Windows HPC Server 2008 provides a productive, cost-effective and high-performance computing (HPC) solution that runs on x64-bit hardware. Windows HPC Server 2008, can be deployed, managed and extended using familiar tools and technologies.
 Solution
Windows HPC Server 2008 combines the power of a Windows 64-bit server platform with
rich, out-of-the-box functionality to improve the productivity, and reduce the complexity,
of your HPC environment. Windows HPC Server 2008, provides a comprehensive set of
deployment, administration, and monitoring tools that are easy to deploy, manage, and
integrate with your existing infrastructure.
Windows HPC Server 2008 enables broader adoption of HPC by providing a rich and
integrated end-user experience scaling from the desktop application to the clusters. A
wide range of software vendors, in various verticals, have designed their applications to
work seamlessly with Windows HPC Server 2008 so that users can submit and monitor
jobs from within familiar applications without having to learn new or complex user
interfaces.
Developing parallel programs requires integrated development environments along
with support for distributed computing standards. Visual Studio® 2008 provides a
comprehensive parallel programming environment for Windows HPC Server 2008. In
addition to supporting OpenMP, MPI, and Web Services, Windows HPC Server 2008 also
supports third-party numerical library providers, performance optimisers, compilers, and
a native parallel debugger for developing and troubleshooting parallel programs.
Enhanced Productitivy
Windows HPC Server 2008 allows you to accomplish more, in less time, with reduced
effort, by leveraging users' existing skills
and integrating with the tools already in use in your enterprise. Administrators
and developers can save time with a common productivity and development
platform across technical workstation and cluster. Administrators can use existing
Microsoft management tools to centrally manage their entire Windows Server
infrastructure. Command line interfaces are fully supported for administrators,
and end users are not required to master command-line interfaces (CLIs) to harness
the power of HPC. Application developers can work with familiar development tools,
such as Visual Studio's native parallel debugger, to develop and troubleshoot
parallel programs. Teams can manage their projects and
collaborate using workflows by using Microsoft Office SharePoint® Server 2007.
Scalable Performance
Windows HPC Server 2008 is built on proven Windows Server® 2008 x64-bit technology.
Windows HPC Server 2008 can efficiently scale to thousands
of processing cores and includes management tools that help system
administrators proactively monitor system health and maintain system stability.
Integration with Windows Server 2008 Enterprise and Microsoft® Windows SQL
Server® provides failover capabilities in the event of system failure.
Quickly Deploy a Manageable Infrastructure
Windows HPC Server 2008 includes wizards, tools, built-in
management consoles, and a To Do List
to simplify the provisioning of compute nodes while
enhancing Windows Server 2008 Deployment Services
to allow you to create, modify, and deploy systems
images across a cluster. New management tools and
enhancements include the following:
NEW!
The time and effort for setup and configuration
are reduced through the use of system templates.
Templates allow an IT professional to create standardized
system images or apply patches that can be consistently
deployed across the cluster.
ENHANCED!
Operating system images can quickly be
deployed using Windows Deployment Services (WDS).
Progress can be monitored through the new HPC Pack
Administration console.
ENHANCED!
Manage clusters programmatically though
command-line interfaces, PowerShell, or other scripting
languages such as Perl.
Built-in wizards, tools, management consoles, and a To Do List help simplify the provisioning of compute nodes so that consistent system images can be created, modified, and deployed across a cluster.
Monitoring, System Health and Reporting
Built on Windows Server 2008 64-bit technology,
Windows HPC Server 2008 includes scalable management
tools built using the new System Center interface.
The heat map (see below) provides an overview of
system utilisation. New tools and enhancements for
administrators include:
NEW!
Windows Server 2008 Enterprise includes Failover
Services. The combination of Windows Failover Services
and SQL Server database clustering provide head node
redundancy in the event of a hardware failure.
NEW!
Administrators can create groups based on
hardware, software, or networking characteristics.
Groups allow for simplified administration and allow
optimal utilisation by mapping jobs requirements to an
appropriate set of servers.
NEW!
Built-in diagnostic tools allow administrators
to quickly identify and diagnose hardware, software,
or network problems across the cluster. Additional
capabilities such as: automating routine, redundant tasks,
and intelligent reporting and monitoring are possible
through the use of System Center Operations Manager
2007.
The Heat Map provides at-a-glance views of cluster health, and helps to quickly identify and diagnose hardware, software, or network issues across the cluster.
A Foundation for High-Performance Service Oriented Applications (SOA)
The Microsoft HPC Pack 2008 includes a flexible job
scheduler that includes a command-line, graphical user
interface (GUI), and application programming interface
for submitting jobs to the cluster. The 2008 job scheduler
addresses both batch and newer service oriented
applications. It can also be deployed together with partner
job schedulers for additional advanced policies or mixed
environments. New job scheduling tools and enhancements
include:
NEW!
Job scheduling supports new SOA workloads.
Compute nodes can communicate with the submitting
systems through Windows Communication Foundation
(WCF) brokers. WCF brokers are dedicated nodes that
scale out cluster performance, and act as proxies facilitating
communication between public network clients and
compute nodes on the private networks. Additional WCF
brokers can be added to scale as required.
ENHANCED!
The HPC Pack job scheduler is multi-core
aware, allowing sophisticated scheduling by core, by socket,
or by server. The job scheduler supports existing policies
such as: priority-based first come, first served (FCFS),
backfilling, non-exclusive and license-aware scheduling, and
also includes new policies for job profiling, preemption, and
growing and shrinking of jobs.
NEW!
The HPC Pack reduces the complexity of integrating
with existing clusters by supporting industry standards
such as the Open Grid Forum's (OGF) HPC Profile for job
scheduler interoperability. Additionally, the Subsystem for
UNIX-based Applications (SUA) integrates applications to
maximize existing investments while extending UNIX- and
LINUX-based applications to Windows systems.
Networking and High-Speed Interconnects
Windows HPC Server 2008 significantly reduces the
complexity of deploying multiple networks across a cluster
by providing an updated networking wizard, new network
diagnostics tools, and a simplified way of performing
network driver management. New functionality and
enhancements include:
ENHANCED!
The Microsoft Message Passing Interface
(MS MPI) is based on the Argonne National Labs
implementation (MPICH2) of the MPI2 standard. MS MPI
can utilise any interconnect that is supported on Windows
Server 2008.
NEW!
NetDirect is Microsoft's new Remote Direct Memory
Access (RDMA) interface for high-speed, low-latency
networks such as those running on Infiniband, 10 Gigabit
Ethernet and Myrinet. With an architecture that directly
bypasses OS and TCP/IP overhead, NetDirect achieves
better performance for massively parallel programs that
can utilise very low-latency, high-bandwidth, and enables
efficient CPU utilization.
A networking wizard and network diagnostics tools, plus simplified driver management, help to reduce the complexity of deploying multiple networks across a cluster, while use of NetworkDirect helps to achieve better performance and enable efficient CPU utilization.
Windows HPC Server 2008 Architecture
The Windows HPC Server 2008 head node:
- Controls and mediates all access to the cluster resources.
- Is the single point of management, deployment, and job
scheduling for the cluster.
- Can failover to a backup head node in the case of failure.
Windows HPC Server 2008 uses the existing corporate
infrastructure and Microsoft Active Directory® for:
- Security
- Account management
- Operations management using tools such as Systems Center Operations Manager 2007.
Windows HPC Server 2008 takes advantage of the overall Microsoft ecosystem to improve the functionality, usability, and manageability of the cluster.
Interactive Sessions Through the WCF
An interactive session through the WCF includes the following steps:
- Clients create a session by specifying the WCF service
used to perform the calculation.
- The job scheduler assigns a broker and launches a
service instance pool on multiple nodes.
- Client connects to the broker and the job scheduler
provides the End Point Reference (EPR) of the broker to
the client.
- Client connects to the EPR.
- Standard WCF request/response messages occur.
Windows HPC Server 2008 System Requirements
The minimum system hardware requirements are similar
to the hardware requirements for Windows Server 2008,
Windows HPC Server 2008 supports up to 128 gigabytes
(GB) of RAM. Supported processors include AMD Opteron,
AMD Athlon 64, Intel Xeon with Intel EM64T, and Intel
Pentium with Intel EM64T.
|
CPU Requirement | x64 Architecture computer with Intel Pentium
or Xeon family processors with Intel Extended
Memory 64 Technology (EM64T) processor
architecture; AMD Opteron family processors;
AMD Athlon family processors; compatible
processor(s).
| |
Multiprocessor
support |
Windows Server 2008 HPC Edition and Windows
Server 2008 Standard Edition support up to four
processors per server. Windows Server 2008
Enterprise Edition supports up to eight processors
per server.
| |
Minimum RAM | 512 MB
| |
Maximum RAM | 128 GB
| |
Minimum Disk
Space for Setup
| 50 GB
| |
Disk Volumes | A single system volume is required for the
head and compute nodes. Redundant array of
independent disks (RAID) is supported but not
required. The system volume must be MBR.
Additional volumes can be MBR or GPT.
| |
Network Interface
Cards |
At least one network interface card (NIC) is
required. If a private network is used, the head
node requires at least two NICs, and compute
nodes at least one. Each node may also require a
high-speed NIC for a MPI network.
|
|
 More Information |
Products
Bright Cluster Manager™
