Scalable Rack Units (SRUs) are flexible building blocks of compute that can be scaled up or down to meet the needs of the customer. These can be used to simplify the cluster design process. The focus of SRUs is on the compute, which means the SRU have to be combined with corresponding infrastructure servers, storage, and core networking. Each SRU contains all the network infrastructure for that rack with enough ports to allow for a fully non-blocking configuration to the rest of the cluster. The number and configuration of core switches required to connect the SRUs to the cluster will vary based on size and chosen blocking factor.

The SRUs come in two flavors: Scalable Compute Rack Unit (SCRU) and Scalable Accelerated Compute Rack Unit (SACRU). These SRUs can be mixed and matched as necessary up to the limits of the Cornelis Omni-Path Express networking solution (27k nodes).