US 11,995,476 B1
Client-configurable retention periods for machine learning service-managed resources
Ramyanshu Datta, Campbell, CA (US); Ishaaq Chandy, Bellevue, CA (US); Arvind Sowmyan, Seattle, WA (US); Wei You, Kirkland, WA (US); Kunal Mehrotra, Kirkland, WA (US); Kohen Berith Chia, Seattle, WI (US); Andrea Olgiati, Gilroy, CA (US); Lakshmi Naarayanan Ramakrishnan, Redmond, WA (US); and Saurabh Gupta, Sammamish, WA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Sep. 22, 2021, as Appl. No. 17/482,276.
Int. Cl. G06F 9/46 (2006.01); G06F 9/50 (2006.01)
CPC G06F 9/5038 (2013.01) [G06F 9/5022 (2013.01); G06F 9/5055 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system, comprising:
one or more computing devices;
wherein the one or more computing devices include instructions that upon execution on or across the one or more computing devices:
cause, by a machine learning service of a provider network, a first machine learning task to be performed at a first compute instance of a virtualized computing service, wherein the first compute instance is utilized for the first machine learning task by the machine learning service in response to a first task request from a requester of a first set of requesters, wherein the first task request does not specify that the first compute instance is to be used for the first machine learning task, and wherein metadata stored by the machine learning service indicates that (a) execution of the first compute instance is not to be terminated during a post-task-completion retention period after a task performed at the first compute instance is completed, and (b) the first compute instance is not to be used for a task which has been requested by a requester that is not a member of the first set of requesters;
cause, by the machine learning service, a second machine learning task to be performed at the first compute instance after the first machine learning task is completed, wherein an indication of the second machine learning task is received from a requester of the first set of requesters before an expiration of the post-task-completion retention period relative to completion of the first machine learning task, and wherein the request for the second machine learning task does not specify that the first compute instance is to be used for the second machine learning task; and
in response to (a) obtaining a task request for a third machine learning task at the machine learning service from a requester of the first set of requesters and (b) determining that a threshold criterion associated with the post-task-completion retention period and the set of requesters has been satisfied:
cause, by the machine learning service, the third machine learning task to be performed at a second compute instance of the virtualized computing service, wherein after the third machine learning task is completed, the second compute instance is terminated by the machine learning service without waiting for a post-task-completion retention period to expire.