Microsoft on Tuesday hit another high-performance computing milestone by placing its serverfor the first time in the top 10 on the list of the Top 500 super computers as judged by Top500.org.
Just a year ago, the best Microsoft could do was 116th place based on rankings from Top500.org, which has been benchmarking supercomputers since 1993 with its bi-annual tests it calls “runs.”
Download the latest Network World Executive Guide - Virtualization Reality Check
Windows HPC Server 2008, a 64-bit system that shipped Nov. 1, came in at No. 10, achieving 180.6 teraflops with 77.5% efficiency at the Shanghai Supercomputer Center and Dawning Information Industry Co.
Despite the high ranking, Microsoft’s biggest high-performance computing challenge is likely in front of the vendor – creating easy-to-use developer tools for writing applications for the platform.
The official launch of HPC 2008 Server will be September 22, blogger Mary-Jo Foley reports. It will launch as part of the 2008 “High Performance on Wall Street” conference in New York.
Hard to believe Microsoft needs yet another version of its flagship server operating system but so be it. This variant went into its final “beta” in June, 2008, and has been mostly overlooked in terms of hype, says blogger Alex Lewis. Features include more management tools, job scheduling and built-in clustering … or for a full list view the chart below.
Feature
Implementation
Benefit
Operating system
Windows Server 2008
Inherits security and stability features from Windows Server 2008.
Processor type
x64 (AMD64 or Intel EM64T)
The large memory model and processor efficiencies of x64 architecture.
Node deployment
Windows Deployment Services
Image-based deployment, with full support for multicasting.
Head node redundancy
Windows Failover Clustering and SQL Server Failover Clustering
Requires Windows Server 2008 Enterprise and SQL Server Standard, but provides a fully redundant head node and scheduler.
Management
New Integrated Administration Console
Integrates all aspects of node and job management, grouping, monitoring at-a-glance, diagnostics, and reporting in a single application.
Network topology
Network Configuration Wizard
Fully automated, with a more intuitive interface, the Network Configuration Wizard facilitates configuring the cluster’s network topology.
Application network
MS-MPI
New, high-speed application network stack using NetworkDirect drivers. New shared memory implementation for multicore processors. Highly compatible with existing MPICH2 implementations.
Scheduler
Job Manager Console
GUI console is integrated into Administration Console, or is stand-alone. Command line supports Windows PowerShell scripting and legacy command-line scripts from Windows Compute Cluster Server (the first version of Windows HPC offering). Greatly improved speed and scalability. Support for SOA applications.
Reporting
Integrated into Administration Console
Standard, prebuilt reports. Extensibility features allow using SQL Server Analysis Services for additional analysis. Historical performance charts.
Monitoring
Integrated into Administration Console
Heat map on cluster or node group. Per node charts. Cluster-wide performance overview.
Diagnostics
Integrated into Administration Console
In-the-box verification tests and performance tests. Store, filter, and view test results and history.
Source: Microsoft
In June, Alex Lewis lamented HPCs lack of software support, but Microsoft marketing materials list about 30 HPC partners when including hardware and software vendors that have been on board with Microsoft Windows Compute Clusters Server 2003. The assumption is that these same makers will offer support of the WS2008 version of HPC, though it is difficult to find any materials which document who, what, when, why. More news on that will no doubt come during the New York gala. Stay tuned.
The Linpack Benchmark
As a yardstick of performance we are using the `best’ performance as measured by the LINPACK Benchmark. LINPACK was chosen because it is widely used and performance numbers are available for almost all relevant systems.
The LINPACK Benchmark was introduced by Jack Dongarra. A detailed description as well as a list of performance results on a wide variety of machines is available in postscript form from netlib. Here you can download the latest version of the LINPACK Report: performance.ps. A parallel implementation of the Linpack benchmark and instructions on how to run it can be found at http://www.netlib.org/benchmark/hpl/.
The benchmark used in the LINPACK Benchmark is to solve a dense system of linear equations. For the TOP500, we used that version of the benchmark that allows the user to scale the size of the problem and to optimize the software in order to achieve the best performance for a given machine. This performance does not reflect the overall performance of a given system, as no single number ever can. It does, however, reflect the performance of a dedicated system for solving a dense system of linear equations. Since the problem is very regular, the performance achieved is quite high, and the performance numbers give a good correction of peak performance.
By measuring the actual performance for different problem sizes n, a user can get not only the maximal achieved performance Rmax for the problem size Nmax but also the problem size N1/2 where half of the performance Rmax is achieved. These numbers together with the theoretical peak performance Rpeak are the numbers given in the TOP500. In an attempt to obtain uniformity across all computers in performance reporting, the algorithm used in solving the system of equations in the benchmark procedure must conform to LU factorization with partial pivoting. In particular, the operation count for the algorithm must be 2/3 n^3 + O(n^2) double precision floating point operations. This excludes the use of a fast matrix multiply algorithm like “Strassen’s Method” or algorithms which compute a solution in a precision lower than full precision (64 bit floating point arithmetic) and refine the solution using an iterative approach.
Microsoft Breaks Into Top 10 of World’s Most Powerful Supercomputers
AUSTIN, Texas — Nov. 18, 2008 /PRNewswire/ — Today at the Supercomputing 2008 conference, Microsoft Corp. debuted in the top 10 of the world’s most powerful supercomputers with Shanghai Supercomputer Center and Dawning Information Industry Co. Ltd., which ranked at No. 10 with 180.6 teraflops, the parallel computing speed, and 77.5 percent efficiency. A truly incredible achievement considering that 12 months ago in Reno, Nevada, Microsoft was at 116 on the Top500 list at Top500.org. This is on the heels of Windows HPC Server 2008 releasing to the manufacturing industry in September.
Reduces costs and complexity of high-performance computing
Windows HPC Server 2008 makes supercomputing more accessible to end users by allowing them to harness computing power through a familiar Windows desktop environment. It also reduces the complexity of Top500 runs and increases efficiency. Microsoft announces the availability of the Top500 Excellence Kit, which includes a Top500 guide containing best practices and internal knowledge from Microsoft developers on how to achieve the highest-efficiency LINPACK runs. As a part of the kit, Microsoft is including several management and performance tools used in its 180.6-teraflop run, a High-Performance LINPACK (HPL) Wizard that automatically tunes HPL for your cluster environment. More information is available at http://windowshpc.net/Resources/Pages/Programs.aspx.
A broad platform for software vendors and an expanded playing field for hardware manufacturers
Microsoft and Cray Inc. teamed up in September with an announcement to drive high-productivity computing further into the mainstream in a broad array of markets with the Cray CX1 supercomputer. Now they’re giving one away! More information is available at http://www.superduosupersweeps.com.
Deep investments in HPC and commitment to driving innovation
Parallelizing code is not easy given that programming languages, frameworks, developer tools, and even the majority of developers have grown up in a largely serial age. So the software development industry is taking strides to make parallelism more accessible to developers, and Microsoft is leading that charge. With Visual Studio 2010, Microsoft is delivering the first wave of powerful developer tools such as Task Parallel Library, Parallel LINQ and Coordination Data Structures for managed code to ease the transition to parallel code. These technologies, along with MPI, MPI.Net and Cluster-SOA, extend parallelism to clusters of thousands of nodes using Windows HPC Server 2008. More information on taking parallelism mainstream is available at http://msdn.microsoft.com/en-us/concurrency/default.aspx.
Executive Quotes
Vince Mendillo, director of the HPC Division at Microsoft:
“Beyond the top 10 number itself there are two things that are very impressive — Windows HPC Server 2008 worked at this scale, and we demonstrated that we have expertise inside Microsoft to lead the integration of a top 10 HPC system. We have come a long way in the past four years from building our first 10-gigaflop cluster to the top echelon of the world’s fastest supercomputers.”
Yao Jifeng, technical director of HPC at Shanghai Supercomputer Center:
With the adoption of Windows HPC Server 2008 for our Top500 run, Shanghai Supercomputer Center was able to accelerate the pace of technology innovation in China. Our growing partnership with Microsoft benefits the many colleges, universities, research institutes and corporations that are our customers.”
Sha Chaoqun, vice general manager at Dawning:
“During our Top500 run we achieved 180.6 teraflops and 77.5 percent efficiency while using the familiar and intuitive Windows operating system. We chose to do the Top500 on Dawning 5000A run using Windows HPC Server 2008 for two reasons: performance and ease of use. Our growing partnership with Microsoft helps Dawning accelerate the pace of technology innovation in China.”
Sash Sunkara, vice president of marketing at Mellanox Technologies:
“We are pleased to partner with Microsoft to deliver the world-leading performance, efficiency and scalability of Mellanox InfiniBand while keeping the simplicity and the ease-of-use capabilities of Windows Server platforms. The tight integration of Windows HPC Server 2008 and Mellanox InfiniBand end-to-end hardware and software solutions not only enabled the tremendous performance achievement at the Shanghai Supercomputer Center, but also ensured the needed application productivity at scale.”
