Ibm power570 web_sphere_7_ net_benchmark_winsrv2008

Microsoft .NET Framework 4.0 vs. IBM
®

WebSphere 7 StockTrader Benchmark Report
®

7/2/2010
© Microsoft Corporation 2010

.NET StockTrader Technical Documentation Page 1

This document supports the release of Windows Server® 2008 R2 and the Microsoft .NET Framework 4.0.

The information contained in this document represents the current view of Microsoft Corp. on the issues disclosed as of
the date of publication. Because Microsoft must respond to changing market conditions, this document should not be
interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any
information presented. This document is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS
OR IMPLIED, IN THIS DOCUMENT.

Information in this document, including URL and other Internet Web site references, is subject to change without notice.
Unless otherwise noted, the companies, organizations, products, domain names, e-mail addresses, logos, people, places and
events depicted in examples herein are fictitious. No association with any real company, organization, product, domain name,
e-mail address, logo, person, place or event is intended or should be inferred. Complying with all applicable copyright laws is
the responsibility of the user. Microsoft grants you the right to reproduce this guide, in whole or in part.

Microsoft may have patents, patent applications, trademarks, copyrights or other intellectual property rights covering subject
matter in this document, except as expressly provided in any written license agreement from Microsoft, the furnishing of this
document does not give you any license to these patents, trademarks, copyrights or other intellectual property.

© 2010 Microsoft Corp. All rights reserved.

Microsoft, Windows Server, the Windows logo, Windows, Active Directory, Windows Vista, Visual Studio, Internet Explorer,
Windows Server System, Windows NT, Windows Mobile, Windows Media, Win32, WinFX, Windows PowerShell, Hyper-V, and
MSDN are trademarks of the Microsoft group of companies.

The names of actual companies and products mentioned herein may be the trademarks of their respective owners.

Contents
Executive Summary....................................................................................................................................... 3
Introduction .................................................................................................................................................. 5
.NET StockTrader Sample Application and Performance Kit .................................................................... 5
Multiple Clients with Open Integration to Middle Tier via WCF............................................................... 6
Full Disclosure Notice................................................................................................................................ 6
Middle-Tier Hardware Platforms Tested ...................................................................................................... 7
IBM Power 570 (based on Power6 Architecture) running IBM AIX 5.3 .................................................... 7
Hewlett-Packard C7000 Blade System with 4 HP ProLiant BL460c blades ............................................... 7
Test Methodology and Reporting Metrics .................................................................................................... 9
Web Application Benchmark .................................................................................................................... 9
Trade Middle Tier Web Service Benchmark and WSTest Web Service Benchmark ............................... 10
Running the Benchmarks and Capturing Reported Metrics ................................................................... 10
.NET Framework 4.0 vs. .NET Framework 3.5 ......................................................................................... 11
Fair Benchmark Comparisons between .NET StockTrader and IBM WebSphere 7 Trade .......................... 12
Database Access Technology/Programming Model ............................................................................... 12
Caching .................................................................................................................................................... 12
Enable Long Run Support ........................................................................................................................ 13
Database Load......................................................................................................................................... 13
Database Configuration .......................................................................................................................... 14
Test Methodology for the Trade Web Application Benchmark .............................................................. 14
Test Scripts .......................................................................................................................................... 14
Simulated User Settings ...................................................................................................................... 15
IBM HTTP Server vs. Port 9080 ........................................................................................................... 16
Web Application Pages Exercised by the Test Scripts......................................................................... 16
Test Methodology for the Trade Web Services Benchmark and WSTest Web Service Benchmark ....... 17
Test Scripts .......................................................................................................................................... 17
IBM HTTP Server vs. Port 9080 ........................................................................................................... 17
Benchmark Results...................................................................................................................................... 18
Trade Web Application Test.................................................................................................................... 18
Trade Web Application Benchmark Discussion ...................................................................................... 19

IBM WebSphere 7/Power6 vs. Microsoft .NET 4.0/HP BladeSystem Benchmark Report Page 1

Trade Middle Tier Web Services Test ..................................................................................................... 21
Trade Middle Tier Web Services Benchmark Discussion ........................................................................ 22
WSTest Benchmark ................................................................................................................................. 24
WSTest Benchmark Discussion ............................................................................................................... 25
Conclusion ................................................................................................................................................... 27
Appendix A: Pricing ..................................................................................................................................... 28
Hardware and Operating System Pricing ................................................................................................ 28
IBM Power 570.................................................................................................................................... 28
Hewlett Packard BladeSystem C7000 ................................................................................................. 31
Pricing for the Application Server Used in the Tests .............................................................................. 32
IBM WebSphere 7/IBM Power 570 ..................................................................................................... 32
IBM WebSphere 7/ Hewlett Packard BladeSystem C7000 ................................................................. 32
Total Hardware, Operating System and Application Server Costs for IBM WebSphere 7/AIX on IBM
Power 570 as tested................................................................................................................................ 33
Total Hardware, Operating System and Application Server Costs for IBM WebSphere 7/Windows
Server 2008 on Hewlett Packard BladeSystem C7000 as tested ............................................................ 33
Total Hardware, Operating System and Application Server Costs for Microsoft .NET/Windows Server
2008 on Hewlett Packard BladeSystem C7000 as tested ....................................................................... 33
Appendix B: Tuning Parameters................................................................................................................. 34
WebSphere Tuning – IBM Power 570/AIX .............................................................................................. 34
IBM HTTP Server Tuning AIX ............................................................................................................... 34
WebSphere Tuning – Windows Server 2008 R2 ..................................................................................... 35
IBM HTTP Server Tuning Windows Server 2008 R2 ............................................................................ 36
.NET Tuning ............................................................................................................................................. 36


Executive Summary
This paper presents updated benchmark results based on extensive performance and scalability testing
of Microsoft .NET Framework 4.0 using the .NET StockTrader benchmark workload as well as the WSTest
workload. The .NET 4.0 framework benchmark results are compared to the equivalent Java Enterprise
workloads running on IBM WebSphere 7, both on a 64-bit Power6/AIX server and on Windows Server
2008 R2. Both the peak throughput rate and the relative cost of each platform are compared, so that
customers can better understand both the performance and cost characteristics of the platforms tested.
All tests on Windows Server were conducted on an HP BladeSystem 7000 running Windows Server 2008
R2 release (64-bit), with the final RTM release of the .NET Framework 4.0 (64-bit) which is downloadable
for free on MSDN. The IBM WebSphere/Java tests were conducted using the latest IBM WebSphere 7.0
release, 64-bit. The entire benchmark kit, including .NET source code and Java source code for the
StockTrader and WSTest workloads is downloadable for free at http://msdn.microsoft.com/stocktrader.
This kit also includes a Capacity Planner/multi-agent benchmark tool that customers can use to perform
their own testing against .NET 4.0, IBM WebSphere and Oracle WebLogic.

Platforms Tested

 IBM Power 570 with IBM WebSphere 7 and AIX 5.3
 Hewlett Packard BladeSystem C7000 with IBM WebSphere 7 and Microsoft Windows
Server 2008
 Hewlett Packard BladeSystem C7000 with Microsoft .NET and Windows Server 2008

The table below summarizes the results of the testing, including the cost of each platform as
tested. Detailed pricing data is included in Appendix A.


IBM Power 570 Hewlett Packard Hewlett Packard
with WebSphere BladeSystem BladeSystem
7 and AIX 5.3 C7000 with C7000 with
(8 cores) WebSphere 7 and Microsoft .NET
Windows Server 4.0 and Windows
2008 R2 Server 2008 R2
Total Hardware +
Operating System $215,728.08 $50,161.00 $50,161.00
Costs
Middle Tier
Application Server $44,400.00 $37,000.00 $0.00
Licensing Costs
Total Middle Tier
System Cost as $260,128.08 $87,161.00 $50,161.00
Tested
Trade Web
Application 8,016 transactions 11,121 transactions 11,020 transactions
Benchmark per second per second per second
Sustained Peak TPS
Trade Middle Tier
Web Service 10,571 transactions 14,468 transactions 22,316 transactions
Benchmark per second per second per second
Sustained Peak TPS
WSTest EchoList
Test Sustained 10,536 transactions 15,872 transactions 23,772 transactions
Peak TPS per second per second per second


Introduction
To perform the comparative analysis for these three application server platforms, the performance
of three core workloads is examined on each platform tested:

1. Data-driven Web application. This comparison is based on a WebSphere 7 implementation
of the IBM Trade performance benchmark application compared to the .NET StockTrader
4.0 benchmark application based on the .NET Framework 4.0.
2. Middle-tier Web Service benchmark with middle-tier business logic and database
transactions. This comparison is based on a WebSphere 7 implementation of the IBM
Trade performance benchmark application compared to the .NET StockTrader 4.0
benchmark application, testing the middle tier Web Service layer using a Web-Service
based benchmark client.
3. Web Service Benchmark based on WSTest. This benchmark uses the WSTest 1.5
benchmark workload (as originally defined by Sun Microsystems) to focus on just raw Web
Service performance, with no business logic or middle tier data access. As such, this test
focuses on just the JAX-WS performance of IBM WebSphere 7 compared to the
performance of the equivalent .NET 4.0 Windows Communication Foundation (WCF) Web
Services.

.NET StockTrader Sample Application and Performance Kit
All sources are downloadable from MSDN as part of the .NET StockTrader sample application,
available at http://msdn.microsoft.com. This download includes:

 The IBM WebSphere 7-optimized Trade benchmark application and Java source code.
 The IBM WebSphere 7-optimized WSTest 1.5 benchmark and Java source code based on
JAX-WS.
 The .NET StockTrader 4.00 benchmark application and source code based on ASP.NET and
WCF 4.0.
 The .NET implementation of the WSTest 1.5 benchmark and source code based on WCF
4.0.
 The Capacity Planner multi-agent benchmark tool, including source code, which enables
customers to test the Web Service workloads (both for Trade and for WSTest) on their
own, using the hardware/software platform of their choosing.

For the benchmark comparisons using the IBM WebSphere Trade application (originally created by
IBM as a best-practice performance application for IBM WebSphere) , Microsoft created an
application that is functionally equivalent to the IBM WebSphere Trade application, both in terms
of user functionality and middle-tier database access, transactional and messaging behavior. This
application was created using best-practice programming techniques for .NET and the Microsoft
Application Development platform. The resulting application, the .NET StockTrader, is now
published on MSDN as a best-practice .NET enterprise application at
http://msdn.microsoft.com/stocktrader. The .NET StockTrader is a service-oriented application that


utilizes Windows Communication Foundation (WCF) for its underlying remoting and messaging
architecture. The user interface is an ASP.NET/C# Web application that is equivalent to the IBM
WebSphere Trade Java Server Pages (JSP) application. Additionally, the middle-tier services,
written in C#, mirror the functionality and transactional characteristics of the backend IBM
WebSphere Trade services which are based on JAVA EE 5 and the IBM WebSphere 7 application
server. Microsoft also modified the IBM WebSphere Trade 6.1 application to take advantage of
IBM WebSphere 7, including the use of a streamlined JSP/Servlet/JDBC architecture (vs. EJBs), and
the latest JAX-WS Java Enterprise Web Services software stack.

For the benchmark comparisons using the WSTest workload, Microsoft created the equivalent Web
Services using IBM WebSphere 7 and JAX-WS; as well as a C# implementation based on Microsoft
.NET and WCF Web Services. These sources are also included in the .NET StockTrader download on
MSDN.

Multiple Clients with Open Integration to Middle Tier via WCF
As a service-oriented application based on WCF, multiple interoperability scenarios are enabled
with the .NET StockTrader. Since both the IBM WebSphere 7-based Trade application and the .NET
StockTrader application expose their middle-tier services as industry-standard Web Services, the
two applications can be seamlessly integrated with no code changes required. The JSP Trade front-
end application can fully utilize the .NET middle tier services and messaging capabilities of the .NET
StockTrader; and the ASP.NET StockTrader front-end application can fully utilize the Java-based
WebSphere 7 Trade middle tier services and messaging capabilities. This interoperability is possible
with the .NET StockTrader since WCF, Microsoft’s component remoting and distributed application
programming model, is fundamentally based on open Web Service standards including SOAP, XML
and the latest WS-* industry standards. In addition to the ASP.NET Web based application that
integrates via services with the middle tier, the sample also includes a Windows Presentation
Foundation (WPF) desktop client (also developed in C#), that provides a smart-client interface to
the middle tier. The WPF client can also seamlessly connect to either .NET middle tier services, or
the IBM WebSphere 7 Trade middle tier services simply by changing the services URL in the
configuration page—no code changes are required to achieve this interoperability.

Full Disclosure Notice
The complete source code, test script flow and all testing methodology for this benchmark are available
online. Any reader may download and view the actual code for all implementations tested, and may
further perform the benchmark for themselves to verify the results. The benchmark kit can be
downloaded from http://msdn.microsoft.com/stocktrader. Extensive time was taken to generate
results that represent optimal tuning for the platforms tested, and we are quite confident in the results.
We encourage customers to download each kit and perform their own comparative testing and
functional and technical reviews of each workload tested.


Middle-Tier Hardware Platforms Tested
A key goal of the testing was to document the performance and relative cost of IBM WebSphere 7
running on an IBM Power 570 server (Power6-based) vs. IBM WebSphere 7 on a Hewlett Packard
BladeSystem (Intel-based) running Microsoft Windows Server 2008 R2; and then to compare these
results to the identical workloads running on the HP BladeSystem with Microsoft .NET and Windows
Server 2008 R2. Complete pricing details, including hardware and software costs, are documented in
Appendix A of this paper. For the first two workloads, which include database access, each platform was
tested against the same database hardware and disk configuration, which was carefully monitored to
ensure the databases were never a bottleneck for any middle-tier platform tested. In addition, network
monitoring was carefully conducted to ensure the network was never an artificial bottleneck. Hence,
the performance measured accurately captures the peak TPS rate achieved on the middle tier hardware
and software platform tested. In all cases, server CPU saturation rates were at or above 96% CPU
utilization at peak TPS, reflecting the fact no external bottleneck (database, network) artificially
limited the results achieved.

Additionally, in all testing, database CPU processor utilization remained below 50%, with no significant
disk queuing.

IBM Power 570 (based on Power6 Architecture) running IBM AIX 5.3
 8 IBM Power6 cores, operating at 4.2GHz
 32 GB RAM
 AIX 5.3
 4 x 1 GB NICs

Hewlett-Packard C7000 Blade System with 4 HP ProLiant BL460c blades
 4 Hewlett Packard ProLiant BL460c blades
 Each blade has one Quad-Core Intel® Xeon® E5450 (3.00GHz, 1333MHz FSB, 80W) Processor
 Each blade has 32 GB RAM
 Each blade is running Microsoft Windows Server 2008 R2/64-bit
 Each blade has 2 x 1 GB NICs

For the Trade Web Application benchmark and the Trade Web Services benchmark, two separate
databases are utilized to ensure enough database and disk I/O capacity to handle the large volume of
transactions each platform is able to produce.

 For the IBM WebSphere 7 on IBM Power 570 platform tests for the Web application and Trade
Web Services test, two instances of IBM WebSphere (nodes) are running on the IBM Power 570
server, each configured to communicate with a separate and dedicated backend trade database.
IBM DB2 v9.5 Enterprise Edition (64-bit, Windows Server) is used for each of these databases,
configured on identical hardware (see the database section for details).
 The WSTest benchmark does not utilize a database; a single instance of IBM WebSphere 7 was
used on the Power 570 server and on each HP Blade.


 For the IBM WebSphere 7 on HP BladeSystem C7000 platform tests (Windows Server 2008 R2),
one instance of IBM WebSphere is running per blade, with two blades communicating to one of
the DB2 databases, and two blades communicating to the other dedicated IBM DB2 database.
 For the Microsoft .NET/Windows Server 2008 platform tests on the HP BladeSystem C7000, a
single .NET 4.0 CLR instance is running on each blade. Two servers are communicating to one
database, two servers to the second database. The databases used for the .NET tests are
Microsoft SQL Server 2008, 64-bit, on Windows Server 2008 R2. They run on the exact same
hardware (servers, disks) as the IBM DB2 9.5 databases used in the WebSphere 7 tests.


Test Methodology and Reporting Metrics
For each result, we report a peak sustained transaction per second (TPS) throughput rate as averaged
over a 30 minute measurement period as tracked by the testing tool. For the Web Application
performance test, HP Mercury LoadRunner was used to drive load and capture the results. For the two
Web Service workloads (Trade Web Services and WSTest); the .NET Capacity Planner (provided for free
with the .NET StockTrader download) was used to drive load and capture results.

Web Application Benchmark
In the Mercury LoadRunner test cases, 32 physical client machines (1.2 GHz, 512MB RAM) were used to
run the test scripts, each client driving several hundred distinct users with a one-second think time.

Figure 1: Test bed used for the Web Application benchmark using IBM WebSphere 7 Trade and Microsoft .NET StockTrader
Web applications. There are two distinct hardware platforms that make up the System Under Test (SUT); however, the HP
BladeSystem runs both the IBM WebSphere 7 Trade workload, and the Microsoft .NET StockTrader workload; representing
three total SUTs. The databases are configured with plenty of capacity for all systems tested, and are not part of the SUTs.


Trade Middle Tier Web Service Benchmark and WSTest Web Service Benchmark
In the .NET Capacity Planner Web Service tests, 10 distinct clients were used to drive load (Intel quad-
core systems @2.4 GHz) with a 0.1 second think time.

Figure 2: Test bed used for the two Web Service benchmarks. Note the databases are only utilized for the Trade Web Service
test; as the WSTest benchmark involves no business logic or database access. There are two distinct hardware platforms that
make up the System Under Test (SUT); however, the HP BladeSystem runs both the IBM WebSphere 7 Trade and Java
WSTest workloads, and the Microsoft .NET StockTrader and .NET WSTest workloads; representing three total SUTs for each
of the two Web Service tests. The databases are configured with plenty of capacity for all systems tested, and are not part of
the SUTs.

Running the Benchmarks and Capturing Reported Metrics
Load balancing is achieved across the 4 HP blades (for both .NET and IBM WebSphere 7) using Windows
Network Load Balancing, built into Microsoft Windows Server 2008. In all cases, users are added to the
system with iterative test runs until peak throughput is obtained. Extensive benchmark runs were done


prior to measurement to ensure proper tuning of the middle tier systems. For each of the three
workloads, and across all SUTs tested, we report the peak sustained TPS rate, and also a calculated
dollar cost per TPS (price performance metric) so customers can better understand what they are
paying for that performance using a standard normalized metric: cost of platform divided by peak TPS
rate measured. This calculation mirrors how testing organizations such as the Transaction Processing
Council (TPC) report price performance metrics.

The cost calculations are based on the middle tier application server software licensing costs and
hardware acquisition costs of the middle tier systems as tested. Notes and details on the pricing of the
middle tier application servers are included in the Appendix, and prices are based on published pricing
from each vendor or as purchased from authorized resellers.

Customers should understand that full .NET capabilities are included in every edition of Windows Server
2008 R2, and upgraded versions of .NET are made available for free download from MSDN (for example,
.NET 4.0 as tested here). Hence, there is no additional or separate application server cost associated
with a .NET application; while commercial JAVA EE 5 application servers such as WebSphere are
separately licensed (and typically quite expensive) products. The dollar cost per TPS rates might,
therefore, surprise some readers.

It is also important to remember that this is a test of specific workloads based on the IBM WebSphere
Trade application and equivalent .NET StockTrader application; and the Java and .NET WSTest
workloads. Each of these workloads, however, utilizes most (if not all) of the commonly deployed
architectural building blocks used in common enterprise application scenarios.

.NET Framework 4.0 vs. .NET Framework 3.5
Finally, the .NET results presented here are for the current .NET Framework 4.0 version running on
Windows Server 2008 R2 with IIS 7.5. For all tests, however, benchmark runs were also completed on
.NET Framework 3.5, the previous version of the .NET Framework, also running on Windows Server 2008
R2. All .NET 4.0 results in these tests were equivalent in performance (with <5% difference in any test)
to the .NET Framework 3.5 results.


Fair Benchmark Comparisons between .NET StockTrader and
IBM WebSphere 7 Trade
Since each application supports many different configurations, it is very important to understand what
constitutes a fair comparison. The configurations compared must be equivalent. This means that the
applications must produce the exact same functionality and processing behavior in the configurations
compared. You cannot, for example, compare one application running with caching and the other with
no caching. The .NET StockTrader, while based on .NET and not JAVA EE 5, was designed to mirror most
of the IBM Trade configurations possible with this testing goal in mind. The WebSphere 7-optimized
Trade application, with Java source provided with the .NET StockTrader download, runs in the fastest
configuration possible: Java Server Pages, Servlets, and straight JDBC database access with no Enterprise
Java Beans.

Database Access Technology/Programming Model
WebSphere 7 Trade:

 JDBC

The JDBC mode eliminates the Entity Beans and CMP, and instead uses direct JDBC calls to the database.
Java model classes are used to pass data between the logical tiers in the application.

.NET StockTrader:

 ADO.NET

The ADO.NET implementation uses C# model classes to pass data between tiers. It uses ADO.NET
DataReaders isolated in a separate data access layer (DAL) as a best-practice performance programming
practice, and to maintain clean separation of database logic from the other tiers of the application.

Caching
WebSphere 7 Trade

 WebSphere is configured for Servlet Caching, and only the Market Summary JSP fragment is
cached. The cache is renewed on a 3 minute cycle (absolute expiration), as defined in the
WebSphere Cachespec.xml file. All other elements of the application are not cached, database
interactions occur on every request. This fragment caching is necessary, since the Market
Summary SQL query is very heavyweight, and with any real data load, would quickly bring the
database and the application to a crawl if run on every visit to the home page. Since the query
is the same in the .NET StockTrader application, the same is true for .NET StockTrader. For
benchmark runs, we tuned to cache to a 3 minute absolute expiration in both cases.

.NET StockTrader

The .NET StockTrader uses the .NET cache API (output caching specifically) to cache the Market
Summary control (fragment) for 3 minutes (absolute expiration). We chose not to implement


further caching in the application because it is simply not realistic. While WebSphere Trade can
cache stock quotes, account data, portfolio data and the like (and .NET StockTrader could too if
implemented), the simple fact is that this is not a realistic approach or “cache policy” for this
application. Consider that the IBM cache, while distributed (it can keep cached items in sync
across clustered servers), is not invalidated by the data source itself. Hence, an update to a
database table by any other application using the same database would result in possibly
corrupt data, or at least presenting incorrect “stale” data to users in the application. Unless a
customer is willing to direct all database updates/deletes/inserts through the same WebSphere
instances, data on the middle tier would quickly become out of sync with the actual database.
In other words, such a strategy keeps the organization from building new applications on other
platforms against a common database. Market Summary information is fine to cache, since its
read-only and can stand to be several seconds stale; user account balances and stock price
information used on trades cannot be stale. Hence, the .NET StockTrader does not implement
further caching beyond Market Summary.

Enable Long Run Support
IBM WebSphere 7 Trade

 Off

This setting was introduced to Trade with WebSphere Trade 6.1, apparently because customers running
the benchmark for long periods of time saw steady degradation of performance as user accounts started
to contain more and more orders. This resulted in heavier queries, and large and steadily increasing
amounts of data being passed between tiers and formatted into the account summary page. Hence,
IBM created this setting. However, since our test database is loaded with many more user accounts,
neither the WebSphere 7 version of Trade or the .NET StockTrader have an issue in running benchmarks
over extended periods of time under heavy load.

.NET StockTrader

There is no equivalent setting for .NET StockTrader; instead we ran the benchmark for both applications
against a larger, much more realistic data load (500,000 accounts, 5 order per account, 100,000 quotes).
We provide a database loader (written in .NET Windows Forms) to load both SQL Server 2008, Oracle
11G, and DB2 V9.5. This loader program will also reset the database between benchmark runs to the
same starting state by deleting added data records.

Database Load
As discussed, we used a default load for all benchmark runs (reset between runs) of 500,000 accounts, 5
orders/holdings per account, and 100,000 quotes. This is much more realistic than the IBM default
settings for Trade.


Database Configuration
The IBM WebSphere 7 Trade application was tested against an all-IBM configuration, using IBM DB2
V9.5 (Enterprise Edition) as the backend databases, and the latest IBM DB2 V9.5 JDBC drivers for data
access. The .NET StockTrader was tested against an all-Microsoft setup, with backend SQL Server 2008
databases (Enterprise Edition). The benchmark is not a database benchmark: enough capacity was
employed for the database hardware to ensure it was not a bottleneck in any benchmark run. For both
the Trade Web Application performance benchmark and the Trade Web Services benchmark, two
backend databases were deployed to eliminate possible database contention under high TPS rates. Each
database was deployed to the same two 64-bit Windows Server 2008 servers. Storeage is a high-
capacity, high-speed HP StorageWorks EVA 4400 Disk Array with a total of 96 15K drives in a RAID 5
configuration.

Database Server Specification

2 x HP BL680c G5 blades

 Each blade configured with 4 Quad-Core Intel XEON CPUs, @2.4GHZ (16 cores in each blade)
 Each blade configured with 64 GB RAM
 Each blade configured with 4 x 1GB NICs
 Each blade running IBM DB 9.5 Enterprise Edition 64-bit and Microsoft SQL Server 2008 64-bit
 Each blade running Microsoft Windows Server 2008 64-bit, Enterprise Edition
 Each blade has 2 4GB HBAs for fiber/sans access to the EVA 4400 storage.

Database Storeage Specification

 HP StorageWorks EVA 4400 Disk Array
 96 15K drives total
 4 logical volumes consisting of 24 drives each
 Database server 1: Logical Volume 1 for logging
 Database server 1: Logical Volume 2 for database
 Database server 2: Logical Volume 3 for logging
 Database server 2: Logical Volume 4 for database

All tuning steps were followed for DB2 according to the IBM Trade 6.1 documentation, however,
additional logging space was configured given the larger data load. The equivalent drive space was
configured for SQL Server logging. The database disk usage was closely monitored to ensure each run
could complete without requiring the database to extend the logging or data file space during a
benchmark run. This is an important consideration for custom tests.

Test Methodology for the Trade Web Application Benchmark

Test Scripts
For the Web Application test, Mercury LoadRunner was used to record test scripts—browser
interactions that exercise most of the functionality in the application. These were run across 32 client


machines (500 MHz Windows XP desktops with 512 MB RAM). User agents were configured to run with
a one second think time between each request. Each benchmark run included a warm up run to get to
steady state, and a 30 minute measurement period. TPS rates were determined by LoadRunner by
averaging across the 30 minutes. Error rates were monitored to ensure they remained at less than .01%
during the measurement period. User loads were run for each application up to a number that
represented peak throughput for that configuration, as determined in many iterative runs (literally
hundreds) during the tuning stages. Extensive time was spent tuning IBM WebSphere (see the
appendix) to achieve peak throughput for the software/hardware configuration tested. IBM does not
publish pre-set tuning guides for Trade, and developers must iteratively test and tune the various knobs
in WebSphere (there are many) to get to an optimal setup for a given hardware configuration and
software workload. .NET does not require nearly as much tuning, and in general will scale quite well out
of the box, given a properly coded application. Some tuning was applied, however, and this is
documented in detail along with the WebSphere tuning in the appendix.

In general, it is fairly straightforward to recognize an in-properly tuned system for both platforms. Given
enough database capacity and network capacity, each application server running under load should be
able to reach close to full saturation (~100% CPU saturation) when properly tuned, assuming good
vertical scaling of the technology/OS across processors, and no external bottlenecks. It is a requirement
to iteratively test after adjusting, individually, the core tuning settings for the application, to determine
peak throughput and user loads that achieve peak throughput (neither over saturating or under
saturating the middle tier server being tested). With WebSphere 7, these tuning settings are extensive,
including several different thread pools, connection pools for databases and queue connection factories,
Java heap sizes, and the like. We are quite confident in the results and the system tuning applied, as in
all tests servers under test achieved between 96% and 100% CPU saturation at peak TPS. However,
should IBM recommend different (specific) settings, we will be happy to re-run the benchmark and
publish new results with these settings. Customers can also run the benchmark—it is a great way to
really judge the capacity of the two platforms for a realistic workload, and then to judge the cost of each
platform and compare the cost to the results achieved.

Simulated User Settings
Mercury agents were set to not download images (this is not a web server/network I/O benchmark)
during runs. They make requests to the application server, and all processing is completed and just the
HTML returned to the agent. This reduces the overhead on the agent machines and network, and helps
ensure the 32 client machines used in the testing never become an artificial bottleneck. Just as
importantly, the agents were configured to reset connections between iterations, to simulate
constantly new users logging into the application, and more fully exercise the underlying networking
stacks and HTTP keep alive system that the application servers and Web servers use to support large
concurrent user bases. Too many benchmarks are simply run with 10-15 threads, no think times and no
network resets between script iterations. These types of benchmarks often produce very different
(often over-inflated) results than real world usage conditions. Our settings are meant to much more
closely mimic the real world. Think times (even if just one second) and connection resets between
iterations make all the difference here.


IBM HTTP Server vs. Port 9080
When benchmarking IBM WebSphere, it is important to understand that while WebSphere provides an
in-process HTTP listener service (port 9080, by default), IBM best practice recommended deployments
are in conjunction with the full-blown IBM HTTP Server (a repackaged version of Apache). Hence, for all
configurations, we used IBM HTTP Server as packaged with WebSphere, configured with the WebSphere
Plug-in. On the .NET side, we used Microsoft Internet Information Services 7.0 (IIS 7), also co-located
with the server being tested. IBM HTTP Server tuning details for Windows and AIX are included in the
Appendix.

Web Application Pages Exercised by the Test Scripts
The test scripts were designed to drive load on the system in a way that exercises most functionality in
the application, and puts a heavier emphasis on transactions; such as adding new registered users and
buying stocks. The precise flow of the test scripts (exactly the same for all test runs on all platforms) is
listed below. A one second think time was used between all URL requests, and in the results, a
‘transaction’ is defined as the successful completion of the URL request to the server, with valid
response/HTML returned.

Part 1 (Existing User)

 Login random registered user (1 to 500,000 users loaded in database; the login includes in both
apps a redirect to the home page, and all the logic to login and display home page)
 Request random quote (1 to 100,000 distinct quotes loaded in database; one post performed
with 1 stock requested)
 Visit Account Page (no account update performed)
 Visit Portfolio Page
 Visit Home Page
 Logout the Registered user via logout page

Part 2 (New User—run in sequence after Part 1)

 Register a new user/submit registration form (this also logs new user in with redirect/display of
home page)


 Buy a random stock symbol (1 to 100,000 stock symbols in database; buy operation involves a
direct post/submit to the order submission pages, which submit the order for all backend
processing)
processing)
processing)
processing)
 Visit Home Page
 Logout

Test Methodology for the Trade Web Services Benchmark and WSTest Web Service
Benchmark

Test Scripts
For these tests, the .NET Capacity Planner tool (a Web Service load-generating client included with the
.NET StockTrader download) was used to capture average TPS rates over a 30 minute measurement
interval. 10 quad-core systems were used to drive load, each running the agent software. Each agent
was running between 50 and 300 user threads, operating at a 0.1 second think time. Iterative testing
was done to determine user loads that resulted in near 100% server saturation and peak sustained TPS
rates over the 30 minute measurement interval. For the Trade Web Services tests, the Capacity Planner
was set to use the Transaction-Light script, of the three selectable scripts in the tool. For the WSTest
tests, the message size was set to 20.

IBM HTTP Server vs. Port 9080
As with the Trade Web Application benchmark, IBM WebSphere was configured with the IBM HTTP
Server. Similarly, the .NET WCF services were hosted via the IIS 7 web server.


Benchmark Results
Trade Web Application Test

Figure 3: Peak TPS Rates for the Web Application Test. IBM WebSphere 7 Trade is running using JSPs/Servlets with direct
JDBC access to the databases. Microsoft .NET StockTrader is running using ASP.NET/Web Forms and ADO.NET access to the
databases. Mercury LoadRunner is used to drive Http requests to the system under test, with the html pages returned to the
virtual users.


Figure 4: Price Performance Ratio for the Web Application Test. See appendix for pricing calculations, which includes the
hardware and software costs of the middle tier application server as tested.

Trade Web Application Benchmark Discussion
This benchmark tests the complete Web applications for the IBM WebSphere 7 version of the Trade
benchmark; and the .NET StockTrader benchmark. As such it shows the relative performance of a
complete data-driven Web application across the three platforms tested. For both applications, the
Web tier directs business service operations to the business service classes, and for most pages, there
are several such operations per page, each with distinct database calls. In both applications, calls into
the business service class from the Web tier and from the business service class into the data access
class are made synchronously, and there are the same number of calls for each equivalent page across
.NET and Java applications.

For the IBM WebSphere 7 version, a combination of Java Server Pages and Servlets are used to create
the data-driven HTML pages for the user. The primary Servlet invokes a logical business service class
which in turn invokes a separate database access class. The database access class in turn uses IBM’s
JDBC driver for DB2 v9.5 to directly access the remote IBM DB2 databases. The entire application is
executing within an IBM WebSphere JVM instance per logical WebSphere node.


For the .NET StockTrader, ASP.NET Web Forms are used to invoke a logically partitioned business service
class, which in turn accesses a logically partitioned data access class. ADO.NET is used with the .NET
Provider for SQL Server to access the Microsoft SQL Server 2008 databases. The entire application is
executing within a single .NET CLR worker process instance per .NET/Windows Server 2008 node.

The following conclusions can be drawn for this benchmark workload:

 .NET on Windows Server 2008 running on the Hewlett Packard BladeSystem C7000 outperforms
IBM WebSphere 7 on the IBM Power 570 server by 37%, yet the IBM Power 570 server costs
419% more (in absolute dollars) than the HP BladeSystem running Microsoft .NET and Windows
Server 2008.
 IBM WebSphere 7 on Windows Server 2008 running on the Hewlett Packard BladeSystem C7000
outperforms IBM WebSphere 7 on the IBM Power 570 server by 39%, yet the IBM Power 570
server costs 198% more (in absolute dollars) than the HP BladeSystem running Microsoft .NET
and Windows Server 2008.
 Overall .NET on Windows Server 2008 shows by far the best price performance metric. IBM
WebSphere 7 on IBM Power 570 has a price performance metric over 8 times higher than
Microsoft .NET on Windows Server 2008. This means that for the performance actually
delivered, IBM WebSphere 7 on the IBM Power 570 server is actually over 7 times as expensive
(in dollars per TPS) than running the same workload on Microsoft .NET and Windows Server
2008 deployed to the Hewlett Packard BladeSystem.


Trade Middle Tier Web Services Test

Figure 5: Peak TPS Rates for the Trade Middle Tier Web Service Test. The .NET Capacity Planner Web Service test tool is used
to drive Web Service requests (SOAP requests) to the server, with the web service responses returned to the virtual users.


Figure 6: Price Performance Ratio for the Web Service Test. See appendix for pricing calculations, which includes the
hardware and software costs of the middle tier application server as tested.

Trade Middle Tier Web Services Benchmark Discussion
This test is a Web Service workload, with backing business logic and database access and database
transactions defined within the middle tier Web Service operations. This benchmark also tests the
Trade/.NET StockTrader application; however, it tests the middle tier service operations via remote Web
Service access. The service tier consists of a Web Service façade built on top of the business service
operations of the IBM Trade and .NET StockTrader applications. It is important to note that in this
benchmark, unlike the previous Web Application benchmark, the definition of a “transaction” (for TPS
rates) is finer-grained than in the previous test. In this test, the peak throughput is measured on all
platforms tested as a single remote Web Service request against the middle tier business operations. In
the previous test, the throughput is measured as complete web page being returned to the benchmark
clients; and each Web page has on average multiple in-process requests to the middle tier, with no
XML/Web Service serialization; but on average more business service operations and database queries
per “transaction” (web page returned). Hence, the TPS rates of the two different benchmarks should
not be directly compared. For both the Java and .NET applications, for instance, we observe higher
overall throughput rates in this test than the previous test; but that is due to the fact multiple business


service operations are being called in the Web Application Test (in-process) for each page returned; vs.
this test measuring just a single business service tier request per Web Service (remote) call.

For IBM WebSphere 7, the Web Service façade is based on JAX-WS, and exposes the services over
industry standard SOAP/WSDL. IBM HTTP Server is used to direct requests to the Web Service tier
running with the IBM WebSphere 7 application server. For the .NET StockTrader, the Web Service façade
is based on Windows Communication Foundation (WCF); which also exposes the services over the same
industry standards for SOAP/WSDL. Internet Information Services 7.0 (IIS) is used to direct requests to
the WCF-exposed services. Each application uses the exact same service and data contracts; and hence a
single Web Service client can seamlessly access either tier with no changes required.

In this test, the .NET Capacity Planner Web Service test tool (included with the .NET StockTrader
download, including source code) is used to drive load against the service tiers. 10 quad-core agents are
configured to drive as much load as required to achieve peak throughput for the service tier being
tested (without over saturating the servers). The Web Service responses (in the form of SOAP
messages) are returned to the clients. The script flow is defined as the Transaction-Light benchmark
within the .NET Capacity Planner, which tests the range of service operations, with less emphasis placed
on buys than on other service operations (to reduce database load; which helps prevent DB2 or SQL
Server disk I/O from becoming an external bottleneck).


IBM WebSphere 7 on the IBM Power 570 server by 111%, yet the IBM Power 570 server costs
419% more (in absolute dollars) than the HP BladeSystem running Microsoft .NET and Windows
Server 2008.
server costs 198% more (in absolute dollars) than the HP BladeSystem running Microsoft .NET
and Windows Server 2008.
delivered, IBM WebSphere 7 on the IBM Power 570 server is actually over 10 times as
expensive (in dollars per TPS) than running the same workload on Microsoft .NET and Windows
Server 2008 deployed to the Hewlett Packard BladeSystem.


WSTest Benchmark

Figure 7: Benchmark TPS rates for the WSTest benchmark, including EchoList, EchoStruct and GetOrder service operations.
This benchmark, which involves no business logic or database access, isolates the performance of the Web Service stack
within the application server (XML serialization, http network access, etc). The .NET Capacity Planner Web Service test tool is
used to drive Web Service requests (SOAP requests) to the server, with the web service responses returned to the virtual
users.


Figure 8: Price performance ratios for the WSTest benchmark. See appendix for pricing calculations/detail, which includes
the hardware and software costs of the middle tier application server as tested.

WSTest Benchmark Discussion
This test is a Web Service workload, with no backing business logic or database access defined within
the middle tier Web Service operations (as opposed to the previous test, which has both). This
benchmark is based on the WSTest workload, a Web Service benchmark originally designed by Sun
Microsystems and augmented by Microsoft. The service tier consists of several different operations
(EchoList, EchoStruct and GetOrder). With no business logic, these operations isolate just the
performance of the Web Service software within a given platform, including network access, and XML
serialization/deserialization of SOAP messages and objects. In this test, across all platforms, we notice
that TPS rates produced for each platform are roughly equivalent to those produced in the previous
Trade middle-tier Web service test. Readers might ask why, if there are no business logic operations or
database queries in the WSTest benchmark, are roughly equivalent TPS rates achieved in this test when
compared to the previous test?

The answer is that the Web Service message payload in this test, on average, is larger than in the
previous Trade middle tier Web Service test. Hence, in this test while less work is being performed by


business logic and database access (there is none); more work is being done on XML serialization/de-
serialization due to the larger message size. Hence, readers should not directly compare results of this
test to the previous Trade middle tier Web Service test. While both tests involve remote Web Service
operations, they are independent tests with different workload characteristics. However, it should be
noted the WSTest benchmark in this test used a default message size of 20 (20 repeating objects in all
returned message payloads); and readers are free to use the .NET Capacity Planner tool to test different
message sizes if they desire. They will then observe all platforms (.NET, WebSphere) producing higher
TPS rates with smaller message sizes, and proportionally lower TPS rates with larger message sizes, due
to the relative increase or decrease in the message size and hence XML serialization costs.

For IBM WebSphere 7 implementation of WSTest, the Web Services are based on JAX-WS, and exposed
over industry standard SOAP/WSDL. IBM HTTP Server is used to direct requests to the Web Service tier
running with the IBM WebSphere 7 application server. For the .NET implementation of WSTest, the Web
Services are based on Windows Communication Foundation (WCF); which also exposes the services over
the same industry standards for SOAP/WSDL. Internet Information Services 7.0 (IIS) is used to direct
requests to the WCF-exposed services. Each application uses the exact same service and data contracts;
and hence a single Web Service client can seamlessly access either tier with no changes required.

In this test, the .NET Capacity Planner Web Service test tool (included with the .NET StockTrader
download, including source code) is used to drive load against the service tiers. 10 quad-core agents are
configured to drive as much load as required to achieve peak throughput for the service tier being
tested (without over saturating the servers). The Web Service responses (in the form of SOAP
messages) are returned to the clients. The script flow uses a message size of 20 (meaning 20 repeating
elements within the SOAP message in each test).


IBM WebSphere 7 on the IBM Power 570 server by roughly 120% (depending on the specific
WSTest operation compared), yet the IBM Power 570 server costs 419% more than the HP
BladeSystem running Microsoft .NET and Windows Server 2008.
server costs 198% more than the HP BladeSystem running Microsoft .NET and Windows Server
2008.
delivered, IBM WebSphere 7 on the IBM Power 570 server is actually over 10 times as
expensive (in dollars per TPS) than running the same workload on Microsoft .NET and Windows
Server 2008 deployed to the Hewlett Packard BladeSystem.


Conclusion
This paper presents an extensive array of benchmark comparisons between IBM WebSphere 7 and .NET
4.0/Windows Server 2008 running three different application server workloads. The results show that
running IBM WebSphere 7 on an Hewlett Packard BladeSystem C7000 with Windows Server 2008 offers
significantly better throughput than running the equivalent workloads on an IBM Power 570 server with
AIX---and at roughly 1/3rd the cost of the middle tier hardware and software. In addition, the tests show
that the best throughput and lowest cost overall is Microsoft .NET running on Windows Server 2008—
offering up to 111% better performance than the IBM Power 570; and at roughly 1/5th the total cost of
the middle tier hardware and software. With published source code for all benchmark workload
implementations and a free test tool provided, we encourage customers to perform their own
comparative testing on the platform of their choosing; and also to use the .NET StockTrader application
as a learning sample for various features of .NET 4.0, ASP.NET and WCF Microsoft enterprise
development technologies.


Appendix A: Pricing
The following pricing was used as the base pricing and price performance metric calculations. Pricing is
based on published suggested retail pricing obtained directly from the vendor, or quotes obtained by
authorized resellers. Pricing does not include sales tax.

Hardware and Operating System Pricing

IBM Power 570
Item Quantity Price Extended Price
9117-MMA IBM SYSTEM P 570 1 $6,932.60 $6,932.60
9117-MMA-0265 AIX PARTITION 1 NA NA
SPECIFY
9117-MMA-0266 LINUX PARTITION 14 NA NA
SPECIFY
9117-MMA-1845OPERATOR PANEL 1 $135.32 $135.32
9117-MMA-3646 73 GB 15K RPM 4 $448.12 $1,792.48
SAS DISK DRIVE
9117-MMA-3660 PROC. FABRIC 1 $1,360.00 $1,360.00
CABLE,2
9117-MMA-4650 RACK INTEG.-NOT 1 NA NA
FACTORY INTEG
9117-MMA-5622 4.2GHZ POW6- 4 $5,202.00 $20,808.00
2CORE 12 DDR2
9117-MMA-5625 PROC. POWER 6 $1,020.00 $6,120.00
REGULATOR
9117-MMA-5626 SYSTEMCEC 2 $340.00 $680.00
ENCLOSURE WITH
BEZEL
9117-MMA-5628 AC POWER 4 $1,021.36 $4,085.44
SUPPLY, 1600 W
9117-MMA-5629 MEDIA 1 $125.80 $125.80
ENCLOSURE AND BACKPLA
9117-MMA-5636 2X-1GB VIRTETH- 2 $271.32 $542.64
INTEGR
9117-MMA-5648 SERVICE 2 $680.00 $1,360.00
INTERFACE CARD
9117-MMA-5657 SERVICE 1 $1,360.00 $1,360.00
INTERFACE CABLE
9117-MMA-5663 PROC 2 $1,360.00 $2,720.00
ENCLOSURE AND BACKPLAN
9117-MMA-5666 I/O BACKPLANE 2 $3,060.00 $6,120.00
9117-MMA-5667 SYSTEM 2 $680.00 $1,360.00
MIDPLANE
9117-MMA-5668 SAS DISK
BACKPLANE-26SLOT 2 $714.68 $1,429.36
9117-MMA-5672 1 PROC ACTIV FOR 8 $10,404.00 $83,232.00
PROC FT
9117-MMA-5680 ACTIV OF 1GB 32 $1,030.20 $32,966.40
DDR2 POWER 6 MEM
9117-MMA-5693 8 $1,032.24 $8,257.92
0/4GB(4X1GB),667MHZ,DDR2,POW6
9117-MMA-5699 SYSTEM SHIP 1 $13.60 $13.60
GROUP
9117-MMA-5736 PCI-X DDR 2.0 2 $528.36 $1,056.72


DUAL CHAN.U320
9117-MMA-5767 2-PORT 4 $475.32 $1,901.28
10/100/1000-TX EXP ADP
9117-MMA-6671 PWR 4 $12.92 $51.68
CORD(9FT),DRW.TO IBM PDU
9117-MMA-7164 IBM/OEM RACK- 2 $150.96 $301.92
MOUNT DRW R KIT
9117-MMA-7870 POWER 2 $180.20 $360.40
DISTRIBUTION BACKPLANE
9117-MMA-7942 ADVANCED PWR 8 $890.80 $7,126.40
VIRTUALIZATION
9117-MMA-9300 LANGUAGE 1 NA NA
GROUP SPCF-US ENG
7212-103 STORAGE DEVICE 1 $867.65 $867.65
ENCLOSURE
7212-103-1103 DVD-RAM 2 DRIVE 2 $536.00 $1,072.00
7212-103-1215 RACK MOUNT KIT 1 $100.50 $100.50
7212-103-9006 PSERIES INSTALL 1 NA NA
7212-103-9300 ENGLISH 1 NA NA
PUBLICATIONS
7212-103-9764 4.5 M VHDCI/HD68 2 NA NA
SCSI CABLE
7212-103-9860 RACK POWER 1 NA NA
CORD, AC, SINGLE
5692-A5L PSERIES SPO 1 NA NA
5692-A5L-0999 5765-G34 VIRTUAL 1 NA NA
I/O SERVER
5692-A5L-0999 5765-G34 VIRTUAL 1 NA NA
I/O SERVER
5692-A5L-1101 BASIC N/C DVD 1 NA NA
MEDIA FEATURE
5692-A5L-2924 ENGLISH 1 NA NA
5692-A5L-3435 OTHER MEDIA (AIX) 1 NA NA
DVD SYS W
5692-A5L PSERIES SPO 1 NA NA
5692-A5L-0967 5765-G03 AIX 5L 1 NA NA
FOR POWER V5
5692-A5L-0968 5765-G03 1 NA NA
EXPANSION PACK
5692-A5L-0970 5765-G03 AIX 5L 1 NA NA
UPDATE CD
5692-A5L-0975 MICROCODE UPD 1 NA NA
DISC TOOL V1.1
5692-A5L-0999 5765-G34 VIRTUAL 1 NA NA
I/O SERVER
5692-A5L-1100 DVD MEDIA 1 $50.00 $50.00
PROCESS CHARGE
5692-A5L-2924 ENGLISH 1 NA NA
5692-A5L-3435 OTHER MEDIA (AIX) 1 NA NA
DVD SYS W
5765-G03 AIX 5L FOR POWER V5.3 1 NA NA
5765-G03-0005 AIX 5L FOR POWER 8 $1,041.25 $8,330.00
V5.3 F5
5765-G34 VIOS 1.5 1 NA NA
5765-G34-0002 VIRTUAL I/O SERV 8 NA NA
F5 W/HW
5771-SWM 1 YEAR SWMA PID 1 NA NA


NUMBER
5771-SWM-0496 SWMA FOR AIX 1 8 $616.25 $4,930.00
F5 SWMA 1Y RGST
5771-SWM-0498 SWMA FOR AIX 1 8 $155.55 $1,244.40
F5 SWMA 1Y7X24
5771-VIO VIRTUAL I/O SERVER 1 NA NA
SWMA 1YR
5771-VIO-0617 GRP F5 BASIC OTC 8 $93.50 $748.00
5771-VIO-0618 VIRTUAL I/O SERV 8 $24.65 $197.20
F5 SWMA1Y724
7042-CR4 IBM 7042-CR4 1 $3,192.00 $3,192.00
7042-CR4-0962 HMC LMC V7 1 NA NA
7042-CR4-4650 RACK INDICATOR- 10 NA NA
NOT FACT INTEGR
7042-CR4-4767 HMC CR4 REDUN 1 $184.80 $184.80
POW SUPPLY 670
7042-CR4-6458 POW CBL, DRAW 2 $15.12 $30.24
14-FT, 250V/10A
7042-CR4-7802 ETH CBL,15M,HMC 1 $26.88 $26.88
CSL TO SYSUNIT
7042-CR4-9069 HMC/SRV ORDER 1 NA NA
LINKAGE IND
7042-CR4-9300 LANG GROUP 1 NA NA
SPECIFY - US ENG
7316-TF3 FLAT PANEL CONSOLE 1 $2,160.00 $2,160.00
KIT
7316-TF3-4650 NO FACTORY 1 NA NA
INTEGRATION IND
7316-TF3-8880 USB TRAVEL 1 $108.00 $108.00
KEYB.W/CBL,US ENG
7316-TF3-9300 LANG GROUP 1 NA NA
SPEC-US ENGLISH
7316-TF3-9911 POWER CORD (4M) 1 NA NA
– ALL
5771-RS1 MCRSA FOR HMC 1 1 NA NA
YEAR
5771-RS1-0612 MCRSA FOR HMC 1 $212.50 $212.50
1Y PRC MCRS 1Y
5771-RS1-0614 MCRSA FOR HMC 1 $73.95 $73.95
PRC MCRS 1Y 7X24
5771-RS1-7000 Agreement for 1 NA NA
MCRSA
Total $215,728.08


Hewlett Packard BladeSystem C7000
Hardware Pricing

HP Hardware pricing obtained from retail pricing via HP Direct, online at http://www.hp.com.

Windows Server 2008 pricing obtained based on Microsoft Suggested Retail Pricing at
http://www.microsoft.com/windowsserver2008/en/us/pricing.aspx.

Item Quantity Price Extended Price
Configurable- HP 1 $13,821.00 $13,821.00
BladeSystem c-Class
c7000 Enclosure 3-inch
LCD - Three Phase
Details of included components with BladeSystem C7000 Enclosure
HP BladeSystem c-Class c7000 Enclosure
HP 3 Phase Power Module
HP BLc7000 Management Module

4 HP BladeSystem c7000 Fans
8 ICE-BL 30 Day Evaluation Licenses - Insight Control Environment for BladeSystem
HP c-Class Insight Control Environment for BladeSystem 8 License with 1 year of 24x7 Support
2 GbE2c Ethernet Blade Switches for HP c-Class BladeSystem
HP Redundant Onboard Administrator Option
6 HP 2450W High Efficiency Power Supplies
6 HP Active Cool Fan Option Kits
Warranty - 3 years - parts, labor, onsite - next business day

HP Care Pack 3-year, 4- 1 $752.00 $752.00
Hour, 13x5 c7000
Enclosure Hardware
Support
Configurable- HP 4 $5,403.00 21,612.00
ProLiant BL460c G5
Server Blade
Details of included components with each HP BL460c Blade
HP ProLiant BL460c G5 Server Blade

Quad-Core Intel® Xeon® E5450 (3.00GHz, 1333MHz FSB, 80W) Processor
HP 32GB Fully Buffered DIMM PC2-5300 8X4GB DR LP Memory
Embedded Smart Array E200i SAS Array Controller (without cache)
HP 64MB cache upgrade for E200i
HP 72GB Hot Plug 2.5 SAS 10,000 rpm Hard Drive
HP 72GB Hot Plug 2.5 SAS 10,000 rpm Hard Drive

RAID 0 drive set (requires minimum 1 hard drive)

2 Embedded NC373i Multifunction Gigabit Network Adapters
HP iLO Advanced Blade 1 Server License with 1yr 24x7 Technical Support and Updates

Integrated Lights-Out 2 (iLO2) Standard Blade Edition (integrated on motherboard)

3-Year Limited Warranty Included
HP Care Pack 3-year, 4- 4 $296.00 $1,184.00
hour, 13x5 BL4xxc


Server Hardware
Support; electronic:
Subtotal Hardware $37,369.00
Windows Server 2008 4 1,199.00 $4796.00
Standard (64-bit)
Windows Server 2008, 4 $1,999.00 7,996.00
External Connector
License
Subtotal Operating 12,792.00
System Software
Total (HP $50,161.00
BladeSystem
hardware and
Windows Server 2008
OS)

Pricing for the Application Server Used in the Tests
Pricing includes middle tier software licensing costs for the primary application server used in the
remote tests. Database software costs and middle tier/database hardware costs were not included,
since these were not part of the system under test (SUT). We priced IBM WebSphere 7 Application
Server Edition for the application server, running on both Power6 and Intel XEON quad-core CPUs. Note
the WebSphere 7 Application Server Software costs 140% more per core to deploy to IBM Power6 than
to Intel/AMD platforms running Windows Server 2008 (50 Processor Value Units per core for Intel/AMD
vs. 120 Processor Value Units per core for Power6/IBM Power 570). Note that for the .NET
4.0/Microsoft Windows Server 2008 configuration, no separate application server is necessary: .NET is
integrated into Windows Server and new versions are made available as free downloads on MSDN.
There is also no redistribution license fee to redistribute the full .NET Framework runtime.

IBM WebSphere 7/IBM Power 570
Pricing Obtained from IBM Passport Advantage Published Pricing at http://www-
01.ibm.com/software/lotus/passportadvantage/pvu_licensing_for_customers.html.

IBM WebSphere Application Server Processor Value Unit (PVU) License + SW Subscription & Support 12
Months (D55W8LL)………………………………………..9601 @ $46.25……………………….….……………….…...$44,400.00

Total WebSphere Application Server License Costs…………………………………………………….…...$44,400.00

IBM WebSphere 7/ Hewlett Packard BladeSystem C7000
Pricing Obtained from IBM Passport Advantage Published Pricing at http://www-
01.ibm.com/software/lotus/passportadvantage/pvu_licensing_for_customers.html.

1
120 PVUs per Power6 core; 8 Power6 cores total on system as tested.


IBM WebSphere Application Server Processor Value Unit (PVU) License + SW Subscription & Support 12
Months (D55W8LL)………………………………………..8002 @ $46.25………………………………………….……..$37,000.00

Total WebSphere Application Server License Costs………………………………………………………..…...$37,000.00

Total Hardware, Operating System and Application Server Costs for IBM WebSphere
7/AIX on IBM Power 570 as tested
$260,128.08

Total Hardware, Operating System and Application Server Costs for IBM WebSphere
7/Windows Server 2008 on Hewlett Packard BladeSystem C7000 as tested
$87,161.00

Total Hardware, Operating System and Application Server Costs for Microsoft
.NET/Windows Server 2008 on Hewlett Packard BladeSystem C7000 as tested
$50,161.00

2
50 PVUs per Intel XEON core; 16 Intel Xeon cores total on system as tested.


Appendix B: Tuning Parameters
WebSphere Tuning – IBM Power 570/AIX
Servlet Caching turned on in Web Container

Session State set to 5 minute expiration (in-process session state)

Access Log Turned Off

Performance Monitor Infrastructure Turned Off

App Profile Service Off

Diagnostic Trace Turned Off

System Out Off

JVM Configured not to write System.Out printline messages

HTTP Channel maximum persistent requests = -1

Minimum Web Container threads = 50

Maximum Web Container threads = 50

Minimum ORB threads = 24

Maximum ORB threads = 24

Minimum Default threads = 20

Maximum Default threads = 20

Minimum JDBC Connections in Pool = 51

Maximum JDBC Connections in Pool = 51

Java Heap Size: 2000 min/2100 MB max

IBM HTTP Server Tuning AIX
Access Log Off

Max KeepAlive Requests unlimited

ThreadLimit 500

ServerLimit 22


StartServers 22

MaxClients 11000

MinSpareThreads 25

MaxSpareThreads 75

Threads/Child 500

MaxRequests/Child 0 (unlimited)

WebSphere Tuning – Windows Server 2008 R2
Servlet Caching turned on in Web Container

Session State set to 5 minute expiration (in-process session state)

Access Log Turned Off

Performance Monitor Infrastructure Turned Off

App Profile Service Off

Diagnostic Trace Turned Off

System Out Off

JVM Configured not to write System.Out printline messages

HTTP Channel maximum persistent requests = -1

Minimum Web Container threads = 50

Maximum Web Container threads = 50

Minimum ORB threads = 24

Maximum ORB threads = 24

Minimum Default threads = 20

Maximum Default threads = 20

Minimum JDBC Connections in Pool = 51

Maximum JDBC Connections in Pool = 51

Java Heap Size: 2000 min/2100 MB max


IBM HTTP Server Tuning Windows Server 2008 R2
Access Log Off

Max KeepAlive Requests unlimited

Default Threads 3200

Threads/Child 3200

.NET Tuning

.NET Worker Process

Rapid Fail Protection off
Pinging off
Recycle Worker Process off

ASP.NET

Forms Authentication with protection level=”All”

Forms Authentication Timeout=5 minutes

IIS 7.0 Virtual Directory

Authentication Basic Only
Access Logging Off

Service Behavior for Trade Business Services:

<behavior name="TradeServiceBehaviors">
<serviceDebug httpHelpPageEnabled="true" includeExceptionDetailInFaults="true"/>
<serviceMetadata httpGetEnabled="true" httpGetUrl=""/>
<serviceThrottling maxConcurrentInstances="400" maxConcurrentCalls="400"/>
</behavior>

.NET StockTrader

Max DB Connections = 60

Min DB Connections = 60

DB2

Logging files expanded to 15 GB

Logging Set to One Drive Array (array a)

Database file on Second Drive Array (array b)

Max Application Connections = 400

SQL/Server

Logging files expanded to 15 GB

Logging Set to One Drive Array (array a)

Database file on Second Drive Array (array b)


Ibm power570 web_sphere_7_ net_benchmark_winsrv2008

More Related Content

What's hot (17)

Similar to Ibm power570 web_sphere_7_ net_benchmark_winsrv2008 (20)

Ibm power570 web_sphere_7_ net_benchmark_winsrv2008