Microsoft Corp.’s release Monday of Windows HPC Server 2008, following the beta release in late 2007, strives to make high-performance computing accessible to regular IT administrators who may not possess the specific expertise.
The release comes with the HPC Cluster Manager, a console that IT managers can use to view the entire cluster and perform tasks ranging from configuration to node management.
Essentially, administrators “can manage a cluster as if managing a typical server environment,” said Nik Garkusha, Microsoft Canada platform strategy manager. “That has not been previously possible.”
Among new features in the Cluster Manager is an SOA-focused job scheduler to let users run iterative, instead of sequential, tasks thereby reducing the computation time from hours to minutes, said Garkusha. Also, IT managers can easily manage all tasks, nodes by viewing the performance of all clusters at any given time.
Besides ease of use for the IT manager, the technology is meant to be easily adopted and managed by individual business departments that would need to run and manage computations on their own cluster. Typically, long queues can result when several departments in the organization attempt to run their computation models within the same time frame, said Garkusha, but the HPC Server 2008 “sits in an assembly that looks like your work station and uses your standard power.”
And, if a department requires more cores for added compute power, said Garkusha, they can “just spin it off to a bigger cluster.” He anticipates the departmental-level cluster will be desirable to the likes of stock brokers and game designers.
The new version is also scalable and designed to do away with the traditional high-powered and costly work stations that needed to be upgraded as technology changed, said Garkusha. “Well, now you can have a cluster that sits in a rack and today you can populate it with four blades and tomorrow you can add another four blades to improve your processing power,” he said.
A customer of Windows HPC Server 2008, Technicolor Canada Inc., a provider of video services for the communication, media and entertainment industries, has reaped the gains of lower cost and smaller staffing demands needed to run and manage typical HPC systems.
Furhermore, Tom Burns, Technicolor’s director of post production infrastructure, said given the rewards his company has seen, it wouldn’t be hard for IT departments to convince their organization’s management of the business benefits.
“It’s easier to get funding for a $25,000 cluster when you don’t have to pay huge consulting dollars for the deployment of a separate task scheduler and you don’t have to hire three rocket scientists to make it go,” said Burns.
Anthony Brown, managing director with Seven Group Inc., a Vancouver-based Microsoft partner that builds IT infrastructures for enterprises and small digital media content companies, said he’s seen HPC adoption expand from predominantly the domain of large enterprises with specialized expertise and the manpower to run it, to smaller less-equipped organizations like those with a thousand or less employees.
With Windows HPC Server 2008, specifically, smaller organizations are deploying clusters faster than ever before, he said, using the example of a school offering a digital media program where Seven Group recently deployed and began running the cluster within one day. And, said Brown, the customer already had the in-house expertise to manage the system.
Microsoft’s HPC release also provides “rapid” HPC application development through integration with Visual Studio 2008 for a “comprehensive parallel programming environment.” And, it supports standard interfaces like OpenMP, multiprocessor interconnect (MPI) and Web services, third-party numerical library providers, performance optimizers, compilers and debugging toolkits.
