Cloud computing
Remote computing with k-Plan
k-Plan executes plans in the cloud using remote high-performance computing (HPC) resources using a HPC-as-a-service paradigm. When a plan is submitted to k-Dispatch, the computing resources needed to run the plan are calculated based on the size of the simulation domain, the number of sonications, and so on. The plans are then submitted for execution on one or more remote HPC facilities. These facilities typically employ a best-effort strategy, which maximises resource utilisation under varying conditions without providing strict performance guarantees.
Managing workloads
The remote computing facilities currently used by k-Plan use fair-share policies. This means that all users compete for available computing resources. Each time a plan is submitted, the fair-share coefficients are recalculated, which can impact the time it takes to complete your plan.
Be mindful when submitting a very large number of plans simultaneously. Doing so can lead to high workload volumes, potentially causing temporary system overloads and longer completion times. By spreading out submissions, you can help maintain a balanced workload and improve overall system efficiency.