Large-scale parallel computation, which Azure serves through CycleCloud or Azure Batch and VM sizes capable of RDMA.
Also called high-performance computing.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains HPC in context, with comparison tables and the common traps.
Terms in this definition
- Azure Batch
Azure service that schedules parallel, HPC and rendering jobs onto pools of compute nodes it scales automatically; multi-instance tasks support MPI workloads.
- RDMA
Short for remote direct memory access: the lowest-latency networking in Azure, running over InfiniBand on HPC VM sizes whose names include an "r".
Related terms
- Accelerated Networking
SR-IOV capability that cuts latency, jitter and CPU overhead for VM networking in Azure. HPC VM sizes with RDMA achieve even lower latency.
- Azure CycleCloud
Tool for creating and operating HPC clusters on Azure around third-party schedulers like Slurm, LSF or PBS Pro.
- LSF
IBM Spectrum Load Sharing Facility, a job scheduler for HPC workloads that Azure CycleCloud supports.
- MPI
Message Passing Interface, a standard by which nodes in an HPC job exchange data. Azure Batch runs such workloads using multi-instance tasks.
- PBS
Portable Batch System. Azure CycleCloud can run this HPC job scheduler, in its PBS Pro form.
- Slurm
A scheduler for HPC jobs, one of those that Azure CycleCloud can run.