New Features in Altair Grid Engine 8.7.0

Shannon Cody_21144
Shannon Cody_21144
Altair Employee
edited December 2021 in Altair HPCWorks

Altair® Grid Engine® version 8.7.0 offers performance enhancements including throughput and scalability updates and flexible resource requests for hybrid, parallel jobs. These performance improvements help Altair Grid Engine users compete in even the most competitive HPC verticals such as FinTech. Altair Grid Engine 8.7.0 is available in the Altair One Marketplace.

New in Altair Grid Engine 8.7.0:

  • Change active advance reservations on the fly​; modify reservations as needed for CPUs, memory, and other resources over a particular time period
  • Linux cgroups memory metrics for highly accurate job monitoring and accounting
  • Bulk modification support for all objects, dramatically improving configuration performance for large clusters
  • GPU support improvements​
  • Altair® Access™ integration

Throughput and Scalability

Advancements in Grid Engine 8.7.0 provide customers with improved job throughput in highly loaded clusters, faster response to client requests, and the ability to scale to even bigger Altair Grid Engine clusters, so users can carve out an even bigger competitive edge.

GPU Acceleration 

Altair Grid Engine 8.7.0 supports GPU acceleration, including NVIDIA Multi-Instance GPU (MIG) for the A100 Tensor Core GPUAltair Grid Engine manages A100 GPUs running in MIG mode, and it also offers support for Docker container workloads — great for artificial intelligence (AI) and machine learning (ML). 

Support for Altair Access: Seamless Job Submission, Visualization, and Management for End Users 

Integration with Access gives Altair Grid Engine users, including researchers and engineers, a seamless, intuitive HPC job submission portal they can use to interact with computational environments on-premises, in the cloud, or a hybrid combination of both via web, desktop, and mobile.  

image