Altair’s industry-leading PBS Professional workload manager is designed to improve productivity, optimize utilization and efficiency, and simplify administration for all kinds of workloads — from the biggest high-performance computing (HPC) loads to millions of small, high-throughput jobs. PBS Professional makes cloud access easy with an intuitive GUI built right in, and now it’s integrated with the Azure CycleCloud HPC workload orchestrator.
Azure High-Performance Computing (HPC)
Adopting cloud resources for HPC and AI workloads, both in cloud-native and bursting scenarios, can deliver significant business value and a low total cost of ownership (TCO). Azure HPC is a complete set of computing, networking, and storage resources integrated with workload orchestration for competitive price/performance compared to on-premises infrastructure.
Modeling and Simulation in the Cloud
Today’s enterprises use simulation and modeling to speed up the design process and deliver better products and services. They often accomplish this using digital twins, where the creation of a virtual entity representative of a product, process, or service allows optimization and tuning in a virtual environment, facilitated by continuous data exchange with the physical world. HPC and AI workloads are key enablers, allowing evaluation and prediction in a safe test bed.
Provisioning on-premises resources for these workloads can lead to forecasting, financial, and technical management challenges. Cloud HPC resources can enable faster time to market, flexibility in exploring new business streams, agile scaling based on market demand, reduced downtime, and increased sustainability.
Azure’s approach to HPC and AI delivers a unique value proposition with purpose-built infrastructure for top price/performance. This approach enables massive scalability, and today several Top 500 supercomputers are running on Azure.
Because a strong partner ecosystem is crucial when it comes to long-term support for customers’ HPC workloads, constant collaboration between Microsoft and Altair empowers customers to get the most out of their solutions on Azure.
PBS Professional in Azure CycleCloud
With Azure CycleCloud, users are empowered to easily migrate their HPC and AI workloads onto the Azure platform while keeping their preferred scheduling and workflow management systems. This allows users and IT administrators to easily plug Azure HPC cloud benefits into their simulation flows without complexities or major adjustments to existing processes or IT architecture.
Azure CycleCloud enables resource autoscaling depending on cluster demand, configuration management through an infrastructure-as-code approach, and granular configuration tuning in terms of hardware and software stack for the specific job type.
Now that Microsoft and Altair support the PBS Professional scheduler inside Azure CycleCloud, PBS Professional users can keep using the powerful scheduler and migrate their workloads to Azure cloud.
The integration allows automated PBS Professional cluster provisioning on Azure. It features execution-node autoscaling based on job queue size, and advanced integration with scheduler resources definition, allowing node allocation with operating system, hardware, and software configuration based on specific job needs— all with Azure CycleCloud’s cost management and administration functionality.
PBS Professional in Action on Azure
The integration with PBS Professional in Azure CycleCloud allows the smooth setup of a PBS Professional cluster directly through a wizard:
Users can select from a wide variety of Azure Virtual Machine (VM) sizes in the user interface, allowing all resources to be available in the PBS Professional cluster as compute nodes.
Once the cluster is created, users can connect directly to the PBS Professional head node and start submitting jobs. The creation of jobs in queue will spin up the requested resources automatically, and they’ll be managed seamlessly by Azure CycleCloud’s autoscale library.
Submitting a few example jobs will automatically create the dedicated resources in Azure CycleCloud:
qsub -N job-1 -l vm_size=Standard_E48bds_v5 -- /bin/sleep 60
qsub -N job-2 -l vm_size=Standard_D32ads_v5 -- /bin/sleep 60
qsub -N job-3 -l vm_size=Standard_D8ads_v5 -- /bin/sleep 60
qsub -N job-4 -l vm_size=Standard_HB120rs_v3 -- /bin/sleep 60
Using the custom resource vm_size, a user can specify the target Azure VM size that should be allocated for the job.
The autoscaling libraries will monitor scheduler-requested resources and, if they’re compliant with Azure CycleCloud’s configured maximum number of cores and VM types, it will trigger the deployment in Azure, automatically adding the nodes to the Azure PBS Professional cluster.
Additional node and cluster customizations can be made using Azure CycleCloud functionality through cloud-init and cluster-init.
This may include software-stack tuning or PBS Professional customizations required by the environment, like munge authentication or additional queue/resource configuration.
Resources that are idle for more than five minutes will be automatically deprovisioned to ensure effective cost management and optimization.
Altair and Microsoft in Cloud HPC
The PBS Professional and Azure CycleCloud integration is the latest step in a close collaboration between Altair and Microsoft, resulting in the most advanced simulation and workflow management tools combined with the advanced performance of Azure HPC. The integration is currently available in preview from the Azure CycleCloud PBS Professional repository in GitHub, and it will be included in a future version of Azure CycleCloud.
This follows the availability on Azure Marketplace of the Altair® Unlimited™ virtual appliance on Azure cloud, providing instant access in a SaaS approach to a range of Altair® HyperWorks® simulation software. It also completes the availability of Altair workload managers on Azure HPC, alongside the Altair® Grid Engine® integration on Azure CycleCloud.
If you’re at International Super Computing in Hamburg, May 22–24, visit Microsoft Booth #C318 for a demo of the PBS Professional and Azure CycleCloud integration on Tuesday, May 22 from 2 to 6 p.m. CET or stop by Altair Booth #F611 to speak to an Altair expert.