Equitable sharing of system resources in workflow execution

Patent metadata

Field	Value
Publication number	US-10554575-B2
Application number	US-201715787827-A
Country	US
Kind code	B2
Filing date	Oct 19, 2017
Priority date	Aug 21, 2014
Publication date	Feb 4, 2020
Grant date	Feb 4, 2020

Abstract

A method may be practiced in a distributed computing environment that provides computing resources to a plurality tenants. The method includes acts for allocating a limited set of system resources to the tenants. The method includes identifying a resource slice. The method further includes identifying an executing tenant workload. Checkpoint characteristics are identified for the executing tenant workload. Based on the checkpoint characteristics and the resource slice, a task eviction event is identified.

First claim

What is claimed is:

1. A computer system comprising:
one or more processor(s); and
one or more computer-readable hardware storage device(s) having stored thereon computer-executable instructions that are executable by the one or more processor(s) to cause the computer system to run a cloud service configured to incentivize use of certain workload checkpointing behavior by providing constructive treatment to workloads demonstrating said workload checkpointing behavior by causing the cloud service to at least:
assign a workload to a resource of the cloud service;
identify an eviction for the workload, the eviction being an event in which an originally assigned use of the resource for the workload expires;
cause the resource to perform work for the workload;
determine that a checkpoint of the workload is scheduled to be performed within a determined proximity to the identified eviction, wherein, as a result of the checkpoint being within the determined proximity to the identified eviction, the workload is determined to demonstrate the workload checkpointing behavior; and
based on determining the workload demonstrates the workload checkpointing behavior, provide the constructive treatment to the workload by permitting an extended use of the resource beyond the originally assigned use.

2. The computer system of claim 1 , wherein the originally assigned use includes a timed amount of processor usage.

3. The computer system of claim 1 , wherein the checkpoint is an event in which information associated with execution of the workload is persisted to storage.

4. The computer system of claim 1 , wherein the eviction occurs when the resource has executed the workload for a determined amount of processor time.

5. The computer system of claim 1 , wherein the eviction occurs when the resource has executed the workload for a determined number of script commands.

6. The computer system of claim 1 , wherein the eviction occurs when the resource has executed the workload for a determined number of processors operations.

7. The computer system of claim 1 , wherein causing the resource to perform the work for the workload includes loading the workload into memory of the resource.

8. The computer system of claim 1 , wherein the cloud service is further configured to notify a tenant that a particular workload will more likely succeed if the particular workload is checkpointed.

9. The computer system of claim 8 , wherein the workload is submitted to the cloud service by a tenant, and wherein a notification is sent to the tenant informing the tenant that the workload will more likely be successfully executed if the workload includes more frequent checkpoints.

10. The computer system of claim 1 , wherein the cloud service is further configured to determine where, within the workload, any checkpoints exist.

11. The computer system of claim 1 , wherein the cloud service is further configured to determine a consistency of checkpoints within the workload.

12. The computer system of claim 1 , wherein, upon a condition in which the resource is not assigned any other workloads, the resource continues to execute the workload beyond the originally assigned use.

13. The computer system of claim 1 , wherein the cloud service receives a request comprising the workload, and wherein the request is received at a queue of the resource.

14. A method for incentivizing use of certain workload checkpointing behavior by providing constructive treatment to workloads demonstrating said workload checkpointing behavior, the method being implemented by a cloud service executing on a computer system, the method comprising:
assigning a workload to a resource of the cloud service;
identifying an eviction for the workload, the eviction being an event in which an originally assigned use of the resource for the workload expires;
causing the resource to perform work for the workload;
determining that a checkpoint of the workload is scheduled to be performed within a determined proximity to the identified eviction, wherein, as a result of the checkpoint being within the determined proximity to the identified eviction, the workload is determined to demonstrate the workload checkpointing behavior; and
based on determining the workload demonstrates the workload checkpointing behavior, providing the constructive treatment to the workload by permitting an extended use of the resource beyond the originally assigned use.

15. The method of claim 14 , wherein the eviction occurs when the resource has executed the workload up until a determined amount of non-volatile storage resources have been used.

16. The method of claim 14 , wherein the eviction occurs when the resource has executed the workload up until a determined amount of volatile memory resources have been used.

17. The method of claim 14 , wherein the resource is one of multiple resources that are provided to a plurality of tenants.

18. The method of claim 14 , wherein the eviction for the workload is at least partially based on how many checkpoints are scheduled for the workload.

19. The method of claim 14 , wherein the workload includes multiple checkpoints such that information associated with execution of the workload is persisted to storage multiple times.

20. One or more hardware storage device(s) having stored thereon computer-executable instructions that are executable by one or more processors of a computer system to thereby cause the computer system to run a cloud service configured to incentivize use of certain workload checkpointing behavior by providing constructive treatment to workloads demonstrating said workload checkpointing behavior by causing the cloud service to at least:
assign a workload to a resource of the cloud service;
identify an eviction for the workload, the eviction being an event in which an originally assigned use of the resource for the workload expires;
cause the resource to perform work for the workload;
determine that a checkpoint of the workload is scheduled to be performed within a determined proximity to the identified eviction, wherein, as a result of the checkpoint being within the determined proximity to the identified eviction, the workload is determined to demonstrate the workload checkpointing behavior;
based on determining the workload demonstrates the workload checkpointing behavior, provide the constructive treatment to the workload by permitting an extended use of the resource beyond the originally assigned use.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Zlati Nicoleta
• Cho Felix Man Toe
• Kallstrom Erik Jonas
• Garg Anirudh
• Sedukhin Igor
• Nayar Manoj
• Enein Mohamed

Classifications

• G06F9/46

  Multiprogramming arrangements · CPC title

• G06F9/4418

  Suspend and resume; Hibernate and awake · CPC title

• G06F9/485 Primary

  Task life-cycle, e.g. stopping, restarting, resuming execution (G06F9/4881 takes precedence) · CPC title

• G06F9/5083

  Techniques for rebalancing the load in a distributed system · CPC title

• H04L47/783 Primary

  Distributed allocation of resources, e.g. bandwidth brokers · CPC title