The essence of High Performance Computing (HPC) is processing power, and Windows Azure’s ability to provide on-demand access to a vast pool of computing resources will put HPC at the fingertips of a broader set of users.  With this in mind we're excited to announce the release of a second ‘service pack’ (SP2) for Windows HPC Server 2008 R2.  This release is focused on providing customers with a great experience when expanding their on-premises clusters to Windows Azure, and we’ve also included some features for on-premises clusters. Here’s an overview of the new functionality:

Windows Azure integration gets better

We’ve continued our work to provide a single set of management tools for both local compute nodes and Windows Azure compute instances. We’ve now integrated directly with the Windows Azure APIs to provide a simplified experience for provisioning compute nodes in Windows Azure for applications and data. Scaling on-premises cluster workloads to the Cloud is easy for administrators to configure, while ensuring complete transparency to end users. Furthermore, our new features include a tuned MPI stack for the Windows Azure network, support for Windows Azure VM role (currently in beta), and automatic configuration of the Windows Azure Connect preview to allow Windows Azure based applications to reach back to enterprise file server and license servers via virtual private networks.

Parametric sweep applications

Not all HPC applications are MPI applications. The most significant growth in the HPC market comes from conveniently parallel applications like parametric sweep applications and the WCF-based cluster SOA model to build scale-out calculation services. These programming models are commonly used in Monte Carlo analysis or risk calculation in finance. Applications using these scale-out models are ideal for the Cloud because they can typically scale linearly. Add more resources and get your work done faster!

Support for LINQ to HPC beta

In the coming days, we will release a second beta for building data intensive applications (previously codenamed Dryad). We’re now calling it ‘LINQ to HPC’. LINQ has been an incredibly successful programming model and with LINQ to HPC we allow developers to write data intensive applications using Visual Studio and LINQ and deploy those data intensive applications to clusters running Windows HPC Server 2008 R2 SP2. Programmers can now use thousands of servers, each of them with multiple processors or cores, to process unstructured data and gather insights.

New scenarios for on-premises clustering

We also added several features in SP2 that will enable new scenarios for on-premises cluster customers:

  • Resource pools allow partitioning cluster resources by groups such that users submitting jobs to their group’s pool are guaranteed a minimum number of resources, but idle resources can be shared with other groups when they have more work to compute.
  • A new REST API and web portal makes it easy to submit jobs from any platform and monitor the status from a user “heat map.”
  • Soft card authentication support enables users to submit jobs to a cluster in organizations that use smart cards instead of passwords.
  • Expanded support for adding Windows 7 desktops to a cluster for “desktop cycle stealing.” These resources can now reside in any trusted Active Directory domain.
  • New APIs that support staging and accessing common data that is required by all calculation requests within a session. The client application can include code that sends the data to the cluster and makes it available to each instance of the SOA service on the cluster.

Be sure to learn about all of the features in the release and try it out. I’m sure you’ll be pleased to see what we have in store for you!

Ryan Waite, Partner Director at Microsoft


Tweet