The beta is the first release candidate (RC), which Microsoft says is feature complete. An RC is a final beta before the code is considered finished. Microsoft officials said they plan to have one more RC before development is complete.
Microsoft made the announcement at the start of the International Supercomputing Conference in Dresden, Germany, where it also placed for the first time in the Top 25 of the world’s largest supercomputers.
Windows High Performance Computing (HPC) Server 2008 is Microsoft’s entry into the battle with Linux to provide platforms for research and other compute-intensive workloads. Linux is clearly the dominant player in the market with Microsoft battling to prove its mettle. The previous version of HPC was originally called Windows Compute Cluster Server (CCS) 2003. It rose from a Microsoft Research project introduced in 2000. CCS 2003 shipped in August 2006.
HPC Server 2008, which is based on Windows Server 2008, features high-speed networking, cluster management tools, advanced failover capabilities, a service-oriented architecture (SOA) job scheduler, and support for third-party clustered file systems.
The server, built to scale to thousands of cores, also includes a new high-speed NetworkDirect RDMA, Microsoft’s new remote direct memory access interface, and cluster interoperability through standards such as the High Performance Computing Basic Profile specification produced by the Open Grid Forum.
The platform combines into a single package the operating system with a message passing interface and a job scheduler developed by Microsoft.
Microsoft also plans to integrate HPC Server with its System Center tools for application-level monitoring and rapid provisioning, and SQL Server Reporting Services for capacity planning and auditing.
The Top 25 cluster powered by HPC Server 2008 was built by the National Center for Supercomputing Applications (NCSA) using a beta version of the software. The cluster came in at No. 23, achieving 68.5 teraflops and 77.7% efficiency on 9,472 cores.
The most powerful supercomputer was IBM’s $100,000 million Roadrunner system at the U.S. Department of Energy’s Los Alamos National Laboratory. Roadrunner achieved performance of 1.026 petaflop/s, the first supercomputer to hit that performance mark.
In addition, computer scientists at Umea University in northern Sweden, showed a supercluster featuring HPC Server 2008 running for the first time publicly on IBM hardware. The cluster, running on 672 IBM blade servers, achieved 46 teraflops and 85.5% efficiency on 5,376 cores.
Microsoft, however, isn’t looking to just be a workhorse.
“Our goal with HPC is not just the speeds and feeds,” said Bill Hilf, general manager of platform strategy at Microsoft. “It is really to bring the classic Microsoft playbook into the picture, which is simplify computing in this space.”
Hilf says NCSA was impressed with the fact it was able to deploy 1,200 nodes in less than four hours.
As part of the effort to simplify the technology and bring HPC out of its small niche, Microsoft wants to make it available to a wider set of users. Last November, Microsoft unveiled the Parallel Computing Initiative, which focuses on multicore desktops and clusters. The initiative adds Parallel Extensions to the .Net Framework that enable developers to improve efficiency and scalability of parallel applications. In addition, Microsoft and Intel recently announced a joint investment to create two new Universal Parallel Computing Research Centers at the University of California, Berkeley, and the University of Illinois at Urbana-Champaign.