Tailor-Made HPC in the Cloud
Clouds and Distributed Computing
TimeTuesday, 19 November 201911:30am - 12pm
DescriptionHPC in the cloud should simplify and optimize application execution and management, but care needs to be taken to ensure that on-premise issues don’t become cloud issues. Workload I/O contention, scheduling, and efficiency issues can all be reduced with Google Cloud. The key to avoiding lift and shifting issues is to ensure that the cloud infrastructure is tailor-made to your applications. Our goal at Google is to allow users to focus on application needs, and not on the toil of deployment, maintenance, and optimization.
This session focuses on how Google Cloud users can renew their focus on application requirements. In one example, file system deployments can be tailored to the IOPs and bandwidth requirements for individual workloads, which is simply not possible on-premise. By taking advantage of resources unique to the cloud, many of the existing issues simply no longer exist, and users can now focus on their HPC applications.