Task-level optimization with compute environments

• US 10,402,227 B1
• Filed: 08/31/2016
• Issued: 09/03/2019
• Est. Priority Date: 08/31/2016
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Kinney, Jr., James Edward
• Primary Examiner(s): Lee, Adam
• Application Number:
  US15/253,699
• Time in Patent Office: 1,098 Days
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/5005 to service a request
• CPC Class Codes: G06F 9/5011 the resources being hardware resources other than CPUs, Servers and Terminals
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06F 9/5066 Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs mappping at compile time, see G06F8/451
• CPC Class Codes: G06F 9/5077 Logical partitioning of resources; Management or configuration of virtualized resources specific details on emulation or internal functioning of virtual machines G06F9/455
• CPC Class Codes: H04L 47/70 Admission control; Resource allocation

Abstract

Methods, systems, and computer-readable media for task-level optimization of compute environments are disclosed. Execution is initiated of one or more tasks using a plurality of computing resources provisioned from a multi-tenant provider network. At least some of the computing resources vary in configuration. One or more metrics are determined that are associated with the execution of the one or more tasks. A configuration of the computing resources is selected based at least in part on the one or more metrics. A modified job definition associated with the one or more tasks is generated. The modified job definition indicates the selected configuration.

Claims

1. A system, comprising:

a pool of available computing resources of a multi-tenant provider network; and

one or more computing devices having at least one processor and a memory configured to implement a compute environment management system, wherein the compute environment management system is configured to;

receive a job definition defining an array job comprising a first set of one or more tasks and a second set of one or more tasks;

select, a plurality of available computing resources in the pool for execution of the first set of one or more tasks, wherein the plurality of available computing resources comprise a plurality of different hardware configurations;

initiate execution of the first set of one or more tasks using the selected plurality of available computing resources from the pool;

monitor the execution of the first set of one or more tasks by the plurality of different hardware configurations;

generate one or more performance metrics based at least in part on the monitoring of the execution of the first set of one or more tasks;

select one or more recommended hardware configurations from the plurality of different hardware configurations based at least in part on the generated one or more performance metrics;

modify the job definition to indicate the selected one or more recommended hardware configurations;

provision a compute environment based at least in part on the modified job definition, wherein the compute environment comprises a plurality of additional available computing resources from the pool having the indicated one or more recommended hardware configurations; and

initiate execution of the second set of one or more tasks across the compute environment according to the modified job definition such that the second set of one or more of the tasks is executed by the plurality of additional available computing resources having the indicated one or more recommended hardware configurations.

2. The system as recited in claim 1, wherein the compute environment management system is further configured to:

generate a recommendation indicating the one or more recommended hardware configurations; and

solicit user input approving or declining the recommendation.

3. The system as recited in claim 1, wherein the compute environment management system is further configured to:

determine at least some of the hardware configurations used for the execution of the first set of one or more tasks based at least in part on the job definition.

4. The system as recited in claim 1, wherein the compute environment management system is further configured to:

select the one or more recommended hardware configurations to optimize the system for cost or performance.

5. A computer-implemented method executed by a compute system having at least one processor and a memory, the method comprising:

receiving a job definition defining an array job comprising a first set of one or more tasks and a second set of one or more tasks;

initiating execution of the first set of one or more tasks on a plurality of different types of available computing resources provisioned from a pool of available computing resources of a multi-tenant provider network, wherein at least some of the different types of available computing resources initiated to execute the first set of one or more tasks vary in configuration;

determining one or more metrics that are based on the execution of the first set of one or more tasks by the plurality of different types of available computing resources;

selecting a configuration of the varied configurations of the plurality of different types of the available computing resources based at least in part on the determined one or more metrics;

generating a modified job definition associated with the second set of one or more tasks, wherein the modified job definition indicates the selected configuration;

provisioning, based at least in part on the modified job definition, a compute environment that comprises one or more additional available computing resources from the pool having the indicated configuration; and

initiating execution of the second set of one or more tasks across the compute environment according to the modified job definition such that the second set of one or more of the tasks is executed by the one or more additional available computing resources having the indicated configuration.

6. The method as recited in claim 5, further comprising:

generating a hardware configuration recommendation based at least in part on the determined one or more metrics, wherein the selected configuration is determined based at least in part on user input indicating acceptance of the hardware configuration recommendation.

7. The method as recited in claim 5, further comprising:

determining at least some of the plurality of different types of available computing resources used for the execution of the first set of one or more tasks based at least in part on the job definition.

8. The method as recited in claim 5, wherein:

the determined one or more metrics comprise one or more cost metrics associated with a cost of executing the plurality of different types of available computing resources, and

said selecting the configuration of the varied configurations of the plurality of different types of available computing resources comprises selecting the configuration to optimize for the cost of executing the plurality of different types of available computing resources.

9. The method as recited in claim 5, wherein:

the determined one or more metrics comprise one or more metrics associated with performance of the plurality of different types of available computing resources that execute the first set of one or more tasks, and

said selecting the configuration of the varied configurations of the plurality of different types of available computing resources comprises selecting the configuration of the plurality of different types of available computing resources to optimize for performance of the plurality of different types of available computing resources.

10. The method as recited in claim 5, further comprising:

varying configuration parameter values for at least some of a plurality of instances of an application, wherein the execution of the first set of one or more tasks using the plurality of different types of available computing resources comprises execution of the plurality of instances of the application using the plurality of different types of available computing resources.

11. The method as recited in claim 5, wherein the selected configuration indicated in the modified job definition comprises one or more hardware characteristics of the one or more additional available computing resources.

12. The method as recited in claim 5, wherein the selected configuration indicated in the modified job definition comprises one or more software parameter values for application software that is executed to execute the second set of one or more tasks.

13. A non-transitory computer-readable storage medium storing computer-executable program instructions that when executed by one or more processors perform:

receiving a job definition defining an array job comprising a first set of one or more tasks and a second set of one or more tasks;

initiating execution of the first set of one or more tasks using a plurality of different types of available computing resources provisioned from a pool of available computing resources of a multi-tenant provider network, wherein at least some of the different types of available computing resources initiated to execute the first set of one or more tasks vary in computational or memory characteristics;

determining one or more metrics based on the execution of the first set of one or more tasks by the plurality of different types of available computing resources;

selecting a configuration of the varied computational or memory characteristics of the plurality of different types of available computing resources based at least in part on the determined one or more metrics;

generating a modified job definition associated with the second set of one or more tasks, wherein the modified job definition indicates the selected configuration;

provisioning, based at least in part on the modified job definition, a compute environment that comprises one or more additional available computing resources from the pool having the selected configuration indicated by the modified job definition; and

initiating execution of the second set of one or more tasks across the compute environment according to the modified job definition such that the second set of one or more tasks is executed by the one or more additional available computing resources having the indicated configuration.

14. The non-transitory computer-readable storage medium as recited in claim 13, wherein the determined one or more metrics comprise one or more metrics associated with a cost or performance of the plurality of different types of available computing resources that execute the first set of one or more tasks, and wherein the selected configuration is selected to optimize for the cost or performance.

15. The non-transitory computer-readable storage medium as recited in claim 13,

wherein the multi-tenant provider network comprises a plurality of hardware configurations varying in computational or memory characteristics, and

wherein the selected configuration comprises a selection of one or more of the hardware configurations that vary in the computational or memory characteristics.