High performance computing scenarios
Azure provides on-demand and scalable compute resources that enable you to run large-scale HPC workloads in the cloud.
Take advantage of several HPC solutions available to you in Azure to add more capacity to your existing on-premises HPC cluster, or to run your HPC workloads entirely in the cloud. With Azure, you can readily scale resources up or down, while taking advantage of advanced compute and networking infrastructure that has been specifically configured to run even the most demanding HPC applications. This combination of flexibility and performance helps you to run your workloads on demand, paying only for the time when you are using the resources.
The scenarios listed on this page can help you to select the Azure solutions that best meet your needs and those of your organization.
Deploy a cluster in the cloud
Use scripts and Azure Resource Manager (ARM) templates to quickly and consistently deploy your applications on Windows or Linux infrastructure as a service (IaaS) virtual machines, including your own custom images, and then run HPC workloads by using the job scheduling solution of your choice. Define dependency chains so that you can have the head node up and running before the compute nodes are deployed, whether it is just a few nodes or thousands of nodes. Storage accounts, network resources, cloud services, and other resources can be provisioned and configured in an easy, repeatable, and extensible way.
This scenario can be very suitable if you are using an existing on-premises HPC system and you want to replicate that system in the cloud. Also, this scenario can be beneficial if you want to deploy a new HPC system but do not want to incur on additional hardware expenses.
Get started:
Learn more:
Run your workload on-premises and in the cloud
Combine your on-premises resources with additional Azure resources in a hybrid solution that enables you to easily deploy additional compute resources only when you need them, or efficiently increases capacity on your existing HPC cluster.
Burst to Azure with HPC Pack
Add platform as a service (PaaS) compute resources in Azure only when you need them by leveraging solutions like Microsoft HPC Pack. With HPC Pack you can deploy tens, hundreds, or thousands of additional compute nodes in minutes, on demand or on a schedule. And with the true high-performance computing capabilities available with Azure A8 and A9 high performance compute instances, take advantage of a backend network with remote direct memory access (RDMA) technology that provides latency of under 3 micro seconds.
This cloud burst scenario is a great option to quickly and easily deploy and stop cloud resources, for example when you want to respond to unexpected or cyclical spikes in demand. It is also well suited for shifting workloads to the cloud that you don’t want to, or can’t, run on-premises (for example, due to resource constraints, or to address a backlog of work). Additionally, HPC Pack is a proven and complete solution to run your HPC workloads on existing Windows or Linux on-premises resources.
Get started:
Learn more:
Extend your cluster with Azure VMs
Extend your on-premises HPC or grid computing solution to Azure without incurring more infrastructure costs by deploying Windows or Linux IaaS virtual machines as additional compute resources. Use your existing cluster management tools. Connect the on-premises and cloud nodes through a virtual network with a virtual gateway, or create an extremely fast and secure dedicated ExpressRoute connection between your on-premises cluster and Azure.
This scenario is well suited for existing HPC workloads that can’t take full advantage of PaaS resources, as well as for shifting workloads to the cloud that you don’t want, or can’t, run on-premises (for example, due to resource constraints, or to address a backlog of work).
Learn more:
Run your applications at scale on Azure Batch
By leveraging a fully managed cloud service like Azure Batch you can readily cloud-enable any application that can scale, without having to deploy and manage an HPC system, or having any previous experience of how such systems work. This helps you to focus on your applications and algorithms, and not on managing compute and job scheduling infrastructure. With Batch, you describe the data that needs to be moved to the cloud for processing, how the data should be distributed, and what parameters to use for each task. Batch easily orchestrates the execution of your application across the compute resources.
This scenario is well suited for new and existing enterprise applications that you want to move to the cloud. It is also a great alternative for software vendors and integrators that want to offer their software as a service (SaaS).
Get started:
Learn more:
Visual Studio subscriber?
Get up to $1800 per year of additional Azure services
Startup?
Join the BizSpark program and get free Azure services
Academic use?
Schools and institutions can receive special benefits