Microsoft Powers the Move to the HPC Cloud at PASC18

Print Friendly, PDF & Email

Dr. Lukasz Miroslaw isn HPC & Big Compute Architect at Microsoft

In this video from PASC18 in Basel, Dr. Lukasz Miroslaw from Microsoft describes how the company is helping customers move HPC & Big Compute workloads to the Cloud.

“High performance computing, artificial intelligence, and other advanced workloads are increasingly important to companies across all industries. But rapid changes in technology and workloads make it challenging for HPC teams to provide the resources their users need. Microsoft Azure provides one cloud for every HPC workload so that users can get the resources they need when they need them and only pay for what’s used.”

High performance workloads require high performance compute resources. Azure provides resources with HPC in mind. Azure has the broadest array of NVIDIA GPUs–including the Tesla V100–for HPC, deep learning, and remote visualization workloads. With optional InfiniBand interconnects, you can scale up without compromising performance. For your most demanding HPC jobs, we have an exclusive partnership with Cray that allows you to get a dedicated, custom-build Cray supercomputer in an Azure datacenter.

For your loosely-coupled workloads, Azure offers a variety of virtual machines optimized for CPU, memory, or storage. The Fv2-Series virtual machines features Intel Xeon Platinum processors and our upcoming Lv2-Series feature AMD Epyc processors. With the flexibility and elasticity of the cloud, you can get exactly the resources you need and easily change your compute environment as your requirements change.

HPC workloads also require performant filesystems. The Azure Marketplace includes BeeGFS and Lustre. Many customers choose to run their preferred filesystems on storage-optimized virtual machines. And our recent acquisition of Avere Systems will bring a new option for high performance, hybrid file storage.

Ease of management

High-end resources are only part of the story. The resources also need to be easy to provision and manage. Azure has the tools you need. Azure Batch allows developers to create cloud-native applications. It provides job scheduling and resource management. With accelerators for popular AI and rendering applications, Batch enables you to focus on your workload, not your infrastructure.

Azure CycleCloud provides simple, managed access to compute resources. HPC admins can easily create and manage compute clusters in Azure. Templated configuration means you can configure each cluster the way you need it and then instantiate as many identical clusters as you need. Hybrid submission support, performance monitoring, and cost alerts give you the tools you need to manage Azure clusters as a production resource.

Your ecosystem

Azure virtual machines support CentOS, SLES, RHEL, Debian, Ubuntu, and Windows Server. You can run your applications on the ecosystem you’re used to. Fit your Azure environment to your needs, not the other way around.

Check out our insideHPC Events Calendar