Azure Batch allows its user to schedule the tasks as Docker Container as containers are known as the best way to schedule, package, and execute High-Performance Computing Applications (HPC). Containers and Azure Batch also let you use powerful cloud-native job batch the workload in a consistent manner.
If you are a Singularity container user, then you can use Batch Shipyard release for better support. The basic purpose of Singularity container is to provide help to both administrators and users in the use of HPC applications in cluster computing environments. It also gives access to users on accelerators like specialized interconnects and GPUs.
Batch Shipyard is a highly useful system that helps the users to share and execute their Singularity workloads on Azure platform. It’s a simple and configuration based system that makes execution effortless. If you talk about the Azure’s GPU, then NCv2, RDMA, and ND are ideal instances which have immense compatibility with these types of workloads. It automatically scales your tasks and executes them on low priority VMs to save 80% storage.
The features of Batch Shipyard are not limited to Singularity only, but it also supports Windows containers YAML based configuration support and integrated Azure Batch Container Support. It has the ability to provide support ARM image-based custom images and gives references to multiple private registries. The other features of Azure Batch are pre-built binaries for the CLI and migration from one system to another. You can directly download Batch Shipyard from Azure Cloud Shell which requires no installation.