Introducing High performance computing

I should write about this.


Last one and half months of my professional life have been just amazing . The day i was introduced to microsoft high performance computing.

It's thrilling. There's birth and death and frustration and victory in learning HPC. It's like a little microcosm of life is built into it .

Some applications run just fine on a single computer. For others, though, one computer isn’t enough. To run fast enough to be useful, these applications need to be broken into many different components, then have each of those components run simultaneously across many different machines. An application that needs to do lots of processing, for example, can benefit from using many CPUs in parallel, while an application that needs to access a large amount of data can improve its performance by using many disk drives at once.



When an application needs simultaneous access to many CPUs, many disk drives, or both, it can benefit from running on a cluster. A cluster is a group of computers, each with processing power and storage, capable of running an application in parallel across those machines. In virtually every case, the goal of using a cluster is to make applications run faster.

To support Windows-based clusters, Microsoft provides Windows HPC Server 2008 R2. This technology offers a way to create and manage clusters built using Windows servers, Windows desktop workstations, and Windows Azure instances in the public cloud. As its name suggests, the goal is to support high-performance computing (HPC), a category that includes a range of applications that can benefit from being distributed across a cluster. To allow this, Windows HPC Server 2008 R2 Service Pack (SP) 2 supports four major types of parallel applications.


Clusters are used for a range of problems today. Traditionally, cluster computing has been synonymous with certain kinds of HPC, primarily scientific and technical applications such as simulating car crashes or modeling nuclear reactions. These kinds of workloads are still an important aspect of cluster computing, but they’re no longer the whole story. The applications that run on clusters today are broader than this, spreading out to address more business-oriented problems. Yet whatever a cluster-aware application does, the cluster it runs on has common requirements for management, job scheduling, and more. It makes sense to provide these in a single product, such as Windows HPC Server, then support a range of applications on this common foundation.


And with today’s broader notion of cluster, which includes desktop workstations and cloud instances as well as traditional on-premises servers, adopting this technology has gotten easier. Organizations need no longer make a large up-front investment in servers just to get started with cluster computing. Combined with the increasing breadth of applications supported on clusters—they’ve moved well beyond traditional HPC—this lower barrier to entry suggests an ever-broader role for this technology. Long relegated to a supporting role, cluster computing just might be headed for center stage.