The Azure Big Compute team is attending SC14 this week, the annual conference for high performance computing and analysis. It’s been a great year for HPC on Azure, and we look forward to catching up with customers and partners about doing large-scale compute in the cloud.
Just two weeks ago, we released the preview of Azure Batch. This new managed service provides “job scheduling as a service” with auto-scaling of compute resources, making it easy to run large-scale parallel and high performance computing (HPC) work in Azure. We have new customers building large-scale cloud solutions with Batch in a range of industries including genomics, financial risk, and image progressing. One customer was amazed by how easily his organization could use 10,000 cores, and another customer is eager to test with 150,000 cores. Whether your need is large or small, we’re ready for you.
HPC on Azure
The SC conference is a great time to talk about performance. The A8 and A9 VMs in Azure is the only public cloud offering with low-latency networking using Mellanox QDR InfiniBand and remote direct memory access (RDMA). With these VMs, we provide you bare-metal performance for the high performance applications you run today, on demand in the cloud.
Intel Corporation is one of our key collaborators for HPC in Azure. We work closely with them on performance, and use Intel processors for the A8 and A9 VMs. Many of our customers use Intel MPI in their applications and want to use the cloud to run additional jobs. We are excited to announce that the Intel MPI Library now runs on Microsoft Azure and takes full advantage of the InfiniBand network and its RDMA capabilities. Stop by our booth at SC14 and we’ll show you the great performance for MPI applications running on Microsoft Azure.
Flexera Software recently announced that FlexNet Licensing is supported to run in Microsoft Azure. FlexNet Licensing is used in over 20,000 applications to enable software monetization, including many important to high performance computing (HPC) users. This support helps software vendors offer new licensing models in the cloud for on-demand and peak use. Customers are also able to bring their existing license servers and expand their cluster capacity into the cloud. ANSYS is one of the first software vendors to take advantage of FlexNet Licensing in Azure and our great MPI performance.
No matter what commercial or open source tools you want to use, Windows or Linux, Azure is a great place to run all of your applications. In our booth at SC14, we’ll be showing a genomics analysis application from Affymetrix running on Linux. Also joining us for a demo is Univa, showing Grid Engine™ running in Azure, scheduling work on a Linux cluster built with Azure virtual machines. We’re delighted that Univa also now has a native Windows version of Grid Engine™ available.
New HPC Pack and MS-MPI Update
I’m excited to announce that Update 1 for HPC Pack 2012 R2, and a new version of Microsoft MPI (MS-MPI) are now available for download. HPC pack is Microsoft’s free cluster manager and job scheduler. HPC Pack makes it easy to set up an on-premises cluster, and to extend that cluster into Azure. We call this Azure Burst. A wizard takes care of setting up the virtual machines, and the HPC job scheduler runs the jobs connecting over the Internet, VPN, or ExpressRoute.
Update 1 also makes it even easier to deploy a cluster completely in the cloud. Start with the HPC Pack VM Image from the Azure Marketplace, and use PowerShell scripts to deploy a cluster and compute nodes. It’s the easiest way to get started running HPC jobs in Azure.
These improvements are part of our ongoing efforts to make it as simple as possible for customers to achieve big things with the power and scale that Azure provides for both hybrid and cloud-native solutions. We’re working closely with our early adopter customers and have exciting plans for 2015. We invite you to come and meet with us at SC14, brainstorm with us about what you want to do with HPC In the cloud, or talk with us in the Azure Batch Forum.
Alex Sutton
Azure HPC and Batch Computing Team