We are excited to announce the extension of A8 and A9 support in Azure to include virtual machine (IaaS) options for these sizes running either Windows Server or Linux operating systems. Compute-intensive workloads can take advantage of bringing their own images to Azure and leverage the higher memory, increased number of cores per VM and the unique capability of a second backend network. This network supports remote direct memory access (RDMA) up to 40Gbps for lower latency and high-throughput communication between compute nodes. RDMA connectivity is available today only on Windows Sever virtual machines and requires the HpcVmDrivers VM extension. You can learn more about this extension, here, at the MSDN library.
These instances are designed for High Performance Computing (HPC) applications such as computational fluid dynamics, finite element analysis, and weather forecasting. Other industries will benefit from this new offering as well, including but not limited to manufacturing, energy exploration, life sciences, et. al.
"The new Azure A9 instances reduce our video transcoding time by 70% compared to A7 [instances]”
--Alexander Leschinsky, Managing Director, G&L Systemhaus, Germany
Today you can specify the A8 or A9 size when creating VMs through PowerShell or the Azure REST APIs. Soon you’ll see the sizes in the Azure Portal. More information on deploying these new instances can be found on the Azure blog. RDMA capabilities are accessed through the Network Direct interface. Network Direct is supported in Microsoft MPI (MS-MPI), and we are working with partners to enable Network Direct in other MPI stacks, as well as bringing RDMA capabilities to Linux virtual machines.
Customers have asked us to provide this capability in Windows Azure because their on-premises HPC clusters are at capacity or they need a way to scale models that are too large or complex to run on existing infrastructure, e.g. a workstation. HPC Pack 2012 R2 clusters are fully supported to be deployed in Azure with A8 or A9 VM types. Now you can shift your on-premises HPC clusters to Azure.
We’re excited to bring these expanded capabilities. You can get more information about Azure Big Compute, here, at the MSDN Library. To learn more about the initial offering of A8 and A9 sizes for PaaS workloads, view this January blog post. Contact us at firstname.lastname@example.org with your feedback and suggestions.
Can a new Virtual network be created that combines the A8/A9 with A1/A2/A3/A4 and allow the A8 to act as a SQL server running Enterprise SQL?
With Regional Virtual Networks it is now possible to combine instances sizes such as A8 and A9s in combination with A1, A2, A3s, etc.