sizing. When reporting on UNX Universes or SAP BW including SAP BW
on HANA, additional sizing is required. Apply sizing using the data source specific sections below as
appropriate.
UNX Universe Data Source Sizing
The sizing calculation for the DSL Bridge for UNX Universes is 0.25 GB per active concurrent user, 8GB
minimum, 30GB maximum.
If the calculation results in more than 30GB, additional DSL Bridge service instances should be created so
that no DSL Bridge instance has more than 30GB allocated to it.
 In your sizing document, add in the memory for the DSL Bridge service to handle UNX Universes.
Note: The processing power (SAPS) for UNX Universes is accounted for by the SAPS allocated to the
WebI Processing Server.

BW Data Source Sizing
Querying report data delivered via BW requires additional processing and memory resources. Sizing is
done relative to the amount of data you expect to be retrieved by an average query and also by the
frequency of queries submitted to BW.
Note: it is very important to have a fully patched BW system and also for your BW system to be sized
and tuned to your BI query needs. It is highly recommended that your BW system be checked and tuned
by a BW expert before connecting a BI system to it.
An example query is 100,000 rows with 10 fields, equivalent to 1 million cells. Your data set(s) may vary
in shape and size.
With a 2000 SAPS per core CPU, 1 million cells take approximately 10 seconds to process and requires
1GB of memory for the processing. You need to judge the processing and memory requirements of your
queries accordingly, taking into account the SAPS per core rating of your computer. The processing time
is linear, so 500,000 cells should take about 5 seconds to process, for example. Similarly, if your CPU is
rated at 1000 SAPS, then 1 million cells will take about 20 seconds to process.
Once you know the amount of processing to be done for each query, you need to determine how often
queries will be submitted. This can be new queries or report refreshes. The refresh frequency may be
quite low in the case of scheduled reports, or may be quite high if a sales team needs to refresh their
reports all at the same time of day. It is important to get an accurate prediction of the peak load, since
under-rating this requirement can result in significant degradation of performance as the system
attempts to process multiple processing and memory-intensive operations at the same time.
Divide the number of queries expected per minute by the time is takes to process a query to get the
number of cores needed to process BW queries.
For example, to support 30 queries per minute where the query returns 1 million cells:
cores required =
30
(
60
10
)
=
30
6
= 5 cores
To size the memory requirement, multiply the expected query size by 1000 and round up to the nearest
GB. This amount of memory will be needed for each core to process the BW queries.
For example, 1,000,000 cells per query x 1000 = approximately 1GB per core.

The additional processing and memory is to be allocated to the DSL Bridge service. In case the amount
of processing and/or memory exceeds the machine on which is runs, multiple DSL Bridge instances can
be created to scale-out the load.
 In your sizing document, add in the SAPS and memory for BW query processing for the DSL Bridge.
When receiving and processing data from BW, the DSL Bridge requires memory to hold the query
results. This has the same memory requirement as UNX Universes: 0.25 GB per active concurrent user,
8GB minimum, 30GB maximum.
If the calculation results in more than 30GB, additional DSL Bridge service instances should be created so
that no DSL Bridge instance has more than 30GB allocated to it.
 In your sizing document, add in the memory for the DSL Bridge service to buffer the BW query
results.
JDBC Considerations
For medium to high usage scenarios that make use of JDBC connectivity, including HANA JDBC
connectivity, additional memory should be allocated for the Java environment created by the
Connection Server. You should add 2GB of memory to the node that hosts the Connection Server and
change the configuration of the Connection Server to allocate that additional memory for the JVM. See
the Connection Server tuning section Connection Server on page 41 for details.
Lifecycle Management Services
‘Life Cycle Management’ refers to the following group of services:
 Visual Difference Service
 Lifecycle Management ClearCase Service
 Lifecycle Management Service
 Trace Log Service
It is recommended that these services be grouped into a LCM-specific APS for medium and larger
deployments. For medium to large sized deployments, this APS should be allocated at least 750 MB of
memory. For very large deployments, this APS should be allocated at least 1GB of memory.
For processing calculations, a nominal amount of SAPS of 1000 should be used.
In your sizing document, add in the SAPS and memory for the LCM
Note: For deployments prior to BI 4.1 SP3, only one instance of the LCM service should be run in a
cluster.

Platform Search Services
Sizing of the Search Service depends on your company’s intended use of search. It can be turned off,
used lightly or used heavily. If you intend to use search, the amount of resources to allocate to it
depends on how much and how frequent content is changed in your system.
For a system where content doesn’t change frequently during the day, you can schedule search to index
at night.
It is recommended that a separate APS be created just for the Search Service. This will allow for easy
monitoring and configuration of resources for search.
For light usage of search, allocate 500 MB of memory and 1000 SAPS. For heavier use of search, allocate
3 GB of memory and 3000 SAPS.
In your sizing document, add in the SAPS and memory for the Platform Search Service.
Data Federation Service
If you intend to use multi-source Universes, you need to size for the Data Federation service.
It is recommended that an APS be created for the purpose of running the Data Federation, especially for
large deployments. This allows the data federation services to be monitored and configured easily.
For medium deployments, allocate 1 GB of RAM. For larger deployments, allocate 3 GB of RAM.
In your sizing document, add in the SAPS and memory for the Data Federation service.
For more thorough sizing of the Data Federation service, see the Data Federation Service Sizing and
Tuning Companion Guide, located on SAP Service Market Place (SMP).
Adaptive Connectivity Service
For medium and larger deployments, the Adaptive Connectivity service needs sizing. For medium
deployments, 500 MB of memory is recommended. For large deployments, 1 GB of memory is
recommended.
It is recommended that an APS be created for the purpose of running the Adaptive Connectivity services,
especially for large deployments. This allows the data federation services to be monitored and
configured easily.
In your sizing document, add in the SAPS and memory for the Adaptive Connectivity service if needed.

Step 6: Sizing Analysis and Scale-out
Scaling-out your deployment is often necessary, where the size of the deployment is larger than a single
machine that is expected to run the BI system. Additionally, some analysis and thought is needed to
selectively place services on nodes in a cluster so that optimal performance and reliability is achieved.
Once you have the basic sizing as calculated in the above sections, you should have a list of all the
services with SAPS and memory requirements for each. The next step is to ensure the hardware used in
node of the deployment can handle the load.
Before assigning services to nodes in the system can occur, you need to determine what the constraints
of the system are. There will very likely be processing power and memory limits on each node. Some
nodes may be different.
For each of the tiers: Intelligence DB tier, Intelligence tier and Application tier and for each
of the BI tools to be used in your deployment, do the following for each of the workloads:
1. Start by assigning the workload to a new node in the system.
2. Add in memory to account for the operating system and basic BI infrastructure
components such as the SIA. 2GB1
of memory and 2000 SAPS are recommended for
these functions.
3. Check to see if either of the processing power of the node or the memory limit of the
node has been exceeded for the workload that has been assigned to it. If it fits, you can
move on to the next workload. If it doesn’t fit, you need to add additional nodes until
the full processing power and/or memory for the workload have been accommodated.
For workloads made up of APS-based services, an APS specific to that workload should be created on the
node for those services. A node may have other services defined on it, possibly as a result of the
standard installation. Any services not intended for use on a node should be disabled. See the APS
Configuration section below for more information.
In cases where more than one workload can fit on a node, it is fine to put multiple workloads on the
same node. However, be sure to take into account expected future growth and also mixing of similar
services. For example, Crystal Reports and WebI are both very I/O intensive so they shouldn’t be put on
the same node unless you don’t expect them to be used at the same time. Similarly, the Application Tier
hosting the Web server can be bound by memory and processing power, which the DSL Bridge can also
be – these services shouldn’t be put on the same node without some thought to the usage of the
system.
1
More than 2GB of memory may be needed depending on the hardware platform and operating system used.
Consult your machine vendor for more information.

Step 7: APS Configuration
The Adaptive Processing Server (APS) is the application that hosts a lot of the BI services. Multiple APSes
may be defined on multiple nodes within a deployment. In almost all cases more than one APS will be
needed in the system, both for management and maintenance of the running services. Often nodes in
the system need to run only a small subset of the available services in the default APS created at install
time, making at least some amount of APS configuration necessary on every machine. Configuring APSes
is an essential part of sizing BI4.
In BI 4.0, one APS is defined at installation. The rest of this section shows the basics of APS
configuration.
APS Service Groupings
When establishing your APSes, the following APS definitions are recommended:
Core APS WebI APS
Insight to Action Service Web Intelligence Monitoring Service
Publishing Post Processing Service Document Recovery Service
Publishing Service Rebean Service
Security Token Service Custom Data Access Service
Translation Service Excel Data Access Service
Monitoring APS Visualization APS
Monitoring Service Visualization Service
Client Auditing Proxy Service Web Intelligence Monitoring Service
Search APS WebI DSL Bridge APS
Platform Search Service DSL Bridge Service
Web Intelligence Monitoring Service
Auditing APS
Client Auditing Proxy Service DF APS
Data Federation Service
LCM APS
Visual Difference Service Connectivity APS
Lifecycle Management ClearCase
Service
Adaptive Connectivity Service
Lifecycle Management Service
Analysis APS
Multi Dimentional Analysis Service
BEx Web Application Service

Note: The color-coded groupings shown above can be used when deploying APSes for redundancy and
fail-over with minimal machine dependencies. The red group (right-hand column) services are related
while the blue group (left-hand column) are mostly unrelated but have no dependencies on the red
group. In cases where you want to duplicate services and also reduce or eliminate machine
dependencies, this partitioning can be used.
Basic APS Configuration Instructions
It is strongly recommended that you not alter the original APS defined by the installer. If that APS is not
needed, it should be left intact, stopped and disabled.
How to Create a New APS
 In the CMC, go to the Services listing
 On the Manage menu, choose New and then New Server
 On the Create New Server dialog box, choose the most appropriate Service Category
 Select the first of potentially many services from the Select Service list, then press Next.
 To add more services to the APS, select them in the Available Additional Services list press > to
add them to the Selected Services list. Press Next when done.
 Give the APS a unique name.
 Chose Create.
At this point the APS will exist but will not be enabled or started. You can enable and start your new APS
by right-clicking on the new service in the CMC and selecting the appropriate options.
How to Permanently Disable an APS
 Right-click on the server (APS) that you wish to modify
 Choose Disable Server
 Choose Properties
 Find the option: ‘Automatically start this server when the Server Intelligence Agent starts’.
 Uncheck the option.

How to Change the Memory Setting of an APS
 Locate the Command Line Parameters
 Scroll to the right until you locate -Xmx1G (or similar - it may have a different number)
 Change the number after the -Xmx. For example to set the APS to use 6GB, change the
command line to read -Xmx6G
How to Change the services running in an APS
 Choose Select Services
See the BI4 Admin Guide for in-depth information on configuring APSes.
Adaptive Job Server
Note that the Adaptive Job Server does not require “splitting” or configuration as described above for
the APS. This is because the Adaptive Job Server runs services as separate dedicated processes.

Step 8: Deployment and Monitoring
An important part of BI4 deployment is monitoring the operation of the various services to make sure
they are running at a reasonable level of utilization. For BI systems, an average utilization goal is 60% of
a node’s resources. This goal is prescribed because of the bursty nature of BI, which doesn’t run in a
steady state but is very influenced by user workload.
During initial installation and testing, monitoring of the system is very important since the best sizing
exercise can’t fully know how a system will behave and perform. Simulating user load is highly
recommended, using products such as Apache JMeter or HP LoadRunner. Start with a small number of
simulated users and bring the workload up slowly, watching how the system reacts. Proceeding
gradually will uncover any trouble spots that can be dealt with before progressing to a full workload.
It is critical to monitor all aspects of resource usage in the system:
 Disk
 Network
 CPU
 Memory
This needs to be done across the entire BI4 landscape, including the CMS database and reporting
databases. The monitoring services in BI4 provide a subset of the needed tools to do this.
Depending on your monitor solution, you should enable the following:
a. Proactively address
Scaling-out BI4 Servers
A very important aspect of sizing that is inferred by the scenarios above is the creation and deployment
of dedicated BI4 servers. The goal is to have dedicated machines running just the services that are
necessary for its role. This is a common requirement of a scale-out deployment. See the BI4 Admin
Guide for guidance on creating clustered deployments.
The Sizing Guide suggests that BI4 be installed as a full installation on each server and then any
unnecessary services be turned off. This allows for a more simple installation procedure and also allows
for future changes to the deployment. For example, if the role of a machine changes, all the BI4
components are there to be enabled and disabled as necessary.
Building the System
It is very important to methodically “build out” the system.
a. Start with a smaller landscape, using a smaller number of users to gain confidence in your
configuration. Gradually increase the load in increments of 50 to 200 users, only adding
services/servers as necessary.
b. Carefully monitor and analyze the performance and resource usage across the entire landscape,
including the CMS repository database, the Web application tier, and any other SAP BI 4
platform servers involved in the test to identify bottlenecks in either the underlying
infrastructure or server configuration.

c. Take the appropriate actions to resolve problem areas as you find them. For example, adding
another CMS host if CMS CPU utilization is above 80%. Testing gradually and solving issues as
they are found is important, since progressing immediately to a full workload can make
identifying the root cause of issues challenging.
The CMS database is key to the overall performance and scalability of the BI platform. A dedicated CMS
database running on dedicated hardware is always recommended. Work with your DBA to ensure the
CMS database server is correctly sized, configured, and monitored by referring to database vendor
materials on sizing.
The underlying infrastructure, including machines and network, is critical to the overall performance and
scalability of the BI platform; work with your infrastructure administrators to ensure the environment is
correctly sized, configured, and monitored.
When starting your SAP BI4 landscape, it’s recommended to methodically start each of your server
intelligence agent (SIA) nodes and ensure all services are correctly started before starting another node.
Monitoring BI4
The BI4 suite has a number of built-in monitoring capabilities that can be used to measure the
performance of the system as you build it up and continue into production.
Check the Processing Server Metrics
It is recommended that you frequently check the Processing Server Metrics for the BI processing
services that your deployment uses, especially in the setup phase of your project. Metrics are
maintained for each processing service such as Average Processing Time, Maximum Processing Time and
Minimum Processing Time.
Here is an example for the Web Intelligence Processing Server:

Some of the information shown includes:

Configure Report Processing Time Alerts
The Monitoring features of BI4 allow you to receive alerts when specific reports take too long to
process. You can configure probes that run a Web Intelligence or Crystal Reports document periodically.
You can be alerted if any of the documents take too long to run.
Step 1: In the Monitoring area of the CMC, open the properties for the Crystal Reports Service
(Processing Server) to configure the probe for Crystal Reports. Open Interactive Analysis to configure the
probe for Web Intelligence. You should already have documents created that you would like to monitor
the performance of. See the BI4 Admin Guide for more information on creating and configuring probes.
While the system is in testing and validation phases, report processing probes should be run frequently
(every few minutes) in order to effectively monitor the performance of the system.
Once a system is placed into production, it is suggested that you run report probes hourly so you can be
alerted if system performance degrades.
You can change the schedule for a report probe as shown here:

Step 2: In the Watchlist, create a new Watch that has rules based on the execution time of the probe.
Set the Caution to 5000 (milliseconds) to have a caution alert occur when the report takes longer than 5
seconds. This is the most time a user expects a report to require to process.
Set the Danger Rule to be some larger amount of time, such as 10 seconds.
See the BI4 Admin Guide for more information, including information on creating CMS Database alerts.
The performance of the CMS is vital to the overall operation of the BusinessObjects environment and
setting probes to monitor its performance is recommended.
Using Audit Reports
Usage information for your BI4 deployment can be obtained by using audit reports. This can be very
helping in determining how many users are using certain BI tools, certain documents, etc. You can
download a set of reports that are helpful for reporting on the BusinessObjects audit database here:
http://scn.sap.com/docs/DOC-6175

Monitoring CPU and Memory Usage
Monitoring and recording of the basic CPU and memory usage of the machines in your deployment is
also recommended. The goal is to have an average CPU usage of 60% to 65% in order to handle peaks in
the range of 80%. Analyzing the historical usage of your deployment can help you determine if more
resources are needed for a particular node. Similarly, if memory is frequently being fully used, you may
be experiencing reduced performance and need to add more.
BI Tool Simulation Workflows
Creating simulated user workflows are very important in the testing of a new deployment. The following
workflows are starting points for your user simulation. Be sure your simulations make use of the
features that most users will use.
Crystal Reports
1. Logon to OpenDocument web application
2. View the report in HTML viewer
3. Enter the Dynamic Cascading Parameters for live data and view first page
4. Go to the 2nd page
5. Go to the page in the middle of the report
Dashboards
1. Open BI launch pad page
2. Logon to BI launch pad
3. Navigate to Document List Tab
4. Expand Public Folder
5. Open the folder containing the Dashboard
6. View the Dashboard
7. Refresh the dashboard
8. Close the Dashboard
9. Logoff
Web Intelligence
1. Open BI launch pad page and login
2. Navigate to the folder that contains the WebI report
3. Open the report
4. Open Advanced Prompt and Run Query for retrieve live data
5. Navigate to Sub Report, “Nation”
6. Start Drilling Mode, Drill Down, and End Drill
7. Close Document
8. Logoff
Analysis OLAP
9. Open BI launch pad page
10. Logon to BI launch pad

11. Navigate to Document List
12. Open Saved Workspace
13. Swap Axes
14. Print Analysis
15. Close Workspace
16. Logoff
Suggested Service Settings and Limits
The following is a list of specific services and associated configuration information and suggested best
practices.
Central Management Service (CMS)
The Central Management Server (CMS) can scale horizontally and vertically. The CMS provides server,
user, session management, and security (access rights and authentication) management.
A single CMS instance can service a large number active concurrent users, depending on the resources
of the machine on which an instance runs. 500 active concurrent users per instance is a conservative
number useful for the purposes of planning your deployment. The recommended method for adding
CMS instances to your deployment is to monitor the CPU and memory consumption of each CMS
instance with a target goal of around 65% utilization. Adjust the number of CMSes while under
representative user load to achieve approximately 65% utilization. Too many CMSes can add
unnecessary synchronization traffic between instances. Too few CMSes will hinder performance of the
deployment.
Keep in mind that all CMS instances share the same repository database. If your repository DB vendor
doesn’t offer low-latency horizontal scalability, you’ll need to ensure you follow your DB vendor’s
guidance on how to size and scale vertically, especially for deployments with thousands of active
concurrent users.
** The performance of the CMS database is very important to the overall performance of the system. It
is strongly recommended that your CMS database be sized and tuned for high performance operation by
a qualified DBA. See the section, CMS Database Tuning CMS Database Tuning, for more information.
Crystal Reports Cache Service
The Crystal Reports Cache Service scales both up and out and can support 400 active concurrent users
per instance. You will very likely not need more than 1 instance on any given machine, since the Cache
Service is never enough of a bottleneck to make this necessary.
Because of the volume and frequency of communication between the Processing Service and the Cache
Service, it’s recommended that you deploy in pairs. That is, on each machine where there is a Cache
Service, also deploy a Processing service and vice versa. This can yield better performance than having
the Cache Service and Processing Service on different machines.

Dashboard Cache Service, Dashboard Processing Service
Because the technology used for the Crystal Reports and Dashboard Cache and Processing Services are
largely the same, the scalability characteristics and recommendations are also the same. Refer to the
Crystal Reports Cache Service and Crystal Reports Processing Service and use the same information for
Dashboards.
Required memory is calculated as follows:
# of children = # of logical cores ÷ 2
Each child may use up to 1.5 GB of RAM
Crystal Reports Processing Service
When sizing for workflows involving a small number of users viewing Crystal Reports, you may notice
high values for recommended memory. You may also notice that for a larger number of users the
required memory is almost constant (it does not increase much when more users are added). This is the
expected behavior and it is due to the internal functionality of the Crystal Reports processing servers
and engine, which are optimized to offer smoother overall performance, especially for a higher number
of users over a longer period of time. The optimal allocated memory is based on an algorithm that uses
the number of CPUs on the machine rather than the number of users in the system. Because of this, it’s
not recommended to have more than one Crystal Reports Processing Service on any given machine. As
you scale-out your deployment, it’s better to add instances to other machines that don’t current have
one.
Required memory is calculated as follows:
Crystal Reports 2011:
# of children = # of logical cores * 2.5
Each child can use up to 500 MB of RAM
 1.25 GB per core
Crystal Reports for Enterprise:
# of children = # of logical cores ÷ 2
Each child can use up to 3 GB of RAM
 1.5 GB per core
Because of the volume and frequency of communication between the Processing Service and the Cache
Service, it’s recommended that you deploy in pairs. That is, on each machine where there is a
Processing Service, also deploy a Cache Service and vice versa.
Data from the database is written to disk in the temporary directory during processing. The speed of
this disk will have an effect on processing performance.
Note: Each Crystal Reports child process (as noted above) can process a number of report processing
jobs at once. A job can potentially be shared among multiple user requests depending on sharing
criteria. Crystal Reports 2011 child processes default to 40 jobs each while Crystal Reports for Enterprise
child processes default to 20 jobs each. The number of jobs per child determines how many child
processes are created. The number of users served by a child process can be greater than the job limit if

job sharing is possible. Sharing can occur when a job’s report processing criteria match, such as database
security and parameter prompts. It is the role of the cache server to control job sharing.
File Repository Service (FRS)
While your deployment can have multiple Input and Output FRS instances, the first FRS pair (input +
output) to register with the CMS become the only active FRS pair. Other FRS instances are considered
passive backups. Although all FRS instances run simultaneously, only the active FRS pair handles
requests. If an active FRS fails, a passive FRS is changed to active status. When the previously active FRS
becomes operational again, it is registered as a passive FRS with the CMS.
Good disk and I/O performance is critical for the operation of the FRS.
Web Intelligence (WebI) Processing Service
You can scale the Web Intelligence Processing Service up or out. For every 200 active concurrent users,
it is suggested you create a WebI Processing Service instance. This amount is suggested given the I/O
processing that WebI usually requires. You can adjust the number of active concurrent users allowed
per server by adjusting the Maximum Connections setting.
In order to ensure the load on WebI servers is spread evenly among multiple instances, ensure the sum
of all Maximum Connections settings is greater than or equal to the number of active concurrent users
expected. For example, if you had 400 active concurrent users and two WebI Processing Service
instances, each should be set to at least 200 for Maximum Connections, for a total of 400.
When adding additional WebI instances, it is recommended that each instance run on a separate
machine in order to maximize the I/O capacity available to the servers. If you need to accommodate
more users on a single machine and it has the I/O capabilities, processing and memory to handle more
users, you can just increase the Maximum Connections setting (in addition to increasing the memory
allocation).

The following shows how the Maximum Connections setting is configured.
When tuning your WebI server instances, be sure to review all the settings shown here to be sure they
make sense for your intended usage. For example, if you intend to scale WebI up to take advantage of a
very capable machine but don’t adjust the Memory Maximum Threshold setting to match, you may find
the WebI service recycles itself unnecessarily. See the BI4 Admin Guide for more details on each of these
settings.
Data Source Considerations
When sizing WebI, it is important to consider the data source(s) being used. If you are using UNX
Universe or SAP BW data sources, the DSL Bridge service is employed by WebI, which hosts the data
connectivity. See the

Web Intelligence section above regarding sizing of the DSL Bridge in addition to the next section.
DSL Bridge Service
The DSL Bridge service is used by WebI when it connects to SAP BW and UNX-format Universes. It hosts
the data connectivity to these data sources.
Memory is calculated based on the number of expected active concurrent users as follows:
 0.25 GB per active concurrent user
 8 GB minimum
 30 GB maximum
The recommended maximum number of active concurrent users per DSL Bridge service is therefore 120.
If you require support for more active concurrent users, you can create additional DSL Bridge service
instances.
You can create DSL Bridge service instances with smaller maximum heap sizes if that suits your
landscape better.
Connection Server
For data connectivity scenarios that use the Connection Server (commonly WebI and Crystal Reports),
the recommended database connectivity is ODBC or native middleware, where available.
For Java-based (JDBC) middleware, including JDBC drivers for HANA, additional performance
configuration may be required for the Java Virtual Machine (JVM) that is created by the Connection
Server that hosts the JDBC driver. For larger user loads, the default JVM memory settings may be
inadequate. The memory allocated to the JVM created by the Connection Server can be increased by
editing the cs.cfg file, located in …/installdir/sap_bobj/enterprise_xi40/dataAccess/connectionServer on
Unix and Linux and here on Windows: C:Program Files (x86)SAP BusinessObjectsSAP BusinessObjects
Enterprise XI 4.0dataAccessconnectionServer.
Set the –Xms and –Xmx settings as shown below. Here they are being set to 1GB at startup and 2GB
maximum memory usage.
<JavaVM>



<Options>
<Option>-Xrs</Option>
<Option>-Xms1024m</Option>
<Option>-Xmx2048m</Option>
</Options>
</JavaVM>

Web Application Server
As you scale out your application tier, you should add an additional application server instance for every
500 additional active concurrent users. For better performance, it’s also recommended that for each
instance you configure at least 8 GB for heap size and 900 maximum threads. This is a minimum. See the
Resource Usage Estimator values to see if more memory is necessary for your deployment. These
recommendations are based on Tomcat; capacity for other Web Application Server vendors may vary.
When deploying your Web Application Server, it is more important to have a low latency connection to
the other BI 4 services than it is between the Web Application Server and the Web client. As such, you
should optimize the network connection between the servers where possible.
To learn more about deploying multiple instances of Tomcat, see the SAP Business Intelligence Platform
Pattern Books.
Scale-out Memory Expectations
As you scale out your deployment, you will need to know how much memory to allocate to each node in
the cluster of machines. The amount of memory required will depend on which services each node is
running.
The following table lists the expected memory requirements for these services. You should use these
values when calculating the memory requirements of a cluster node.
Service Expected Memory Requirements
Crystal Reports Cache Service 250 MB
Dashboard Design Cache Service 250 MB
CMS 1 GB
SIA 500 MB
Visualization Service (CVOM) 2 GB
See the Business Intelligence Platform Administrator Guide for a full list of services and how to configure
each service.

Sizing Example
The following is an example customer requirement and walkthrough of how to apply the sizing
methodology specified above.
Note: this sample doesn’t demonstrate every aspect of sizing. For your sizing be sure to follow all the
steps outlined in the section Deployment and Sizing Methodology, “Deployment and Sizing
Methodology” on page 15.
For this example, the following are the requirements set out by the customer. All user counts are
considered active-concurrent users.
 Web Intelligence:
o 150 business users, 10 expert users
o Data source: BW. Average query returns 200,000 cells. The customer has
indicated that a peak report refresh rate of 40 per minute needs to be
accommodated by the system.
 Crystal Reports 2011:
o 80 business users, 20 expert users
o Data source: relational database
 Analysis OLAP:
o 15 expert users
o Data source: BW
 The customer has indicated they won’t be using search heavily.
 The customer has indicated the system will be used internally only, so the Web server
doesn’t need extra network security, etc. required for exposure on the Internet.
 The customer’s standard machines have a SAPS rating of 25,120 SAPS delivered by 12
cores for a SAPS/core value of 2093.
The rest of this section demonstrates how to follow the methodology described above in the sizing of
this customer scenario.

Step 1: Resource Usage Estimator Setup
Enter the user numbers into the BI4 Resource Usage Estimator as shown here:
Step 2: Intelligence Tier DB
The Intelligence DB (the CMS database) must have processing headroom of 4322 SAPS and 10 GB of
memory available for processing CMS queries. Add these values to the sizing worksheet.
Step 3: Intelligence Tier
The CMS itself requires 8320 SAPS and 11 GB of memory to function properly. Add these values to the
sizing worksheet.
Step 4: Application Tier
The Web server is the main portion of the processing that occurs on the Application Tier. In this
deployment it needs 7911 SAPS and 8 GB of memory. Add these values to the sizing worksheet.

At this point your sizing worksheet might look like this:
Step 5: Processing Tier
This is the tier that includes all the BI tools.
Analysis OLAP:
To determine how much processing and memory resources are needed to support the deployment’s
Analysis OLAP needs, put just those numbers into the Resource Usage Estimator and record the
Processing Tier numbers.
Since Analysis OLAP will be using straight forward data connectivity, no special sizing is necessary.
Add these values to the sizing worksheet.
Crystal Reports 2011
Crystal Reports 2011 needs, put just those numbers into the Resource Usage Estimator and record the
Since Crystal Reports will be using straight forward data connectivity, no special sizing is necessary.
The memory required for Crystal Reports 2011 is calculated according to the processing needed. The
SAPS/core rating of the computer selected is 2033.
6509 ÷ 2033 = 3.2 cores. To ensure enough processing headroom, assume four cores.
Memory for Crystal Reports 2011 is sized by 1.25 GB per core, or 6 GB.
Web Intelligence

Web Intelligence needs, put just those numbers into the Resource Usage Estimator and record the
The Web Intelligence performance and memory numbers can go into the sizing worksheet as shown.
2 GB needs to be allocated to the Data Visualization service. Add that to the worksheet.
Since the data source is going to be BW as described, additional sizing is required for the data
connectivity to BW. The connectivity is hosted in the DSL Bridge service and so the DSL Bridge service
will need additional sizing.
To size for BW as a data source, this key information is known: the peak refresh rate of 40 refreshes per
hour and the size of the average query results: 200,000 cells.
Using the benchmark of a 1,000,000 cell query requiring 10 seconds to process on a 2000 SAPS CPU, we
can conclude that a 200,000 cell query will take 2 seconds to process.
Recall the way to calculate the queries per core per minute and the resulting cores required:
For this case, the queries per core per minute are:
60s/minute
2s/query
= 30 queries/minute
cores required =
40 queries/minute
30 𝑞𝑢𝑒𝑟𝑖𝑒𝑠/𝑚𝑖𝑛𝑢𝑡𝑒
= 1.33 cores. Round up to 2 cores.
To calculate the memory required per core by the DSL Bridge, multiply the number of cells by 1000:
200,000 x 1000 = 200 MB of memory per core, or 400 MB total.
To ensure a smooth running system, round the memory number up to the nearest GB, so in this case
allocate 1GB of memory for the DSL Bridge and also account for 2 cores being needed by it.
The BW query processing in the DSL Bridge needs to be buffered. This is allocated at 0.25 GB per active
concurrent user. In this scenario 160 active concurrent users are specified.
160 x 0.25 GB = 40 GB

Lifecycle Management Services
The LCM services for this size of deployment, a medium size deployment, requires 750 MB of memory.
1000 SAPS are adequate for the processing needed.
Platform Search Services
Platform Search can be sized for such a deployment, considering they won’t be using search heavily, at
500 MB of memory and 1000 SAPS of processing power.
Note: Considering the customer wishes to make light usage of Search, it would be wise to configure the
Search Services to work in a limited fashion. See the BI4 Admin Guide for more details for configuring
Search Services.
At this point, your sizing worksheet might look like this:
Sizing Analysis and Scale-out
For this deployment, recall that the customer has 12 core machines with 2093 SAPS per core. In the
spreadsheet shown above, the number of cores needed for each service has been calculated by dividing
the SAPS for each service by the SAPS per core of the specified processor.

Now that we know the number of cores needed, we can focus on distributing the services across
machines, taking into account the number of cores per machine. It is assumed that memory is relatively
cheap compared to additional machines, so some machines will have requirements for more memory
than others.
Remember that for each new machine, an operating system needs to be accounted for as well as the
basic communication infrastructure of the BI system.
Distributing the Intelligence and Application Tiers
For this deployment, the relatively small processing load required by the CMS and Web server could
allow it to be deployed to the same machine. However, the focus of performance for user interaction is
primarily the Web application server and the CMS, so it is not suggested to run them on the same
machine.
Machine #1 will house the CMS services and also LCM. LCM runs infrequently and makes use of the CMS
so running them on the same machine for this scenario makes sense. Machine #2 will host the Web
application server. At this point the processing usage is low but this leaves room in the future for
additional load by mobile services, etc. Since the processing load on this machine is relatively low,
Platform Search can be allocated to it as well.
Machine #1: CMS SAPS Memory Cores
OS and BI Core 2000 2 1
Intelligence DB 4322 10 2
Intelligence Tier (CMS) 8320 11 4
LCM 1000 0.75 0.5
32 8
Machine #2: Web App SAPS Memory Cores
Application Tier (Web Server) 7911 8 4
Platform Search 1000 0.5 0.5
11 6
Distributing the BI Processing Services
For this deployment, the Analysis OLAP and Crystal Reports processing fit on one machine.
Machine #3: AOLAP, CR SAPS Memory Cores
Analysis OLAP 5528 10 3
Crystal Reports 2011 6509 6 4
18 8

The Web Intelligence sizing for this customer requires 14 cores for the processing services and 2 cores
for the DSL Bridge. 16 cores exceed a single machine’s 12, so two machines will be specified for WebI
processing. Spreading the processing and memory for the Web Intelligence services (processing server,
DSL Bridge and Data Visualization) evenly across the two machines will provide processing headroom for
the future, which is desirable.
Machine #4, Machine #5: WebI SAPS Memory Cores
Web Intelligence 28469 6 7
Data Visualization (CVOM) 2
DSL Bridge 2093 21 1
31 9
Each of the WebI Processing Server instances should be configured in this case to have a Maximum
Connections setting of 160 ÷ 2 = 80 connections. The WebI Processing Service on each machine is
allocated 6GB to ensure it has adequate resources for its job.
APS Configuration
APSes need to be created on four of the five machines. Following the APS Configuration guidance in Step
6 above on page 27, the following APSes should be defined and allocated to the machines as follows:
Machine #1: CMS
 Core APS
 Monitoring APS
 Auditing APS
 LCM APS
Machine #3: AOLAP, CR
 Analysis APS
Machine #3, #4: WebI
 WebI APS
 Visualization APS
 WebI DSL Bridge APS
 Connectivity APS
Deployment
When deploying BI4 to these machines, all services should be installed and then selectively disabled
once the cluster is created. This will aid future expansion and patching, so that services can be turned
back on where and when they are needed. For example, patching of machines can be done in parallel if
a CMS is running on each node in the cluster. For day-to-day processing, having a CMS running on each

node is not necessary. However, it is relatively easy to bring up a CMS instance on each node if it is
already installed.
Redundancy and Virtualization
This sample deployment doesn’t take into consideration redundancy or virtualization. Those are
important topics that would be part of any Enterprise software deployment. For further details on
deploying BI4, see the BI4 Admin Guide.

Scheduling and Publishing
Sizing for Scheduling and Publishing is very different than sizing for interactive users. You typically have a
time budget to work with but are not as concerned with response time. However, if you want to have
your jobs run in a time window, sizing is needed.
When considering a starting point for sizing, specify the user count in the Resource Usage Estimator as
Expert users. This most closely matches the demand made by the job schedulers, which don’t have any
human ‘think time’ at all. For the number of active concurrent users, specify the number of concurrent
jobs that may be running.
When sizing for scheduling and publishing, you need to consider the time of day that those operations
are done with respect to the workload of your interactive users. If they overlap, the amount of workload
from scheduling and/or publishing needs to be added to the workload of your users. If the schedules do
not overlap, for example if scheduling is done at night and your users only use the system during your
day, then you size for the larger of the two scenarios.
Figure 1: Workloads: Combined or separate
For Publishing, it is especially important to understand how many concurrent personalization jobs will
be running as part of each publication. For example, if you need to serve three regions in a Publishing
job and thus have three separate database queries, that would be equivalent to three active concurrent
users. If you are publishing with personalizations that require a query-per-recipient, you need to
determine the number of queries that might be able to be processed at once and use that number as
the active concurrent user count.
If you are unsure about how much time is being spent running scheduled reports, you can deploy the
auditing reports available from SAP. These reports help you to report on auditing information stored in
the BusinessObjects audit database. You can download the reports and learn more here:
http://scn.sap.com/docs/DOC-6175
Workloads
Active Users Scheduling

Scheduling and Publishing Best Practices for Performance
Use Dedicated Machines
If both CPU and memory for the Adaptive Processing Server (APS) are heavily utilized during job
processing, move that APS to its own machine. A faster machine is good to consider as well.
Be sure BI4 Feature Pack 3 or later is installed. This feature pack introduced better scaling and CPU
utilization for scheduling and publishing.
For Publishing, isolate the Publishing Service and the Publication Post Processing Service on dedicated
APS instances on dedicated machines. This will allow post processing to occur while a subsequent
publishing job is processed.
Scaling
Horizontally scaling out the Publishing Service and Publication Post Processing Services across multiple
APSes will enable more publication instances to proceed concurrently.
For publications with many recipients, vertically scale the APS on machines that have more CPU cores
and memory. This will enable the Publishing Service to concurrently process more recipients and the
APS to generate more jobs.
I/O
Scheduling and especially Publishing are very I/O intensive. A machine with fast I/O for disk and/or
networking to a SAN is important for the location of the FRS folders.
For more guidance on ensuring scheduling and publishing jobs are configured to run as efficiently as
possible, see the BI4 Admin Guide.

Service Tuning
Configuration optimizations for the services listed below may improve their performance.
Java Garbage Collection
Note: BI 4.1 has this setting configured out of the box.
A number of services in the BI4 environment are Java based. Java provides a garbage-collected
environment, meaning that as portions of memory inside a process are no longer being used, the Java
environment will collect it and make it available for new tasks. Data that stays in use past a number of
collections is considered part of the old generation of data.
The default garbage collector used by Java 6 for old generation data is a single-threaded garbage
collector. While this garbage collector has the smallest memory overhead of all the available garbage
collectors, it runs with a single thread and doesn’t take advantage of the modern multi-core and multi-
CPU machines that BI4 is often installed on. This can cause long pause times while collecting the garbage
from the large number of tasks that can be running on multi-core hardware. This can cause unnecessary
performance degradation and availability issues by having all but one CPU core active while doing
periodic garbage collection.
It is recommended that the Java Parallel garbage collector be used for BI4 processes. These processes
include the APSes as well as the Tomcat Web server.
Setting an APS Garbage Collector
 Scroll to the end of the parameters

 Add the parameter -XX:+UseParallelOldGC as shown here:
Setting the Tomcat Garbage Collector on Windows
 In Windows, locate and run the Tomcat Configuration application in your server’s Start menu or
Start screen
 Select the Java tab
 In the Java Options box, specify -XX:+UseParallelOldGC as shown here:
See the Apache Tomcat administration guide for more details

Setting the Tomcat Garbage Collector on UNIX and Linux
On UNIX and Linux, changing the Tomcat configuration requires changing the script file that starts the
server. This is the setenv.sh file, located in the BI4 Tomcat folder:
<install location>/sap_bobj/tomcat/bin/setenv.sh
In that file, change the JAVA_OPTS parameter to include -XX:+UseParallelOldGC. It will look similar to
this:
Minimum Memory Settings
For Java-based processes such as the APS-based servers in a deployment, setting a nominal minimum
memory setting aids Java in its management of memory. The default for Java process such as the APS is
32MB, which is very small. For most APSes, the minimum suggestion is 512 MB, depending on the
expected load of the system.
How to Change the Initial (minimum) Memory Setting of an APS
 Scroll to the right until you locate -Xms parameter, if present.
 Change or add the parameter, setting the number after the -Xms with no space. For example to
set the APS to start with 512 MB, set the command line to read -Xms512M

Web Intelligence Tuning
DSL Bridge
You should ensure there are one or more dedicated APS instances running only the DSL Bridge Service.
The recommended heap size is 0.25 GB per active concurrent user, 8GB minimum, 30 GB maximum.
More DSL Bridge service instances can be created if necessary to support the number of active
concurrent users required.
Maximum Connections
Under heavy load, the WebI Processing Server can benefit from having a maximum connection setting
that is higher than the maximum number of expected concurrent connections. This is because under
heavy load, the time to close a connection can increase and WebI may not be ready to accept the next
connection in time. See the section Web Intelligence (WebI) Processing Service Web Intelligence (WebI)
Processing Service for more information on setting the maximum connections.
Multiple Processing Instances
It is not recommended that you run multiple WebI Processing Server instances on the same machine.
The two services can degrade the performance of each other as they make heavy use of I/O bandwidth,
etc. If you wish to run more than one server instance, they should be run on separate machines.
SAP BW
If you are using SAP BW, the recommended DSL Bridge Service parameters are:
 BICSResultChunkSize <# of rows of the biggest rowset> / 50
 BICSLOVChunkSize <# of items of the biggest LOV> / 20
These are both registry settings that are set differently on Windows and UNIX.
On Windows: Go to [HKEY_LOCAL_MACHINESOFTWARESAP BusinessObjectsSuite XI
4.0defaultWebIntelligenceCalculator] and create the above as DWORDs (decimal).
On UNIX: Go to INSTALL_BASE_DIR/sap_bobj/data/.bobj/registry/64/software/sap
businessobjects/suite xi 4.0/default/webintelligence/calculator and add two keys using parameter
names as above to the .registry file.
Visualization Service
The Visualization Service (CVOM) generates graphics and charts for WebI. This service should run in one
or more dedicated APS instances. 2GB of memory is recommended for this service.
Web Application Server
For deployments that involve sizable numbers of users, mobile users or remote users, it is
recommended that the Web tier be split to deliver static content by a separate Web application server.
Static content (HTML pages, images, documents, JavaScript and Cascading Style Sheets) – content that
doesn’t change – can be delivered by a Web server dedicated to that task. The Apache Web server is
typically used for this.

You can learn more about splitting the Web tier by consulting the Web Application Deployment Guide
found on http://help.sap.com. Additional help on doing a split Web tier can be found in the SAP
Community Network document, “Improving the User Experience in SAP BI Platform 4.0 with Apache and
WDeploy.” An update to this document is available that covers SAP BI 4.1 and the latest version of
Apache httpd, version 2.4. You can find that document here: http://scn.sap.com/docs/DOC-50325
A simplified guide has been created that walks you through the steps of the Apache Split for Windows
deployments: Quick Step-by-Step Guide to Apache Split Deployment with BI4.1
To learn more about deploying multiple instances of Tomcat, see the SAP Business Intelligence Platform
Pattern Books.
Crystal Reports for Enterprise Tuning
Controlling Job Creation
To ensure the number of job processing child processes created by the Crystal Reports Processing
Service doesn't exceed your needs, it is recommended that you configure the Maximum Concurrent Jobs
entry to match the number of users you expect to support. Open the Properties of the
CrystalReportsProcessingServer and locate the Crystal Reports Processing Service section as shown here:

Increasing Memory Allocation for Improved Response Time
Depending on your reporting needs and the design of your reports, it may be possible to improve the
processing time of reports by allocating more memory for Crystal Reports to work with. This is done by
increasing the amount of memory allocated to each Crystal Reports child process. You may want to do
this in the case where you have extra processing and memory headroom on a machine beyond your
essential sizing requirements. Allowing Crystal Reports to more quickly process reports has obvious
responsiveness benefits and also allows the system to be better able to handle spikes in demand.
Each Crystal Reports processing child process has a default memory allocation of approximately 2GB.
You can increase this by setting the Java memory parameter in the Java Child VM Argument list, as
shown below in the settings for the CrystalReportsProcessingServer in the CMC. In this example, the
memory is changed to 8GB by specifying 8192 MB using the parameter setting Xmx8192M.
You may want to increase the number of concurrent jobs if your goal is to increase the number of
reports that each child process can deliver. As shown here, the number of concurrent jobs has been set
to 80.

Dashboards Tuning
To ensure the number of job processing child processes created by the Dashboards Processing Service
doesn't exceed your needs, it is recommended that you configure the Maximum Concurrent Jobs entry
to match the number of users that you expect to support.

Analysis OLAP Tuning
Analysis OLAP (Multi-Dimensional Analysis Service or MDAS) works the best when run in its own APS.
See the section Step 7: APS Configuration, “Step 7: APS Configuration”.
Memory
Once in its own APS, it is recommended that MDAS be configured with a minimum of 4GB of memory.
The section Step 7: APS Configuration outlines the steps for configuring the memory settings of an APS.
Sessions
The MDAS “Maximum Client Session” setting is often set to a low number by default. It is recommended
that it be set in a range of 45 to 100 depending on the expected load.
Each MDAS can support up to 100 client sessions. To support more sessions, create additional APS
instances that host an MDAS service. It is recommended that additional MDAS APSes be deployed on
separate machines.

SAP BW Considerations and Recommendations
When using BI4 to report on SAP BW data, the following recommendations are made to ensure the best
performance possible:
BW Configuration
 Apply the latest fixes available from SAP by applying the latest support packs to your BW system
 It is strongly suggested that your BW be sized for the anticipated query load expected from the
BI4 system.
 Have a BW expert tune the system generally and BEx queries specifically for your BI4 usage
Limiting Query Result Size
Consider the BW Safety Belt feature, which helps to ensure report developers don’t create reports with
result sets that are larger than desired. The Safety Belt feature is described here:
https://scn.sap.com/docs/DOC-31900
BEx Query Performance
Check the performance of BEx queries. See the SAP note: BI Query Runtime Statistics which is found on
the SAP Support Web site at
http://help.sap.com/saphelp_nw70/helpdata/EN/43/e39fd25ff502d2e10000000a1553f7/frameset.htm
You can determine how much time the execution of certain user actions require in the front-end and in
the SAP BW analytic engine.
Navigational Attributes
A large number of navigational attributes defined in the underlying InfoProvider may impact the overall
performance (please refer to SAP best practices for data modeling). This can lead the BI4 tools to
generate a lot of "Crossjoin" operators, causing extra unused data to be returned in queries. Revisit your
BEx queries to be sure they include only the information necessary for BI.
Characteristic attributes can be converted into navigational attributes. They can be selected in the query
in exactly the same way as the characteristics for an InfoCube. In this case, a new edge/dimension is
added to the InfoCube. During the data selection for the query, the data manager connects the
InfoProvider and the master data table ('join') in order to fill the Query.
From a pure performance point of view, you should model an object on a characteristic rather than on a
navigational attribute.

BW Specific Tuning and Configuration
Set Specific Properties
 Check the option: Use Selection of Structure Elements
 Choose the Read Mode option: Query to Read When You Navigate or Expand
Checking the Read Mode property helps improve performance in situations where there are many
Restricted Key Figures and other calculations in the Query Definition.
If analysis of the query performance indicates very high EVENTID 3200 times and/or the FEMS number is
very high, try enabling this property and check performance.
Checking the Use Selection of Structure Elements option ensures the structure elements are sent to the
database for processing.
You can see these settings shown below:

Activate OLAP Cache and Delta Cache
A new cache mode is delivered with Support Package 16 for BI 7.0 and currently, the new OLAP cache in
B2P is not active yet (SP17 for BI7.0 is implemented in B2P). The Delta cache, which is especially
important to uncompressed InfoCubes, is not in use for some queries at this moment. This means that
the OLAP has to fetch the data from database again once new requests are sent to InfoCubes, while
delta cache is not enabled. This can result in an increase of OLAP time as well as Database time and
eventually the entire query runtime. It is recommended to activate the new OLAP Cache mode by
running the report SAP_RSADMIN_MAINTAIN and setting "Update cache objects in delta process".
Additional Content
More BW performance tuning information can be found here:
 Performance Optimizing SAP BusinessObjects Reports Based Upon SAP BW Using BICS Connectivity
http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/d0e3c552-e419-3010-1298-
b32e6210b58d?overridelayout=true
 Business Intelligence Performance Tuning
https://www.sdn.sap.com/irj/sdn/bi-performance-tuning
 Performance & Scalability Best Practices and Tuning Recommendations for SAPBO XI 3.1 on BW
http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/006b1374-8f91-2d10-fe9c-
f9fa12e2f595?QuickLink=index&overridelayout=true&48747879633011

CMS Database Tuning
The CMS database performance has a significant impact on the performance of the system. The
following Sybase database tuning tips are offered. These tips may apply to other database servers as
well.
1. Data Cache [buffer pool]
Data cache allows efficiently reducing IO. Having a sufficient size of data cache can speed up
queries.
2. Procedure Cache
Configure the Procedure Cache so frequently used database procedures are cached. This allows
the database to not need to reload procedures from disk as often.
3. Lock Granularity
The CMS does a lot of reading and writing from/to the database. By default, the granularity is
set to page level, which is too high. Page level locking promotes contention by the CMS when,
for example, several records are stored together in a same page. One of them may be being
updated but another one is being read. One of these two sessions has then to wait until other
has finished.
Setting Lock Granularity to use row level locking can avoid contention and deadlock issues.
4. Parallel Processing:
This is important to set so Sybase can fully utilize all the CPU cores available on the machine.

Virtualization: Service Level and Performance
BI4 works very well in virtualized environments. Testing has shown it performs equally well on
virtualized hardware as it does on dedicated hardware. However, in order to achieve equal
performance, the hardware it runs on must be dedicated to the virtual machines that the BI services run
in.
SAP strongly recommends that BI virtual machines have reservations for the memory and
CPUs assigned to them.
BI is a bursty application workload. It can be very I/O intensive when heavily loaded by users. When
experiencing maximum load as defined by your sizing, all parts of the system can go from moderately
loaded to very busy momentarily. If the BI hardware is being shared with other workloads in a
virtualized environment, the response to bursts in activity can result in slow-downs between dependent
BI services and ultimately end user response time.
It is recommended that you build your system to target 65% utilization so that bursts in activity can be
handled. Performance of both physical and virtual systems degrades when system utilization is greater
than 80%.
If you are unable to secure memory and CPU reservations for your virtualized deployment, it is
recommended that additional scale-out be done to provide adequate BI processing resources for peak
periods. The amount of additional scaling out will depend on your IT infrastructure and performance
policies.
The goal of Sizing BI4 is to have a deployment that can handle the workload needed and provide good
response time for end users. If you deploy to a shared virtualized environment, extra care is needed to
ensure your sizing exercise results in a system which has a Service Level and responsiveness that can
be maintained.
For more in-depth information on BI and Virtualization, see the SAP BI 4 Virtualization site at
www.sap.com/bivirtualization

Desktop Virtualization
When running BI4 desktop applications are deployed to virtualized desktop environments such as Citrix
and Windows Server Remote Desktop Services, sizing is required to determine how many concurrent
users can be supported for commonly used BI applications.
The following table shows the memory footprint of applications that you may run in a virtualized
desktop environment. You can use the memory footprint information to calculate the amount of
memory required to service the number of users required.
Product Memory Footprint per User
Web Intelligence Rich Client 1.5 GB
Crystal Reports for Enterprise 1.5 GB
Crystal Reports 2011 150 MB
Interactive Design Tool 500 MB
Universe Designer 500 MB

Sizing SAP BusinessObjects Live Office
SAP Live Office uses the Web Intelligence and Crystal Reports engines depending on which report parts
are inserted in your documents. When performing Sizing of BI4, Live Office usage counts as usage of
these BI services, including data sources, etc.
It is important to realize that if a document contains multiple BI report parts, refreshing that Office
document is the equivalent of the number of users interacting with the BI system all at the same time.
For example, if an Office document has three report parts and it gets refreshed, that’s the same as three
active concurrent users. Be sure to take this into consideration when accounting for the number of
active concurrent users of the BI system.
Here is a more detailed diagram of the Live Office architecture:
For more information on SAP Live Office, see the Live Office Web site here: http://help.sap.com/bolo

Sizing SAP BusinessObjects Explorer
Explorer has two deployment types:
Explorer Data Providers Components
Explorer 4.0: Provides fast
processing of small to large
data sets
 BusinessObjects
Universes (UNX)
 Excel files
 SAP BusinessObjects Explorer
 SAP BusinessObjects
Enterprise
Accelerated Version:
Provides processing of very
large data sets
 NetWeaver BWA
indexes enabled for
Explorer
 SAP HANA in-memory
computing engine
 Excel files
 SAP BusinessObjects Explorer
 SAP BusinessObjects
Enterprise
 SAP NetWeaver Business
Warehouse (BWA)
 SAP HANA
The following is the Explorer architecture:
For information on Sizing Explorer, see the SAP BusinessObjects Explorer Sizing Guide in the
Performance and Scalability area of SAP Service Market Place.

Sizing SAP BusinessObjects Mobile
The SAP BusinessObjects Mobile solution allows end users to access the SAP BI documents (Web
intelligence, Crystal Reports, etc. ) through their mobile devices. It is suited for ad hoc query,
reporting and analysis.
The SAP BusinessObjects Mobile backend services run within the Web application server (typically
Tomcat) in the BI4 environment. The Mobile server provides alternate rendering and workflows for BI in
ways that are optimized for mobile devices.
When sizing for deployments that include Mobile BI, follow the sizing methodology outlined in this
document. Be certain to take into account the load all users will put on the system, including mobile
users. The load created by a mobile user needs to be considered in the same way as a regular desktop
user: Information Consumers, Business Users and Experts need to be identified and their load accounted
for in the sizing exercise.
Depending on the portion of mobile users vs. desktop users, you may want to increase the memory and
CPU resources allocated to the Web application server in order to ensure any increase in processing
required by the SAP BusinessObjects Mobile server is accounted for.
Sizing SAP BusinessObjects Design Studio
Sizing information for Design Studio can be obtained on the SAP Service Market Place in SAP Note
1177020.

Appendix
BI4 P&R Benchmark Testing Details
The following information describes the P&R test environment used to generate the Sizing benchmark
data.
Test report documents and data characteristics
The following table is a summary of the data volume returned for the various documents in the sizing
tests. For more information about the complexity and design of the documents see the document-
specific sections further down. There is a separate section later on, which describes not only the data
volume returned as the result of a query, but also the total size and structure of the data as it existed in
the reporting database.
Product Small Medium Large
Crystal Reports (for Enterprise and 2011) 1,000 rows 10,000 rows 100,000 rows
Dashboards 25 rows 250 rows 2,500 rows
Web Intelligence 100,000 rows 250,000 rows 500,000 rows
Analysis, edition for OLAP 35,000 cells 89,500 cells 375,000 cells
Crystal Reports for Enterprise and Crystal Reports 2011
Information about the composition and complexity of the Crystal Reports documents used in the sizing
tests is summarized below.
Simple Elements Complex Elements
Total Field Text Total Picture Total Chart Crosstab Total
Report Header 4 3 1 0 0 0 0 0 4
Page Header 6 1 4 0 1 0 0 0 6
Group Header 1 7 1 3 3 0 2 2 0 9
Group Header 2 7 1 3 3 0 1 1 0 8
Group Header 3 3 1 1 1 0 0 0 0 3
Group Header 4 5 3 1 1 0 1 0 1 6
Body 3 2 1 0 0 0 0 0 3
Group Footer 4 6 2 3 1 0 0 0 0 6
Group Footer 3 2 0 1 1 0 0 0 0 2
Group Footer 2 2 0 1 1 0 0 0 0 2
Fields in query result set: 23
Tables joined: 4
Record selection clauses: 4
Parameters/Prompts: 4
Interactive filter clauses: 0

Group Footer 1 3 3 0 0 0 0 0 0 3
Report Footer 15 0 8 7 0 0 0 0 15
Page Footer 3 1 2 0 0 0 0 0 3
Total 66 18 29 18 1 4 3 1 70
Web Intelligence
The Web Intelligence documents used in the sizing tests have been summarized below, with respect to
their composition and complexity.
Fields in query result set: 9
Tables joined: 8
Record selection clauses: 2
Parameters/Prompts: 3
Report filter clauses: 0
Simple Elements Complex Elements
Total Cell Total Chart Table Total
Blank Report
Page Header 0 0 0 0 0 0
Body 5 5 0 0 0 5
Page Footer 0 0 0 0 0 0
Revenue_by_part…
Body 1 1 0 0 0 1
Section 1 7 7 1 1 0 8
Section 2 1 1 1 0 1 2
Revenue_by_nation…
Body 1 1 0 0 0 1
Section 1 0 2 1 1 2
Total 15 15 4 2 2 19

Analysis, edition for OLAP
The information below represents the typical Analysis Workspace used in the sizing tests.
Dashboards
Dashboards like the following were used in the sizing tests. There were 2 queries issued every time the
Dashboard was refreshed.

Report data description
This table represents total data set sizes. Other tables earlier in this document show the data volume
returned from a query which is always a subset of the entire dataset. Crystal Reports for Enterprise,
Crystal Reports 2011, and Dashboards tests were executed against DS2 and DS3 on a relational
database. Web Intelligence tests were executed against DS1, also on a relational database.
Data Set LINEITEM ORDERS PARTSUPP PART CUSTOMER SUPPLIER NATION REGION
DS1 367.9 MB
3M rows
81.5 MB
750K rows
55.75 MB
400K rows
12.9 MB
100K rows
11.6 MB
75K rows
0.75 MB
5K rows
0.01 MB
25 rows
< 0.01 MB
5 rows
DS2 72.75 MB
400K rows
18.3 MB
100K rows
13.7 MB
53K rows
3.1 MB
13K rows
2.9 MB
10K rows
0.2 MB
650 rows
0.01 MB
25 rows
< 0.01 MB
5 rows
DS3 4.8 MB
40K rows
1.1 MB
10K rows
0.75 MB
5K rows
0.2 MB
1K rows
0.2 MB
1K rows
0.02 MB
65 rows
0.01 MB
25 rows
< 0.01 MB
5 rows
Analysis, edition for OLAP was tested against an SAP BW data source. Information about the dataset
used in the sizing tests is below. The average BEx query used had 4 characteristics, 5 hierarchies, 3 key
figures, and 0 variables.
Dimension Hierarchy / Characteristic Members Levels
Material Material 100K 1
Material BMK_HIER 100K 6
Material MAT_N2_HIGH_LARGE 60K 11
Material MAT_N6_LINK_NODES 60K 11
Sold-to Party Sold-to Party 100K 4
Sold-to Party CUST_N1_FLAT_LARGE 100K 4
Sold-to Party CUST_N5_UNBALANCED 100K 8
Sales Organization Sales Organization 1K 1
Sales Organization SALESORG_N3_SMALL 500 4
Calendar Day Calendar Day 365 1
Calendar Day QuaDay 365 2
Calendar Day QuaMonDay 365 3
Country of Sold-to Party Country of Sold-to Party 21 1
Calendar Year/Month Calendar Year/Month 12 1
Product Hierarchy Product Hierarchy 10 1
Division Division 9 1

CMS repository data description
Info Object Type # of Info Objects
Folders 686
Documents - WebI 775
Documents – Crystal Reports 29
Documents - Dashboards 3
Documents – Analysis, edition for OLAP 8
Inboxes 604
Personal Categories 604
Users 604
Groups 12
Deployment Files 3048
Others 732
Application Objects 927
Total 8020
Test hardware specifications
Machine CPU
(SAPS)
CPU
Arch
RAM
(GB)
Disk W
(MB/s)
Disk R
(MB/s)
Network
M1 - BusinessObjects BI 4 23,000 x64 48 852 1170 10 Gbit
M2 - Reporting DB (Relational) 23,000 x64 48 863 1120 10 Gbit
M3 - Reporting DB (SAP BW 1) 23,000 x64 48 955 1252 10 Gbit
M4 - Reporting DB (SAP BW 2) 23,000 x64 24 891 1192 1 Gbit
M5 - CMS Repository DB 23,000 x64 48 974 1219 10 Gbit
M6 - Load Generator 12,000 x64 32 25 616 1 Gbit

Document Update History
Date Details
June, 2013 Sizing methodology reworked to be easier to follow as well as more comprehensive.
September, 2013 Additional tuning suggestions added (Apache Split, BI Pattern Books). BI 4.1 sizing
added.
December, 2013 Added (restored) Analysis OLAP tuning, enhanced Scheduling and Publishing
guidance, Resizing guidance, added audit database reporting guidance to the
Monitoring section.
Added sizing checklists.
February 2014 Added Connection Server tuning: JVM memory settings for JDBC connectivity
including HANA.
Updated WebI suggested maximum connections, CMS sizing guidance, importance
of CMS database tuning. Added a CMS Database Tuning section.
Updated notice about multiple Lifecycle Management instances: supported starting
with BI 4.1 SP3.
Added a link to Design Studio sizing. Updated link to the PAM for BI 4.1.
Updated Apache Split links to new content for BI 4.1 and Apache 2.4.

Copyright
© Copyright 2014 SAP AG. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or for any purpose without the express permission of SAP AG.
The information contained herein may be changed without prior notice.
Some software products marketed by SAP AG and its distributors contain proprietary software components of other software vendors.
Microsoft, Windows, Excel, Outlook, and PowerPoint are registered trademarks of Microsoft Corporation.
IBM, DB2, DB2 Universal Database, System i, System i5, System p, System p5, System x, System z, System z10, System z9, z10, z9,
iSeries, pSeries, xSeries, zSeries, eServer, z/VM, z/OS, i5/OS, S/390, OS/390, OS/400, AS/400, S/390 Parallel Enterprise Server,
PowerVM, Power Architecture, POWER6+, POWER6, POWER5+, POWER5, POWER, OpenPower, PowerPC, BatchPipes,
BladeCenter, System Storage, GPFS, HACMP, RETAIN, DB2 Connect, RACF, Redbooks, OS/2, Parallel Sysplex, MVS/ESA, AIX,
Intelligent Miner, WebSphere, Netfinity, Tivoli and Informix are trademarks or registered trademarks of IBM Corporation.
Linux is the registered trademark of Linus Torvalds in the U.S. and other countries.
Adobe, the Adobe logo, Acrobat, PostScript, and Reader are either trademarks or registered trademarks of Adobe Systems
Incorporated in the United States and/or other countries.
Oracle is a registered trademark of Oracle Corporation.
UNIX, X/Open, OSF/1, and Motif are registered trademarks of the Open Group.
Citrix, ICA, Program Neighborhood, MetaFrame, WinFrame, VideoFrame, and MultiWin are trademarks or registered trademarks of
Citrix Systems, Inc.
HTML, XML, XHTML and W3C are trademarks or registered trademarks of W3C®, World Wide Web Consortium, Massachusetts
Institute of Technology.
Java is a registered trademark of Sun Microsystems, Inc.
JavaScript is a registered trademark of Sun Microsystems, Inc., used under license for technology invented and implemented by
Netscape.
SAP, R/3, SAP NetWeaver, Duet, PartnerEdge, ByDesign, SAP Business ByDesign, and other SAP products and services mentioned
herein as well as their respective logos are trademarks or registered trademarks of SAP AG in Germany and other countries.
Business Objects and the Business Objects logo, BusinessObjects, Crystal Reports, Crystal Decisions, Web Intelligence, Xcelsius, and
other Business Objects products and services mentioned herein as well as their respective logos are trademarks or registered
trademarks of Business Objects S.A. in the United States and in other countries. Business Objects is an SAP company.
All other product and service names mentioned are the trademarks of their respective companies. Data contained in this document
serves informational purposes only. National product specifications may vary.
These materials are subject to change without notice. These materials are provided by SAP AG and its affiliated companies ("SAP
Group") for informational purposes only, without representation or warranty of any kind, and SAP Group shall not be liable for errors or
omissions with respect to the materials. The only warranties for SAP Group products and services are those that are set forth in the
express warranty statements accompanying such products and services, if any. Nothing herein should be construed as constituting an
additional warranty.

Bo sizing guide

More Related Content

What's hot (13)

Viewers also liked (8)

Similar to Bo sizing guide (20)

Recently uploaded (20)

Bo sizing guide