Alces Flight, a new entry to the AWS Marketplace, provides an “effortless” yet fully featured High Performance Computing environment (HPC) that can be launched in minutes according to Jeff Barr.
Barr wrote last week that the application can be used on-demand or spot instances and comes with a job scheduler, hundreds of HPC applications and is ready to go. Some applications have built-in collaboration features, such as shared graphic views.
Spot instances of EC2 provide spare AWS capacity at a discount of up to 90 percent from AWS’ On Demand pricing. This can significantly reduce an organization’s core cost. Barr stated that the new HPC offering will help corporate researchers and academics who are increasingly turning to the cloud for high-quality, parallel computing jobs at a reasonable price.
They can start clusters immediately on demand, run the jobs, and then shut them down immediately. This allows them to scale up or launch new clusters as the demand grows.
Barr stated that this self-serve, cloud-based approach favors science and servers and accelerates research and innovation. “Colleagues can have access to shared, cloud-based resources on the same campus as well as halfway around the globe, without worrying about possible issues at the organizational or network boundaries.”
AWS HPC clusters behave exactly like traditional Linux-powered HPC clusters. They can launch with fixed sizes, or with auto scaling. They run in a Virtual Private Cloud with SSH and graphical Desktop connectivity.
Barr stated that you can launch a small cluster of nodes (up to 8) for testing and evaluation, or a larger cluster to conduct research. The AWS CloudFormation template can be downloaded from the Alces website if you subscribe to the product. This template powers all products and can be used to quickly launch all AWS resources required to create the cluster.
Barr explains how to launch an HPC cluster in his last week’s post. The company also provides guidance on a getting started guide.
