Your Guide to the Latest Windows Server Product Information
For those of you interested in parallel computing and high performance computing, you'll take note that today we released the first public beta of V2 of Windows Server 2008 for HPC clusters.
The V2 product is called Windows HPC Server 2008. The product currently in the market is called Windows Computer Cluster Server 2003 (or Windows CCS). I'm told the product has been renamed to reflect an expanded feature set beyond compute for very scalable enterprise clusters. Some of those features include high-speed networking, support for clustered file system (e.g., IBM's GPFS, HP Polyserve, Panasas), new failover capabilities, enhanced management tools and a service-oriented architecture job scheduler. Windows HPC Server 2008 will be available in the 2nd half of 2008.
In the booth at Supercomputing 2007, we’re running a couple demos with mixed clusters (Linux and Windows). This is a first for Microsoft at a conference, and from what I’m told is linked directly to customer input from research, academia, life sciences industries. Here’s an excerpt from today’s news release:
Mixed, dual-boot clusters can also improve cluster efficiency. Because dual-boot clusters flexibly serve both Linux and Windows users, they increase utilization rates by expanding their number of addressable users. Examples of customers deploying large mixed clusters include the University of Iowa, Cambridge University, 3M and Baker Hughes Inc. Leading technology partners that have announced mixed cluster support for Windows HPC Server 2008 include Altair Engineering Inc., Cluster Resources Inc. and Platform Computing.
One demo station, titled “Virtualization for Mixed Clusters” assumes the customer isn’t running MPI, and has SLES running as a child partition on Hyper-V with Windows HPC Server 2008 parent partition. I’m told that this scenario is coming from customers who want to optimize for manageability and cluster utilization and not necessarily peak performance. The second demo station, titled “Linux and Windows: Mixed Cluster Management” assumes a dual-boot cluster (running SLES), storage from Panasas and management tools from Moab. I was told that this scenario could use Hyper-V. A gentleman from Moab explained that customers are seeing more demand for Windows-based apps running on HPC clusters, but the primary HPC environment is Linux. Moab’s tools let customers provision either OS with the chosen app – on an as needed basis. Very cool.
Last item to point out is the new Top500 supercomputer list. For those of you not familiar with this bi-annual benchmark, the Top500 list represents the 500 most powerful computers in the world. We’re talking trillions of computations per second. And for the first time ever, there were six Windows-based clusters on the list. While this seriously pales in comparison to the 426 Linux-based clusters, it’s certainly a start. The most powerful Windows-based cluster is owned by Microsoft in our Rainier (Washington) datacenter, and measured in at 11.7 teraflops. And by upgrading this cluster to Windows HPC Server 2008 (same hardware, new OS), I’m told the Linpack benchmark results improved 30%, and more importantly, ran the test in 2 hours.
To follow news from the Supercomputing 2007 conference, visit http://insideHPC.com or www.winhpc.org.
PingBack from http://linux.x7g.net/2007/11/13/download-beta-of-new-windows-hpc-server-2008/
back to Redmond late yesterday from Reno where I was attending Supercomputing 2007 conference. I suspect SC07 will be best remembered for the power outage that hit the convention center and most of downtown Reno, and Ashlee's stellar headline. While I