I’m pleased announce that Service Pack 1 for HPC Pack 2012 is available for download.
Our focus for SP1 is Azure Burst scale and robustness. A principal of the cloud is to design for failure. We’ve learned a lot from working with customers over the past two years about how to troubleshoot and handle failures in communication, provisioning, or hardware. Our ongoing testing deploys thousands of cores, runs jobs of varying length to, simulates customer scenarios, and injects failures across the stack over and over again. Additional instrumentation and logging help make troubleshooting easier when this is necessary. We feel good supporting our customers that are pushing the limits.
HPC Burst to Azure now supports using the A6 and A7 high memory VMs with 7 GB of RAM per core. A6 is 4 cores with 28 GB of RAM, and A7 is 8 cores with 56 GB of RAM.
With SP1, we introduce two new user roles for managing the cluster--job administrator and job operator. This has been one of our top feature requests. Job administrators can perform all the functions of any job owner as if they were the admin, but the role is not allowed to manage resources. Think of this as a queue administrator. The job operator restricts things further to view, cancel, finish, or re-queue any job.
Also in SP1 is HPC Services for Excel support for Microsoft Excel 2013, updated cluster debuggers for SOA and MPI applications, iSCSI deployment for diskless compute nodes, fine-tuning of heartbeats, and graceful preemption for balanced mode scheduling.
We encourage all customers using HPC Pack to burst to Windows Azure to upgrade to the 2012 SP1 release. There are significant improvements over our 2008 R2 service pack releases.
The team is working on some how-to guides that we’ll post over the next few months. Let us know if there some specific topics that you would like to see us cover.
Alex Sutton
Group Program Manager, Big Compute