Using Placement Sets for Altair PBS Professional

Fairshare is a powerful tool for sharing resources and allocating a set amount of your cluster to entities like users, groups, and projects. This white paper will get you up and running with basic fairshare on Altair PBS Professional.

Fairshare is a PBS Professional scheduling tool designed to share a cluster’s limited resources according to cluster usage history and allocated percentage. It's the most direct option to grant a percentage of shared resources on the cluster to users, projects, or groups. This white paper covers the creation and management of the fairshare tree structure for your PBS complex. It also covers basic settings and describes concepts of shares within the fairshare tree. Understanding, calculating, and managing entity usage data will not be covered herein. It assumes your familiarity with basic fairshare setup along with the limitations and caveats fairshare has with other PBS Professional scheduling tools. Enabling basic fairshare does not consider fairshare tree capabilities and structure or describe how to create and manage the tree. Granting a percentage of the cluster to individual users may not be viable; the use of your cluster may instead be defined with respect to queues, projects, or accounting. Further, some users may span across multiple groups. Your site’s policy may require the restriction of cluster use to only known entities. PBS Professional’s weighted rooted tree structure provides the opportunity to group entities together and finely manage the balance of resources on your cluster. Adapting resource usage allows you to dictate tracked resources and apply a formula to balance those resources according to your site’s needs.

Fairshare is an Altair® PBS Professional® scheduling tool designed to share a cluster’s limited resources according to history of cluster usage and the entities’ allocated percentage. Fairshare is the most direct option to grant a percentage of shared resources on the cluster to users, projects, or groups. This white paper covers how PBS Professional collects and manages fairshare usage. With this document, you should be able to walk through each logical step of fairshare, understanding each number from pbsfs output. It also covers how you may manually alter each entity’s usage data.

PBS Professional provides the user with an interface to submit jobs to a queue, which will be considered by the scheduler and dispatched to computational resources in a cluster, supercomputer, or cloud. This default configuration has no restrictions on the user, and it allows for maximum utilization of computational resources, which PBS calls vnodes. As an administrator, you may be asked to allocate or restrict computational resources to a specific user, group, or project. Another common ask is from a user who needs dedicated computational resources for a special project. PBS Professional provides many ways to allocate and restrict resources. In this document we will describe two common methods of satisfying the requirement by using queues and custom resources.

Altair PBS Professional is a fast, powerful workload manager designed to improve productivity, optimize utilization and efficiency, and simplify administration for clusters, clouds, and supercomputers — from the biggest HPC workloads to millions of small, high-throughput jobs. PBS Professional can be integrated with popular container technologies like Docker, Singularity, and Enroot. In this document, we will highlight PBS Professional integration with Enroot, a simple yet powerful tool to turn traditional container/OS images into unprivileged sandboxes.

High-performance computing (HPC) fuels scientific discovery and innovation across multiple industries. The combination of large datasets, advanced simulation techniques, and machine learning helps organizations generate insights that would not be possible without modern HPC infrastructure. Given HPC’s outsized role in driving business results, selecting the right management software is critical. This is especially true in commercial organizations where time is money. In this paper, we discuss the pros and cons of open-source software in HPC and make the case for commercial workload management. While open-source workload managers are fine in some situations, they can present disadvantages in production environments.