The azure-batch tag should be used to ask questions regarding all aspects of the Azure Batch service including job scheduling, compute node provisioning, and pool orchestration. Extended ecosystem questions are also welcome that are related to Azure Batch accelerator offerings, Azure Portal Batch blades, and Batch Explorer.
Azure Batch is a job scheduling and compute management orchestration solution for the Microsoft Azure cloud. The service is available in all Azure regions and can deploy with nearly all Azure VM sizes available, including those with specialized hardware such as GPUs and Infiniband/RDMA interfaces.
Azure Batch is a managed job scheduling solution where the underlying compute resources are allocated as pools. Pools may be comprised of Linux or Windows VMs and any number of dedicated and low priority nodes and can be automatically scaled through autoscale formulas. An Azure Batch Job can be thought of as a queue for tasks. Each Azure Batch Job can be assigned to a pool. Any number of tasks can be assigned to jobs, including multi-instance tasks (i.e., multi-node) which support MPI jobs. Azure Batch also supports containerized execution.
For a full introductory treatment of Azure Batch, see the Azure Batch Technical Overview. The service team also provides a GitHub repository for issue tracking, feature requests, and change logs.
