The compute resources provided by Scientific Computing are available to all Hutch researchers. As a shared resource in high demand, it’s necessary to manage usage to ensure some measure of equitable access to all research.
This high demand for a shared resource is primarily seen in the Gizmo cluster, where we have both a) limits on how much can be used by individuals and groups and b) algorithms that prioritize pending work. The limits and prioritization algorithms serve to schedule the waiting jobs for available resources.
The cluster account is the basis for these algorithms - the calculation of a job’s priority for scheduling purposes and the entity on which limits are placed.
Thus, the cluster account only provides access to the cluster. After creating the account, individuals are associated with the account, allowing them to submit jobs into that account.
Cluster accounts are created for Hutch faculty and NIH grant principals. Associations to accounts are created manually, either on request from a faculty or designate or by user request when the individual is indicated as a report to faculty in Workday.
Account Versus Allocation
The terms “account” and “allocation” are often conflated in conversations about usage limits. However, it is important to note that we do not “allocate” resources to individuals or accounts in the sense of reserving or designating resources for an account’s use.
These systems are managed as a common pool with systems designed to provide a minimum of resources “instantaneously” to an account with larger volumes of computational resources available to that account over time.
Updated: February 12, 2021Edit this Page via GitHub Comment by Filing an Issue Have Questions? Ask them here.