Migrating from LSF to Accelerator

Shannon Cody_21144
Shannon Cody_21144
Altair Employee
edited February 2023 in Altair HPCWorks

Despite the fact that LSF can be deeply entrenched in design flows and scripts, many customers have used a structured approach to successfully migrate from it to Accelerator, thereby avoiding disruption in operations. In the attached document we give new Accelerator administrators and users a feel for what we offer, and we give guidance in migrating from LSF to Accelerator.

  

This document introduces key terminology, as well as conceptual differences between Accelerator and LSF such as job queues versus job classes, priority versus fairshare, reservations and limits versus preemption, and partitions versus tasker binding. It also covers the key workload management features in Altair Accelerator.

 

The migration process is outlined in five steps:

 

  1. Assessing the Migration Task
  2. Preparing for Migration
  3. Rolling out the Migration
  4. Decommissioning the Old Workload Manager
  5. Optimizing Accelerator

  

In addition to providing a clear path for the transition from LSF to Accelerator, this document provides key information on further optimizing the impact of your Accelerator implementation in the "Optimizing Accelerator" step, through features like storage aware scheduling via Altair's collaboration with Dell Technologies, SAGA, (storage-aware grid acceleration). 

Tagged:

Categories