Vasantha K Tammana

US Patent:

20200012523, Jan 9, 2020

Filed:

Sep 19, 2019

Appl. No.:

16/576637

Inventors:

- San Jose CA, US
Amit Mittal - San Francisco CA, US
Anjali Gugle - Cupertino CA, US
Hetal N. Badheka - Sunnyvale CA, US
Vasantha K. Tammana - San Jose CA, US

International Classification:

G06F 9/48
G06F 9/50
G06F 11/34

Abstract:

Disclosed is a method and system for using a credit-based approach to scheduling workload in a compute environment. The method includes determining server capacity and load of a compute environment and running a first benchmark job to calibrate a resource scheduler. The method includes partitioning, based on the calibration, the compute environment into multiple priority portions (e.g. first portion, second portion etc.) and optionally a reserve portion. Credits are assigned to allocate system capacity or resources per time quanta. The method includes running a benchmark job to calibrate a complexity of supported job types to be run in the compute environment. When a request for capacity is received, the workload is assigned one or more credits and credits are withdrawn from the submitting entity's account for access to the compute environment at a scheduled time.

US Patent:

20190205164, Jul 4, 2019

Filed:

Jan 3, 2018

Appl. No.:

15/861255

Inventors:

- San Jose CA, US
Amit Mittal - San Francisco CA, US
Anjali Gugle - Cupertino CA, US
Hetal N. Badheka - Sunnyvale CA, US
Vasantha K. Tammana - San Jose CA, US

International Classification:

G06F 9/48
G06F 9/50
G06F 11/34

Abstract:

Disclosed is a method and system for using a credit-based approach to scheduling workload in a compute environment. The method includes determining server capacity and load of a compute environment and running a first benchmark job to calibrate a resource scheduler. The method includes partitioning, based on the calibration, the compute environment into multiple priority portions (e.g. first portion, second portion etc.) and optionally a reserve portion. Credits are assigned to allocate system capacity or resources per time quanta. The method includes running a benchmark job to calibrate a complexity of supported job types to be run in the compute environment. When a request for capacity is received, the workload is assigned one or more credits and credits are withdrawn from the submitting entity's account for access to the compute environment at a scheduled time.