Ibm special hpc

Simulation
Risk Analysis
High Throughput Computing
Automotive Price Modelling

HIGH CAEEngineering Aerospace

PERFORMANCE
COMPUTING 2012/13
TECHnOLOgy CAD
COMPASS IBM SPECIAL
Big data Analytics Life Sciences

TECHnOLOgy COMPASS
TABLE Of COnTEnTS And InTrOduCTIOn

EnTErPrISE-rEAdy CLuSTEr &
WOrkLOAd MAnAgEMEnT ..........................................................................4
IBM Platform HPC .......................................................................................................6
What’s new in IBM Platform LSf 8 ...............................................................10
IBM Platform MPI 8.1 .............................................................................................20

HIgH PErfOrMAnCE COMPuTIng ....................................................... 24
Performance Turns Into Productivity .........................................................24

SCALABLE, EnErgy EffICIEnT HPC SySTEMS................................. 26
IBM idataPlex dx360 m4 HPC clusters .......................................................28
idataPlex: for High Performance
and a flexible data Center ................................................................................30

2

MOrE THAn 30 yEArS Of EXPErIEnCE In SCIEnTIfIC COMPuTIng
1980 marked the beginning of a decade where numerous startups transtec HPC solutions deliver ease of management, both in the
were created, some of which later transformed into big players in Linux and Windows worlds, and even where the customer´s envi-
the IT market. Technical innovations brought dramatic changes ronment is of a highly heterogeneous nature. Even the dynamical
to the nascent computer market. In Tübingen, close to one of ger- provisioning of HPC resources as needed does not constitute any
many’s prime and oldest universities, transtec was founded. problem, thus further leading to maximal utilization of the cluster.

In the early days, transtec focused on reselling dEC computers transtec HPC solutions use the latest and most innovative
and peripherals, delivering high-performance workstations to technology. Their superior performance goes hand in hand with
university institutes and research facilities. In 1987, Sun/Sparc energy efficiency, as you would expect from any leading edge IT
and storage solutions broadened the portfolio, enhanced by solution. We regard these basic characteristics.
IBM/rS6000 products in 1991. These were the typical worksta-
tions and server systems for high performance computing then, In 2010, transtec entered into a strategic partnership with IBM,
used by the majority of researchers worldwide. surely one of the biggest players in the HPC world with a very
strong brand. The flexibility and long-year experience of trans-
In the late 90s, transtec was one of the first companies to offer tec, combined with the power and quality of IBM HPC systems
highly customized HPC cluster solutions based on standard constitute a perfect symbiosis and provide customers with the
Intel architecture servers, some of which entered the TOP500 most optimal HPC solution imaginable. IBM idataPlex systems
list of the world’s fastest computing systems. are highly optimized for HPC workload in datacenter environ-
ments, regarding performance, flexibility, and energy, space and
Thus, given this background and history, it is fair to say that cooling efficiency. Platform HPC and LSf are both enterprise-
transtec looks back upon a more than 30 years’ experience in ready HPC cluster and workload management solutions and are
scientific computing; our track record shows nearly 500 HPC widespread in all kinds of industrial HPC environments.
installations. With this experience, we know exactly what cus-
tomers’ demands are and how to meet them. High performance Last but not least, your decision for a transtec HPC solution
and ease of management – this is what customers require to- means you opt for most intensive customer care and best ser-
day. HPC systems are for sure required to peak-perform, as their vice in HPC. Our experts will be glad to bring in their expertise
name indicates, but that is not enough: they must also be easy and support to assist you at any stage, from HPC design to daily
to handle. unwieldy design and operational complexity must be cluster operations, to HPC Cloud Services.
avoided or at least hidden from administrators and particularly
users of HPC computer systems. Have fun reading the transtec HPC Compass 2012/13 IBM Special!

3

EnTErPrISE-rEAdy
CLuSTEr &
WOrkLOAd
MAnAgEMEnT

High performance computing (HPC) is becoming a necessary tool for
organizations to speed up product design, scientific research, and busi-
ness analytics. However, there are few software environments more
complex to manage and utilize than modern high performance comput-
ing clusters. Therefore, addressing the problem of complexity in cluster
management is a key aspect of leveraging HPC to improve time to re-
sults and user productivity.

5

EnTErPrISE-rEAdy CLuSTEr & InTrOduCTIOn
Clusters based on the Linux® operating system have become
WOrkLOAd MAnAgEMEnT increasingly prevalent at large supercomputing centers and
IBM PLATfOrM HPC continue to make significant in-roads in commercial and
academic settings. This is primarily due to their superior price/
performance and flexibility, as well as the availability of com-
mercial applications that are based on the Linux OS.
Ironically, the same factors that make Linux a clear choice for
high performance computing often make the operating system
less accessible to smaller computing centers. These organiza-
tions may have Microsoft Windows administrators on staff, but
have little or no Linux or cluster management experience. The
complexity and cost of cluster management often outweigh the
benefits that make open, commodity clusters so compelling. not
only can HPC cluster deployments be difficult, but the ongo-
ing need to deal with heterogeneous hardware and operating
systems, mixed workloads, and rapidly evolving toolsets make
deploying and managing an HPC cluster a daunting task.
These issues create a barrier to entry for scientists and re-
searchers who require the performance of an HPC cluster, but
are limited to the performance of a workstation. This is why
ease of use is now mandatory for HPC cluster management.
This paper reviews the most complete and easy to use cluster
management solution, Platform HPC, which is now commer-
cially available from Platform Computing.

THE CLuSTEr MAnAgEMEnT CHALLEngE
To provide a proper HPC application environment, system
administrators need to provide a full set of capabilities to
their users, as shown below. These capabilities include cluster
provisioning and node management, application workload

6

management, and an environment that makes it easy to develop, While there are several cluster management tools that ad-
run and manage distributed parallel applications. dress parts of these requirements, few address them fully,
Modern application environments tend to be heterogeneous; and some tools are little more than collections of discrete
some workloads require Windows compute hosts while others open-source software components.
require particular Linux operating systems or versions. Some cluster toolkits focus largely on the problem of cluster
The ability to change a node’s operating system on-the-fly in re- provisioning and management. While they clearly simplify
sponse to changing application needs - referred to as adaptive cluster deployment, administrators wanting to make chang-
scheduling - is important since it allows system administrators es to node configurations or customize their environment
to maximize resource use, and present what appears to be a will quickly find themselves hand-editing XML configuration
larger resource pool to cluster users. files or writing their own shell scripts. Third-party workload
managers and various open-source MPI libraries might be
Essential components of an HPC cluster solution
included as part of a dis-
tribution. However, these
Application-Centric Interface included components are
loosely integrated and
often need to be managed
Provisioning
Workload Parallel Job Adaptive separately from the cluster
& Node
Management Enablement Scheduling manager. As a result the
Management
cluster administrator needs
to learn how to utilize each
additional piece of soft-
Uniﬁed Management Interface
ware in order to manage
the cluster effectively.
Learning how to use a command line interface to power-up, Other HPC solutions are designed purely for application
provision and manage a cluster is extremely time-consuming. workload management. While these are all capable workload
Administrators therefore need remote, web-based access to managers, most do not address at all the issue of cluster
their HPC environment that makes it easier for them to in- management, application integration, or adaptive scheduling.
stall and manage an HPC cluster. An easy-to-use application- If such capabilities exist they usually require the purchase of
centric web interface can have tangible benefits including additional software products.
improved productivity, reduced training requirements, Parallel job management is also critical. One of the primary
reduced errors rates, and secure remote access. reasons that customers deploy HPC clusters is to maximize

7

Enterprise-Ready cluster & application performance. Processing problems in paral-
lel is a common way to achieve performance gains. The
workload Management choice of MPI, its scalability, and the degree to which it is
IBM Platform HPC integrated with various OFED drivers and high performance
interconnects has a direct impact on delivered application
performance. Furthermore, if the workload manager does
not incorporate specific parallel job management features,
busy cluster users and administrators can find themselves
manually cleaning up after failed MPI jobs or writing their
own shell scripts to do the same.
Complexity is a real problem. Many small organizations or
departments grapple with a new vocabulary full of cryptic
commands, configuring and troubleshooting Anaconda kick
start scripts, finding the correct OFED drivers for special-
ized hardware, and configuring open source monitoring sys-
tems like Ganglia or Nagios. Without an integrated solution
administrators may need to deal with dozens of distinct
software components, making managing HPC cluster imple-
mentations extremely tedious and time-consuming.

Re-thinking HPC clusters
“Platform HPC and Platform LSF have been Clearly these challenges demand a fresh approach to HPC
known for many years as a highest-quality and cluster management. Platform HPC represents a “re-think”
enterprise-ready cluster and workload manage- of how HPC clusters are deployed and managed. Rather
ment solutions, and we have many customers than addressing only part of the HPC management puzzle,
in academia and industry relying on them.” Platform HPC addresses all facets of cluster management. It
provides:
A complete, easy-to-use cluster management solution
Integrated application support
Dr. Oliver Tennert Director Technology Management
HPC Solutions User-friendly, topology-aware workload management
Robust workload and system monitoring and reporting

8

Dynamic operating system multi-boot (adaptive scheduling) est and most complete HPC management capabilities while
GPU scheduling reducing overall cluster cost and improving administrator pro-
Robust commercial MPI library (Platform MPI) ductivity. It is based on the industry’s most mature and robust
Web-based interface for access anywhere workload manager, Platform LSF, making it the most reliable
solution on the market.
Most complete HPC cluster management solution Other solutions are typically a collection of open-source tools,
Platform HPC makes it easy to deploy, run and manage HPC which may also include pieces of commercially developed
clusters while meeting the most demanding requirements for software. They lack key HPC functionality and vendor sup-
application performance and predictable workload manage- port, relying on the administrator’s technical ability and time
ment. It is a complete solution that provides a robust set of to implement. Platform HPC is a single product with a single
cluster management capabilities; from cluster provisioning installer and a unified web-based management interface. With
and management to workload management and monitoring. the best support in the HPC industry, Platform HPC provides the
The easy-to -use unified web portal provides a single point of most complete solution for HPC cluster management.
access into the cluster, making it easy to manage your jobs and
optimize application performance. Complete solution
Platform HPC is more than just a stack of software; it is a fully Platform HPC provides a complete set of HPC cluster man-
integrated and certified solution designed to ensure ease of use agement features. In this section we’ll explore some of these
and simplified troubleshooting. unique capabilities in more detail.

Integrated application support Easy to use, cluster provisioning and management
High performing, HPC-optimized MPI libraries come integrated With Platform HPC, administrators can quickly provision and
with Platform HPC, making it easy to get parallel applications manage HPC clusters with unprecedented ease. It ensures maxi-
up and running. Scripting guidelines and job submission tem- mum uptime and can transparently synchronize files to cluster
plates for commonly used commercial applications simplify nodes without any downtime or re-installation.
job submission, reduce setup time and minimize operation
errors. Once the applications are up and running, Platform HPC Fast and efficient software Installation – Platform HPC can be
improves application performance by intelligently scheduling installed on the head node and takes less than one hour using
resources based on workload characteristics. three different mechanisms:
Platform HPC DVD
Fully certified and supported Platform HPC ISO file
Platform HPC unlocks cluster management to provide the easi- Platform partner’s factory install bootable USB drive

9

EnTErPrISE-rEAdy CLuSTEr Installing software on cluster nodes is simply a matter of as-
sociating cluster nodes with flexible provisioning templates
WOrkLOAd MAnAgEMEnT through the web-based interface.
WHAT’S nEW In IBM PLATfOrM LSf 8
Flexible provisioning – Platform HPC offers multiple options
for provisioning Linux operating environments that include:
Package-based provisioning
Image based provisioning
diskless node provisioning

Large collections of hosts can be provisioned using the same
provisioning template. Platform HPC automatically manages
details such as IP address assignment and node naming conven-
tions that reflect the position of cluster nodes in data center
racks.
unlike competing solutions, Platform HPC deploys multiple
operating systems and OS versions to a cluster simultaneously.
This includes red Hat® Enterprise Linux, CentOS, Scientific
Linux, and SuSE® Linux Enterprise Server. This provides admin-
istrators with greater flexibility in how they serve their user
communities and means that HPC clusters can grow and evolve
incrementally as requirements change.

10

Delegation of administrative rights – Empowers line of
business owners to take control of their own projects
Fairshare pre-emptive scheduling enhancements – fine
tunes key production policies

PLATfOrM LSf 8 fEATurES
Guaranteed Resources Ensure Deadlines are Met
WHAT’S nEW In IBM PLATfOrM LSf 8 In Platform LSf 8, resource-based scheduling has been extended
Written with Platform LSf administrators in mind, this brief to guarantee resource availability to groups of jobs. resources
provides a short explanation of significant changes in Plat- can be slots, entire hosts or user-defined shared resources such
form’s latest release of Platform LSf, with a specific emphasis as software licenses. As an example, a business unit might guar-
on scheduling and workload management features. antee that it has access to specific types of resources within ten
minutes of a job being submitted, even while sharing resources
ABOuT IBM PLATfOrM LSf 8 between departments. This facility ensures that lower priority
Platform LSf is the most powerful workload manager for jobs using the needed resources can be pre-empted in order to
demanding, distributed high performance computing environ- meet the SLAs of higher priority jobs.
ments. It provides a complete set of workload management Because jobs can be automatically attached to an SLA class via
capabilities, all designed to work together to reduce cycle times access controls, administrators can enable these guarantees
and maximize productivity in missioncritical environments. without requiring that end-users change their job submission
procedures, making it easy to implement this capability in exist-
This latest Platform LSf release delivers improvements in perfor- ing environments.
mance and scalability while introducing new features that sim-
plify administration and boost user productivity. This includes: Live Cluster Reconfiguration
Guaranteed resources – Aligns business SLA’s with infra- Platform LSf 8 incorporates a new live reconfiguration capabil-
structure configuration for simplified administration and ity, allowing changes to be made to clusters without the need to
configuration re-start LSf daemons. This is useful to customers who need to add
Live reconfiguration – Provides simplified administration hosts, adjust sharing policies or re-assign users between groups
and enables agility “on the fly”, without impacting cluster availability or running jobs.

11

EnTErPrISE-rEAdy CLuSTEr
WOrkLOAd MAnAgEMEnT
WHAT’S nEW In IBM PLATfOrM LSf 8

Changes to the cluster configuration can be made via the bconf
command line utility, or via new API calls. This functionality can
also be integrated via a web-based interface using Platform
Application Center. All configuration modifications are logged
for a complete audit history, and changes are propagated almost
instantaneously. The majority of reconfiguration operations are
completed in under half a second.
With Live reconfiguration, down-time is reduced, and administra-
tors are free to make needed adjustments quickly rather than
wait for scheduled maintenance periods or non-peak hours. In
cases where users are members of multiple groups, controls can
be put in place so that a group administrator can only control
jobs associated with their designated group rather than impact-
ing jobs related to another group submitted by the same user.

Delegation of Administrative Rights
With Platform LSf 8, the concept of group administrators has
been extended to enable project managers and line of business
managers to dynamically modify group membership and fair-
share resource allocation policies within their group. The ability
to make these changes dynamically to a running cluster is made
possible by the Live reconfiguration feature.
These capabilities can be delegated selectively depending on the
group and site policy. different group administrators can man-
age jobs, control sharing policies or adjust group membership.

12

More Flexible Fairshare Scheduling Policies More Sophisticated Host-based Resource Usage for Parallel Jobs
To enable better resource sharing flexibility with Platform Platform LSf 8 provides several improvements to how resource
LSf 8, the algorithms used to tune dynamically calculated use is tracked and reported with parallel jobs. Accurate tracking
user priorities can be adjusted at the queue level. These algo- of how parallel jobs use resources such as CPus, memory and
rithms can vary based on department, application or project swap, is important for ease of management, optimal scheduling
team preferences. The fairshare parameters EnABLE_HIST_ and accurate reporting and workload analysis. With Platform
run_TIME and HIST_HOurS enable administrators to control LSf 8 administrators can track resource usage on a per-host
the degree to which LSf considers prior resource usage when basis and an aggregated basis (across all hosts), ensuring that
determining user priority. The flexibility of Platform LSf 8 resource use is reported accurately. Additional details such as
has also been improved by allowing a similar “decay rate” to running PIds and PgIds for distributed parallel jobs, manual
apply to currently running jobs (run_TIME_dECAy), either cleanup (if necessary) and the development of scripts for
system-wide or at the queue level. This is most useful for cus- managing parallel jobs are simplified. These improvements in
tomers with long-running jobs, where setting this parameter resource usage reporting are reflected in LSf commands includ-
results in a more accurate view of real resource use for the ing bjobs, bhist and bacct.
fairshare scheduling to consider.
Improved Ease of Administration for Mixed Windows and
Performance Scalability Enhancements Linux Clusters
Platform LSf has been extended to support an unparalleled The lspasswd command in Platform LSf enables Windows LSf
scale of up to 100,000 cores and 1.5 million queued jobs for users to advise LSf of changes to their Windows level passwords.
very high throughput EdA workloads. Even higher scalability With Platform LSf 8, password synchronization between environ-
is possible for more traditional HPC workloads. ments has become much easier to manage because the Windows
Specific areas of improvement include the time required to passwords can now be adjusted directly from Linux hosts using
start the master-batch daemon (MBd), bjobs query perfor- the lspasswd command. This allows Linux users to conveniently
mance, job submission and job dispatching as well as im- synchronize passwords on Windows hosts without needing to
pressive performance gains resulting from the new Bulk Job explicitly login into the host.
Submission feature. In addition, on very large clusters with
large numbers of user groups employing fairshare schedul- Bulk Job Submission
ing, the memory footprint of the master batch scheduler in When submitting large numbers of jobs with different resource
LSf has been reduced by approximately 70% and scheduler requirements or job level settings, Bulk Job Submission allows
cycle time has been reduced by 25%, resulting in better perfor jobs to be submitted in bulk by referencing a single file
formance and scalability. containing job details.

13

Enterprise-Ready cluster Simplified configuration changes – Platform HPC simplifies ad-
ministration and increases cluster availability by allowing changes
workload Management such as new package installations, patch updates, and changes to
ibm Platform HPC configuration files to be propagated to cluster nodes automati-
cally without the need to re-install those nodes. It also provides
Resource monitoring a mechanism whereby experienced administrators can quickly
perform operations in parallel across multiple cluster nodes.

Repository snapshots / trial installations – Upgrading soft-
ware can be risky, particularly in complex environments. If a
new software upgrade introduces problems, administrators
often need to rapidly “rollback” to a known good state. With
other cluster managers this can mean having to re-install the
entire cluster. Platform HPC incorporates repository snapshots,
which are “restore points” for the entire cluster. Administrators
can snapshot a known good repository, make changes to their
environment, and easily revert to a previous “known good”
repository in the event of an unforeseen problem. This powerful
capability takes the risk out of cluster software upgrades.

New hardware integration – When new hardware is added to
a cluster it may require new or updated device drivers that are
not supported by the OS environment on the installer node. This
means that a newly updated node may not network boot and
provision until the head node on the cluster is updated with a
new operating system; a tedious and disruptive process. Platform
HPC includes a driver patching utility that allows updated device
drivers to be inserted into existing repositories, essentially future
proofing the cluster, and providing a simplified means of support-
ing new hardware without needing to re-install the environment
from scratch.

14

Software updates with no re-boot – Some cluster managers to run other jobs. Using the unified management interface,
always re-boot nodes when updating software, regardless of administrators can monitor the GPU performance as well as
how minor the change. This is a simple way to manage updates. detect ECC errors.
However, scheduling downtime can be difficult and disruptive.
Platform HPC performs updates intelligently and selectively Unified management interface
so that compute nodes continue to run even as non-intrusive Competing cluster management tools either do not have a
updates are applied. The repository is automatically updated so web-based interface or require multiple interfaces for managing
that future installations include the software update. Changes different functional areas. In comparison, Platform HPC includes
that require the re-installation of the node (e.g. upgrading an a single unified interface through which all administrative tasks
operating system) can be made in a “pending” state until down- can be performed including node-management, job-manage-
time can be scheduled. ment, jobs and cluster monitoring and reporting. Using the
unified management interface, even cluster administrators with
User-friendly, topology aware workload manage- very little Linux experience can competently manage a state of
ment the art HPC cluster.
Platform HPC includes a robust workload scheduling capability,
which is based on Platform LSF - the industry’s most powerful, Job management – While command line savvy users can
comprehensive, policy driven workload management solution continue using the remote terminal capability, the unified web
for engineering and scientific distributed computing environ- portal makes it easy to submit, monitor, and manage jobs. As
ments. By scheduling workloads intelligently according to changes are made to the cluster configuration, Platform HPC
policy, automatically re-configures key components, ensuring that
Platform HPC improves end user productivity with minimal jobs are allocated to the appropriate resources. The web portal
system administrative effort. In addition, it allows HPC user is customizable and provides job data management, remote
teams to easily access and share all computing resources, while visualization and interactive job support.
reducing time between simulation iterations.
Workload/system correlation – Administrators can correlate
GPU scheduling – Platform HPC provides the capability to workload information with system load, so that they can make
schedule jobs to GPUs as well as CPUs. This is particularly timely decisions and proactively manage compute resources
advantageous in heterogeneous hardware environments as against business demand. When it’s time for capacity planning,
it means that administrators can configure Platform HPC so the management interface can be used to run detailed reports
that only those jobs that can benefit from running on GPUs are and analyses which quantify user needs and remove the guess
allocated to those resources. This frees up CPU-based resources work from capacity expansion.

15

Enterprise-Ready cluster Simplified cluster management – The unified management
console is used to administer all aspects of the cluster envi-
workload Management ronment. It enables administrators to easily install, manage
ibm Platform HPC and monitor their cluster. It also provides an interactive en-
vironment to easily package software as kits for application
deployment as well as pre-integrated commercial application
support. One of the key features of the interface is an opera-
tional dashboard that provides comprehensive administra-
tive reports. As the image illustrates, Platform HPC enables
administrators to monitor and report on key performance
metrics such as cluster capacity, available memory and CPU
utilization. This enables administrators to easily identify and
troubleshoot issues.
The easy to use interface saves the cluster administrator
time, and means that they do not need to become an expert
Job submission templates in the administration of open-source software components.
It also reduces the possibility of errors and time lost due to
incorrect configuration. Cluster administrators enjoy the
best of both worlds – easy access to a powerful, web-based
cluster manager without the need to learn and separately
administer all the tools that comprise the HPC cluster envi-
ronment.

Robust Commercial MPI library
Platform MPI – In order to make it easier to get parallel appli-
cations up and running, Platform HPC includes the industry’s
most robust and highest performing MPI implementation,
Platform MPI. Platform MPI provides consistent performance
at application run-time and for application scaling, result-
ing in top performance results across a range of third-party
benchmarks.

16

Open Source MPI – Platform HPC also includes various OS to meet application demands, providing them with the
other industry standard MPI implementations. This includes flexibility to support special requests and accommodate
MPICH1, MPICH2 and MVAPICH1, which are optimized for unanticipated changes. Rather than being an extracost item
cluster hosts connected via InfiniBand, iWARP or other RDMA as it is with other HPC management suites, this capability is
based interconnects. included as a core feature of Platform HPC.

Integrated application support Commercial Service and support
Job submission templates – Platform HPC comes complete Certified cluster configurations – Platform HPC is tested and
with job submission templates for ANSYS Mechanical, ANSYS certified on all partner hardware platforms. By qualifying each
Fluent, ANSYS CFX, LS-DYNA, MSC Nastran, Schlumberger platform individually and providing vendor-specific software
ECLIPSE, Simulia Abaqus, NCBI Blast, NWChem, ClustalW, and with optimized libraries and drivers that take maximum
HMMER. By configuring these templates based on the ap- advantage of unique hardware features, Platform Computing
plication settings in your environment, users can start using has essentially done the integration work in advance. As a
the cluster without writing scripts. result, clusters can be deployed quickly and predictably with
minimal effort. As a testament to this, Platform HPC is certi-
Scripting Guidelines – Cluster users that utilize homegrown fied under the Intel Cluster Ready program.
or open-source applications, can utilize the Platform HPC
scripting guidelines. These user-friendly interfaces help mini- Enterprise class service and support – Widely regarded as
mize job submission errors. They are also self-documenting, having the best HPC support organization in the business,
enabling users to create their own job submission templates. Platform Computing is uniquely able to support an inte-
grated HPC platform. Because support personnel have direct
Benchmark tests – Platform HPC also includes standard access to the Platform HPC developers, Platform Computing
benchmark tests to ensure that your cluster will deliver the is able to offer a higher level of support and ensure that any
best performance without manual tuning. problems encountered are resolved quickly and efficiently.

Flexible OS provisioning Summary
Platform HPC can deploy multiple operating system versions Platform HPC is the ideal solution for deploying and manag-
concurrently on the same cluster and, based on job resource ing state of the art HPC clusters. It makes cluster management
requirements, dynamically boot the Linux or Windows oper- simple, enabling analysts, engineers and scientists from organi-
ating system required to run the job. Administrators can also zations of any size to easily exploit the power of Linux clusters.
use a web interface to manually switch nodes to the required Unlike other HPC solutions that address only parts of the HPC

17

WOrkLOAd MAnAgEMEnT
IBM PLATfOrM HPC
Automotive
Risk Analysis

CAD Life Sciences
Aerospace
Engineering
Price Modelling

Simulation
CAE
Aerospace High Throughput Computing

CAD
Life Sciences
Risk Analysis

Big data Analytics

18
Simulation

management challenge, Platform HPC uniquely addresses all
Capability / Feature Platform HPC
aspects of cluster and management including:
Cluster Provisioning and Management x
Easy-to-use cluster provisioning and management
Initial cluster provisioning x
User-friendly, topology aware workload management
Unified management interface Multiple provisioning methods x

Robust commercial MPI library Web-based cluster management x

Integrated application support Node updates with no re-boot x

Flexible OS provisioning Repository snapshots x
Commercial HPC service and support Flexible node templates x

Multiple OS and OS versions x
By providing simplified management over the entire lifecycle
Workload Management Application
of a cluster, Platform HPC has a direct and positive impact on x
Integration
productivity while helping to reduce complexity and cost. The
Integrated workload management x
comprehensive web-based management interface, and features
like repository snapshots and the ability to update software HPC libraries toolsets x

packages on the fly means that state-of-the-art HPC clusters can NVIDIA CUDA SDK support x

be provisioned and managed even by administrators with little Web-based job management x

or no Linux administration experience. Web-based job data management x

Multi-boot based on workload x

Price Modelling Advanced parallel job management x

Commercial application integrations x

MPI Libraries x

Commercial grade MPI x

Workload and system monitoring,

Automotive reporting and correlation
Workload monitoring
x

x

Workload reporting x
Big Data Analytics System monitoring reporting x

CAE
Workload and system load correlation x

Integration with 3rd party
x
management tools

Engineering
19

WOrkLOAd MAnAgEMEnT
IBM PLATfOrM MPI 8.1


BEnEfITS
Superior application performance
reduced development and support costs
faster time-to-market
The industry’s best technical support

fEATurES
Supports the widest range of hardware, networks and
operating systems
distributed by over 30 leading commercial software vendors
Change interconnects or libraries with no need to re-
compile
Seamless compatibility across Windows® and Linux®
environments
Ensures a production quality implementation

IdEAL fOr:
Enterprises that develop or deploy parallelized software
applications on HPC clusters

20

Commercial software vendors wanting to improve appli- Platform MPI is optimized for both distributed (dMP) and shared
cations performance over the widest range of computer memory (SMP) environments and provides a variety of flexible
hardware, interconnects and operating systems CPu binding strategies for processes and threads, enabling better
performance on multi–core environments. With this capability
THE STAndArd fOr SCALABLE, PArALLEL APPLICATIOnS memory and cache conflicts are managed by more intelligently
Platform MPI is a high performance, production–quality imple- distributing the load among multiple cores.
mentation of the Message Passing Interface (MPI). It is widely With support for Windows HPC Server 2008 and the Microsoft job
used in the high performance computing (HPC) industry and is scheduler, as well as other Microsoft operating environments,
considered the de facto standard for developing scalable, paral- Platform MPI allows developers targeting Windows platforms to
lel applications. enjoy the benefits of a standard portable MPI and avoid propri-
Platform MPI maintains full backward compatibility with HP- etary lock-in.
MPI and Platform MPI applications and incorporates advanced
CPu affinity features, dynamic selection of interface libraries,
superior workload manger integrations and improved perfor-
mance and scalability.
Platform MPI supports the broadest range of industry standard
platforms, interconnects and operating systems helping ensure
that your parallel applications can run anywhere.

fOCuS On POrTABILITy
Platform MPI allows developers to build a single executable
that transparently leverages the performance features of any
type of interconnect, thereby providing applications with
optimal latency and bandwidth for each protocol. This reduces
development effort, and enables applications to use the “latest
and greatest” technologies on Linux or Microsoft Windows
without the need to re-compile and re-link applications.

21

WOrkLOAd MAnAgEMEnT

SuPPOrTEd OPErATIng SySTEMS
red Hat Enterprise Linux 4.6, 5.x and 6.x
SuSE Linux Enterprise Server 10 and 11
CentOS 5.3
Microsoft Windows® XP/Vista, Server 2003/Server 200 8/HPC
Server 2008, Windows 7

SuPPOrTEd InTErCOnnECTS And PrOTOCOLS
Myrinet (Linux) gM MX on X86-64 and Itanium2

InfiniBand (Linux) OfEd, PSM, udAPL on X86-64 and
Itanium2
OfEd 1.1, 1.2, 1.3, 1.4, 1.5
Sdr, ddr, Qdr, ConnectX and
ConnextX2 Mellanox fCA

gigE (Linux) rdMA, udAPL, TCP/IP

Infiniband (Windows) WinOf 2.x, IBAL, WSd, Sdr, ddr,
Qdr, ConnectX(2)

gigE (Windows) TCP/IP on x86-64

22

fEATurES And BEnEfITS

Features Benefits
fully complies with the MPI 2.2 standard, providing Applications port easily to other platforms
dynamic processes, one–sided communications, ex- Protects ISV software investment
tended collectives, thread safety, and updated rOMIO reduces time-to-market
Complete debugging, diagnostic and profiling tools Increased robustness and quality of applica-
Simplicity
Auto-detection of interconnects and dynamic loading tions
of libraries Technical problems resolved quickly and
no re-link required for debugging and profiling efficiently
Supported by the largest dedicated HPC support
organization
Improved shared memory performance, incorporating Takes maximum advantage of available
code and methods from Platform MPI 5.6 (Scali MPI) hardware
75% reduction in job startup and shutdown at scale reduced latency for better performance
Performance Scalability to 17,000 ranks Performance improves without explicit
rdMA message progression coalescing enhance- developer action
ments Better message throughput in streaming
flexible CPu binding options maximize cache effec- applications
tiveness and balance applications to minimize latency Easier to optimize application performance
Automated benchmarking of collective operations
Common source-code base between Linux and Windows Avoid the cost of separate releases for differ-
Binary compatible with applications developed for ent platforms
HP-MPI Easily used with existing MPI applications
MPICH-2 compatibility mode Common mpirun syntax between Linux and
Compatibility Linux Standard Bindings ensure full compatibility Windows
across all major Linux distributions Customers avoid proprietary “lock-in”
Scheduler agnostic with workload manager integra- Avoid floating point issues causing inconsis-
tions for Windows HPC, Platform LSf, PBS Pro, SLurM tent results
and other popular schedulers and resource managers
Supports the widest variety of networks and intercon- develop applications that will run on more
nects platforms
Select interconnects at run-time with no need to re- reduce testing, maintenance and support
compile costs
Flexibility Write applications once and deploy across multiple OS Enjoy strategic flexibility
and hardware topologies
CPu binding features well suited to gPu-aware applica-
tions

23

HPC solution
benchmarking of application
HIgH PErfOrMAnCE COMPuTIng
different systems installation
PErfOrMAnCE TurnS InTO PrOduCTIVITy

continual
improvement

maintenance, integration
onsite
customer into
support hardware
training customer’s
managed services assembly
environment

SErVICES And CuSTOMEr CArE frOM A TO Z

application-, burn-in tests software
individual Presales
customer-, of systems OS
consulting
site-speciﬁc installation
sizing of
HPC solution
benchmarking of application
different systems installation

continual
improvement

maintenance, integration
onsite
customer into
support hardware
training customer’s
managed services assembly
environment

24

important middleware components like cluster management or
developer tools and the customer’s production applications. Onsite
delivery means onsite integration into the customer’s production
environment, be it establishing network connectivity to the corpo-
rate network, or setting up software and configuration parts.

transtec HPC clusters are ready-to-run systems – we deliver, you
HPC @ TrAnSTEC: SErVICES And CuSTOMEr CArE frOM A TO Z turn the key, the system delivers high performance. Every HPC
transtec Ag has over 30 years of experience in scientific comput- project entails transfer to production: IT operation processes and
ing and is one of the earliest manufacturers of HPC clusters. policies apply to the new HPC system. Effectively, IT personnel is
for nearly a decade, transtec has delivered highly customized trained hands-on, introduced to hardware components and soft-
High Performance clusters based on standard components to ware, with all operational aspects of configuration management.
academic and industry customers across Europe with all the
high quality standards and the customer-centric approach that transtec services do not stop when the implementation proj-
transtec is well known for. ects ends. Beyond transfer to production, transtec takes care.
transtec offers a variety of support and service options, tailored
Every transtec HPC solution is more than just a rack full of hard- to the customer’s needs. When you are in need of a new instal-
ware – it is a comprehensive solution with everything the HPC lation, a major reconfiguration or an update of your solution
user, owner, and operator need. – transtec is able to support your staff and, if you lack the re-
sources for maintaining the cluster yourself, maintain the HPC
In the early stages of any customer’s HPC project, transtec solution for you. from Professional Services to Managed Servic-
experts provide extensive and detailed consulting to the cus- es for daily operations and required service levels, transtec will
tomer – they benefit from expertise and experience. Consulting be your complete HPC service and solution provider. transtec’s
is followed by benchmarking of different systems with either high standards of performance, reliability and dependability
specifically crafted customer code or generally accepted bench- assure your productivity and complete satisfaction.
marking routines; this aids customers in sizing and devising the
optimal and detailed HPC configuration. transtec’s offerings of HPC Managed Services offer customers
the possibility of having the complete management and admin-
Each and every piece of HPC hardware that leaves our factory istration of the HPC cluster managed by transtec service spe-
undergoes a burn-in procedure of 24 hours or more if neces- cialists, in an ITIL compliant way. Moreover, transtec’s HPC on
sary. We make sure that any hardware shipped meets our and demand services help provide access to HPC resources when-
our customers’ quality requirements. transtec HPC solutions are ever they need them, for example, because they do not have the
turnkey solutions. By default, a transtec HPC cluster has everything possibility of owning and running an HPC cluster themselves,
installed and configured – from hardware and operating system to due to lacking infrastructure, know-how, or admin staff.

25

SCALABLE EnErgy
EffICIEnT
HPC SySTEMS

data Centers today increasingly face challenges to keep pace with
growth and varying business demands that require rapid scaling and
technology evolution. While each technology refresh cycle delivers a
significant performance boost at roughly the same price, these data
centers face rising facilities costs and mounting energy bills for power
and cooling. The operational costs for labor and the ongoing equip-
ment maintenance costs are other key components of the data Center
Total Cost of Ownership (TCO). In order to get the most out of their cur-
rent and future IT investments, data Center managers must minimize
TCO while achieving desired efficiencies, computing performance and
effective manageability of massive scale-out environments.

27

SCALABLE, EnErgy EffICIEnT InTrOduCTIOn
As a leading provider of High Performance Computing (HPC)
HPC SySTEMS solutions across a wide range of application and industry seg-
IBM IdATAPLEX dX360 M4 HPC CLuSTErS ments over the last few decades, IBM has constantly innovated
and helped alleviate customer pain points through its rich
portfolio for HPC. The new system design of IBM’s idataPlex
dx360 M4 servers leverages Intel’s industry-standard highly
scalable, powerful processors to address energy efficiency and
IBM SySTEM X IdATAPLEX dIrECT WATEr COOLEd dX360 M4 significantly reduce carbon emission footprints of next genera-
tion data centers. Intel’s Xeon processor E5-2600 product family
using 32nm Sandy Bridge processors power the IBM idataPlex
dx360 M4 air-cooled half depth servers to provide more rack
density per square feet of floor space in the data center en-
abling higher compute power for less energy per square foot of
data center floor-space.

Overcoming Data Center Challenges
Business users, engineers and scientists attempting to solve
Highlights: challenging problems in engineering, financial markets, media,
40% efficiency advantage over air cooled systems by public sector, oil gas and the life and earth sciences, rely heav-
eliminating 90% of heat per node ily on HPC systems. These users are some of the most demand-
Provides up to 10% power advantage over air-cooled ing IT solution clients. Their insatiable demand for computa-
systems tional performance continues to drive HPC platforms towards
Two water cooling loops per node to cool proces- increasing system scalability, larger computing density, lower
sors and memory dimms remaining components air power consumption, efficient system cooling, easier and more
cooled reliable system management, and adaptable programming
90% heat recovery can be reused for economical pur- environments.
poses, such as warming other buildings and facilities Multiple industry studies indicate that revenue from HPC
Typical operating conditions: Ambient air servers will continue to grow much faster than overall server
inlet @ 27-35°C and water inlet @ 40-45°C to the nodes revenues with clusters being the dominant platform – over

28

75 percent of HPC servers. The battle for HPC leadership will cooling loops. This water-cooled design eliminates the need for
become increasingly global and strategic. More real-world fans in a 2U chassis and significantly boosts energy efficiency.
applications will run at trans-petaflop speeds. However, these This allows operations that are more reliable and better place-
studies also suggest that the escalating costs of server management of large clusters at HPC data centers – significantly reduc-
ment, power and cooling, reliability, availability and serviceabil- ing the overall power, cooling and facilities management costs
ity (RAS) of HPC clusters with thousands of racks and facilities for the entire data center.
management at HPC data centers will far outpace the costs of
buying new servers. This has caused a severe crisis in HPC data iDataPlex: Scalable, Energy Efficient and Economical HPC
centers. IT solution providers such as IBM have responded by Workhorse
designing innovative solutions that address these issues while With direct water-cooling technology, the iDataPlex can help
retaining all the attractive attributes of industry-standard clus- lower power consumption by up to 40% and can greatly reduce
ter architectures for HPC. the air conditioning need in data centers. Another advantage
is reuse of 90% of the heat removed using direct water-cooling
HPC Cluster Differentiators: iDataPlex Innovations in Packag- technology from the data center for community purposes such
ing, Power and Cooling as heating buildings or water. The direct effect of this innova-
IBM’s iDataPlex dx360 M4 server can significantly reduce data tive architecture is that it economically increases the compute
center costs arising from power consumption. It also helps density in the data center by a factor of five while retaining
reduce floor space requirements for cold aisles through innova- many of the attractive attributes of current integrated rack
tions in cooling technologies and smart system designs that cluster solutions from IBM. As compared to air-cooling, direct
require less floor space. Data centers can deploy the next gen- water-cooling technology also consumes much less energy.
eration dx360 M4 with Intel Xeon processors E5-2600 product
family and obtain more computing density while consuming IBM Continues Active Collaboration with HPC Developers and
the same or lower power. They also benefit from new packaging Partners
and cooling methods – traditionally used in very large enter- Since the iDataPlex can provide a large scale-out cluster using
prise class systems and proven very effective for designing industry standard components, the large portfolio of HPC applica-
energy-efficient clusters where water is pumped through the tions that have been mapped, migrated, and optimized for the
rack and into node-level cooling loops, which will provide data IBM Cluster architecture will benefit greatly from this increased
centers with un-paralleled advantages. density, scalability, performance and cost-effectiveness. This helps
The iDataPlex dx360 M4 has a hybrid air and direct water-cooled to protect customer’s prior investment in application enablement
design. The direct water-cooled solution is innovative node level and related. The iDataPlex provides an efficient, reliable and cost-
cooling that pumps water through the rack and into node-level effective building block for supercomputing HPC clusters.

29

SCALABLE, EnErgy EffICIEnT IdATAPLEX: fOr HIgH PErfOrMAnCE And A fLEXIBLE
dATA CEnTEr
HPC SySTEMS To address the new data center challenges, a new way of
IBM IdATAPLEX dX360 M4 HPC CLuSTErS designing data centers and the server infrastructure that goes
into the data centers is essential. The design must encompass
data center power, cooling, management, and acquisition as
well as operating costs as the chief design goals, in addition
to the performance, cost, and scalability requirements at the
server level. While IBM has years of experience designing server
technologies for scale-up and scale-out that primarily focus on
performance and scalability as the fundamental requirements,
the idataPlex solution focuses on a different set of goals:
reduce the initial hardware acquisition costs and on-going
maintenance costs for data center owners
Improve efficiency in power consumption
Eliminate data center cooling requirements
Achieve higher server density within the same footprint as
the traditional rack layout
Simplify manageability for massive scale-out environments
reduce the time to deployment through pre-configuration
and full integration at manufacturing

As is evident, these design goals go far beyond a single server or
a single rack level; they are goals for the entire data center. With
this new philosophy and the new design, the idataPlex solution
promises to address the data center challenges at various levels:
an innovative rack design for higher node density
a flex node chassis based on industry standard components
shared power and cooling components to improve node as
well as rack level efficiencies
optional rear-door heat exchangers that virtually eliminates
traditional cooling and lastly,
optional direct water cooling technology that can lower
power consumption by up to 40%

30

THE fOLLOWIng fIgurE HIgHLIgHTS EVOLuTIOn Of COOLIng TECHnOLOgIES In IBM’S IdATAPLEX SOLuTIOn

TODAY
Cooling at Rack Level with Rear
Door Heat Exchanger TOMORROW
Cooling at Node Level with Direct
Water Cooling

Increase data center density by eliminating hot/cold aisles node-level cooling loops cool processors and memory, all
Eliminate rack heat exhaust other components air cooled
Same dimensions as standard idataPlex rear door 4” deep Water manifolds for intra-rack distribution
Liquid cooling at the rack is up to 95% more efficient than air 40% efficiency advantage over air cooled servers
cooling by a CrAC 90% heat recover per node
no electrical or moving parts Heat removed from the datacenter can be reused for
no condensation community purposes (ex. heating buildings or water)
Chilled water

31

SCALABLE, EnErgy EffICIEnT iDataPlex Architecture: Cost Effective, Flexible and Scalable
With major cooling innovations, the direct water-cooled
HPC SySTEMS idataPlex promises to be a unique offering for HPC data
IBM IdATAPLEX dX360 M4 HPC CLuSTErS centers. IBM’s enterprise experience and traditional HPC needs
combine to create a low cost platform with no redundant
hardware, easily available industry standard parts, extreme
compute density, increased performance/watt, and quick
and large-scale deployment. With a more efficient standard
form factor, the idataPlex combines servers, chassis, switches,
Pdus, management appliances, and efficient rear door heat
exchanger and latest direct water-cooling technology to
address many HPC needs such as flexible configurations for
nodes, storage and I/O capability including power-optimized
architectures.
With the Intel Xeon processor E5-2600 product family powered
idataPlex M4, Petascale HPC clusters and data centers can
achieve over 55% reduction in overall TCO and as much as 90%
savings in power and cooling through innovations in air cool-
ing, rear door heat exchanger chassis, energy efficient Intel
processors, rack, and optional liquid cooling in an idataPlex.

iDataPlex Features and Benefits at a Glance
The idataPlex is a result of a decade of system innovation by
IBM and promises to close the loop between low-end inex-
pensive volume servers and enterprise class SMP and blade
servers. This massive, scalable computing workhorse includes
a wide-range of flexible options that make it a great choice for
HPC applications and other large-scale applications. The figure
below depicts the salient system details and summarizes the
idataPlex features and benefits.

32

Benefits iDataPlex Features
Racks fully integrated by IBM manufacturing at the factory site
Delivered, installed, and powered up at the Data Center before customer acceptance
Deployable
Minimal on-site time required – shipping dock to image-load

All front access – eliminates need for accessing rear of rack
Blade-like design with chassis docking into power connector
Serviceable Innovative cable layout provides highly efficient cable routing
Flexible support options from self maintenance to 24x7x4 response time

Lower power consumption by up to 40% per rack
90% extracted heat can be reused for community purposes (heating buildings)
Efficient Liquid cooling at the rack is up to 95% more efficient than air cooling by a CRAC
Energy expenses reduced by up to 40%
Minimize heat exhaust with optional Rear Heat Exchanger

Excellent flexibility in node and rack configuration
Factory integrated racks delivered to the Data Center
Flexible
3rd party options and rack support

Shared infrastructure designs cut cost out
Non-redundant components
Affordable
TCO benefits -- Electricity cost ~90% lower, IT Cap Ex ~60% higher but 55% less overall
TCO with iDataPlex.

Blade like thinking in design
Scalable system management software for large-scale configurations
Manageable Rack management appliance for several racks
Ease in servicing from complete front access for upgrades
RAS advantages

33

SCALABLE, EnErgy EffICIEnT THE IdATAPLEX: TAILOrEd fOr HPC APPLICATIOnS
HPC users have always demanded computing solutions that
HPC SySTEMS have the best performance, price/performance, and now
IBM IdATAPLEX dX360 M4 HPC CLuSTErS are increasingly demanding energy-efficient platforms. The
idataPlex with its economical, large scale-out architecture is
an excellent workhorse for a wide-range of HPC applications
in engineering, life sciences, upstream petroleum, financial
services, and scientific research. Often, these applications do
not require hardware redundancy for failure protection, and
software fault-tolerance built into the application and/or
system software is more than adequate for practical pur-
poses.
Over the last decade, in conjunction with application devel-
opers, IBM has made substantial investments to optimize
and tune HPC applications to take advantage of the IBM
Cluster architecture. Together they have achieved substan-
tial performance tuning by careful load balancing, maximiz-
ing single socket performance, maintaining data locality,
minimizing cache misses, and maximizing the computation-
to-communication ratio. In many cases, application perfor-
mance is further enhanced by the use of optimized mathe-
matical libraries and high-performance I/O solutions like the
IBM general Parallel file System (gPfS). All these investments
translate readily to the idataPlex M4.

34

Risk Analysis Automotive
CAD Life Sciences
Price Modelling
Aerospace
Engineering
ice Modelling
Simulation CAE
Aerospace High Throughput Computing

CAD Life Sciences
Risk Analysis

Engineering Big Data Analytics

Data Analytics Simulation

CAE Automotive
38

Always keep in touch with the latest news
Visit us on the Web!

Here you will find comprehensive information about HPC, IT solutions for the datacenter, services and
high-performance, efficient IT systems.

Subscribe to our technology journals, E-News or the transtec newsletter and always stay up to date.

www.transtec.de/en/hpc

39

Automotive
Risk Analysis

Simulation
Engineering CAE
Aerospace

Big Data Analytics

Price Modelling
Life Sciences
Risk Analysis

CAD
transtec Germany transtec Switzerland
Tel +49 (0) 7071/703-400 Tel +41 (0) 44/818 47 00
transtec@transtec.de transtec.ch@transtec.ch
www.transtec.de www.transtec.ch

transtec United Kingdom ttec Netherlands transtec France
Tel +44 (0) 1295/756 500 Tel +31 (0) 24 34 34 210 Tel +33 (0) 3.88.55.16.00
transtec.uk@transtec.co.uk ttec@ttec.nl transtec.fr@transtec.fr
www.transtec.co.uk www.ttec.nl www.transtec.fr

Texts and concept: Dr. Oliver Tennert, Director Technology Management HPC Solutions | Oliver.Tennert@transtec.de
Layout and design: Stefanie Gauger, Graphics Design | Stefanie.Gauger@transtec.de

© transtec AG, June 2012, rev. 1.1
The graphics, diagrams and tables found herein are the intellectual property of transtec AG and may be reproduced or published only with its express permission.
No responsibility will be assumed for inaccuracies or omissions. Other names or logos may be trademarks of their respective owners.

40

Ibm special hpc

More Related Content

Similar to Ibm special hpc (20)

More from TTEC (20)

Recently uploaded (20)

Ibm special hpc