AWS PCS: HPC fully managed from the Amazon cloud
With its Parallel Computing Service, Amazon is now offering HPC clusters from the cloud that are designed to take as much work off administrators as possible.
(Image: iX)
Amazon calls its new, fully provider-managed high-performance computing service the AWS Parallel Computing Service (PCS). It allows users to set up, manage and scale HPC clusters. Simple Linux Utility for Resource Management (Slurm), with which many administrators are already familiar, is used for job scheduling.
PCS is not the first HPC service from the AWS cloud: ParallelCluster made its debut in 2018, but administrators still had to take on many management tasks themselves. PCS is now set to simplify management: For example, Amazon takes over update management completely and provides observability functions ex works.
Interested in securing Amazon Web Services? In this training course, you will learn how attackers exploit misconfigurations of the Amazon Cloud and missing hardening measures, and how to effectively secure your AWS services and cloud identities. Dates and further information can be found at heise.de/s/3x00Y.
AWS PCS can be accessed via the AWS Management Console, the AWS SDK or the AWS CLI. Administrators first configure the cluster and then transfer the desired resource management to the service via Slurm. Users can then access the system directly – via SSH, for example. Alternatively, Amazon provides a remote desktop connection.
AWS PCS is available immediately, including in the Europe region (Frankfurt). Prices are based on the size of the Slurm controller and a fee for the node management of the instances. A sample calculation can be found on the product page. Administrators can find an initial guide to the new service on the AWS blog.
(fo)
