US 12,135,629 B2
Workload placement based on special purpose accelerator requirements and performance metrics
Dharmesh M. Patel, Round Rock, TX (US); and John Kelly, Mallow (IE)
Assigned to Dell Products L.P., Round Rock, TX (US)
Filed by Dell Products L.P., Round Rock, TX (US)
Filed on Jun. 28, 2021, as Appl. No. 17/360,130.
Prior Publication US 2022/0413987 A1, Dec. 29, 2022
Int. Cl. G06F 11/34 (2006.01); G06F 9/50 (2006.01); G06F 11/30 (2006.01)
CPC G06F 11/3433 (2013.01) [G06F 9/5044 (2013.01); G06F 9/5083 (2013.01); G06F 11/3006 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An infrastructure manager for placing workloads for performance across available infrastructure comprising a plurality of on-demand infrastructure and a plurality of dedicated infrastructure, comprising:
a storage device for storing an available infrastructure repository specifying at least types, quantities, and characteristics of different portions of the available infrastructure for placing the workloads; and
a processor programmed to:
intercept a workload placement request for a workload of the workloads, wherein the request specifies at least one selected from a group consisting of a type of the workload, resources required to perform the workload, and a prioritization parameter for performance of the workload, wherein the request is issued by a client;
identify performance metrics by matching the type of the workload to performance profiles associated with the workloads, wherein the performance metrics are used to estimate how efficient the workload will be performed using a combination of the plurality of on-demand infrastructure and the plurality of dedicated infrastructure;
make, based on the request, a determination that the workload has a special purpose hardware requirement;
in response to the determination:
identify, based on the performance metrics and the special purpose hardware requirement, a plurality of potential placement locations (PLs) in the available infrastructure for the workload that each meet the special purpose hardware requirement;
place, based on information obtained from a performance metrics repository, the workload at a first PL of the plurality of the potential PLs,
wherein the information comprises at least a first result specifying a first performance of a test workload using the dedicated infrastructure and a second result specifying a second performance of the test workload using the on-demand infrastructure;
identify, after placing the workload and while the workload is being performed, a performance metric change in the available infrastructure, wherein the workload is impacted by the performance metric change;
identify, based on the performance metric change and the special purpose hardware requirement, a second PL of the plurality of potential PLs is now available to perform the workload in a more efficient manner than the first PL, wherein the second PL is different from the first PL or the second PL is not different from the first PL;
make a second determination that the second PL is different from the first PL; and
migrate, based on the second determination, the workload to the second PL.