Dynamically controlled distributed workload execution

Patent metadata

Field	Value
Publication number	US-10127080-B2
Application number	US-201715413854-A
Country	US
Kind code	B2
Filing date	Jan 24, 2017
Priority date	Feb 3, 2015
Publication date	Nov 13, 2018
Grant date	Nov 13, 2018

Abstract

Methods and systems for optimizing distributed workloads on information handling systems involve determining workload attributes for a computing task for distributed execution among distributed information handling systems. The workload attributes are used to dynamically determine optimal allocation and configuration of distributed hardware resources at the distributed information handling systems. After dynamic confirmation of the allocation and configuration of the distributed hardware resources, the computing task is segmented for distributed execution.

First claim

What is claimed is:

1. A method for optimizing distributed workloads on information handling systems, the method comprising:
receiving a workload request from a first information handling system specifying a computing task for distributed execution;
based on the workload request, determining workload attributes of the computing task, the workload attributes including dependency of the computing task on hardware resources included in information handling systems;
identifying distributed information handling systems accessible via a network for the distributed execution of the computing task, including identifying distributed hardware resources present at the distributed information handling systems;
based on the distributed hardware resources identified, determining a preliminary allocation of the computing task among the distributed hardware resources;
sending a first request to the distributed information handling systems requesting confirmation of the preliminary allocation;
responsive to the first request, receiving a preliminary allocation response from at least some of the distributed information handling systems; and
based on the preliminary allocation response, determining a confirmed allocation of the computing task among the distributed hardware resources, wherein the computing task is executable using the confirmed allocation of the distributed hardware resources.

2. The method of claim 1 , wherein the preliminary allocation response specifies first distributed resource information describing distributed hardware resources currently available at the distributed information handling systems responding to the first request, and further comprising:
based on the confirmed allocation:
sending, to confirmed distributed information handling systems including at least some of the distributed information handling systems, a second request to make the distributed hardware resources described in the first distributed resource information exclusively available for the computing task; and
segmenting the computing task among the confirmed distributed information handling systems for the distributed execution.

3. The method of claim 2 , further comprising:
sending the computing task segmented to respective ones of the confirmed distributed information handling systems;
receiving segmented output of the computing task from respective ones of the confirmed distributed information handling systems, the segmented output corresponding to the computing task segmented;
assembling the segmented output to generate assembled output of the computing task; and
sending the assembled output to the first information handling system.

4. The method of claim 1 , wherein the workload request further specifies an application executing on the first information handling system associated with the computing task, and wherein the workload attributes are selected from at least one of:
an amount of multithreading used by the computing task;
a thread priority used by the computing task;
an instruction set architecture used by the computing task;
a memory capacity used by the computing task;
a memory data throughput used by the computing task;
a memory cache capacity used by the computing task;
a memory cache architecture used by the computing task;
an amount of non-uniform memory access permitted by the computing task;
a network latency permitted by the computing task;
a storage capacity used by the computing task;
a storage data throughput used by the computing task;
a storage latency permitted by the computing task;
an amount of acceleration by a graphics processing unit used by the computing task;
a vertical synchronization setting permitted by the computing task;
an amount of digital signal processing used by the computing task;
an amount of integer processing used by the computing task;
an amount of background execution permitted by the computing task; and
an operating system used by the computing task.

5. The method of claim 1 , wherein the identifying the distributed hardware resources is based on second distributed resource information collected prior to receiving the workload request.

6. The method of claim 1 , wherein the first request includes a request to reserve the distributed hardware resources described in first distributed resource information for the computing task, and wherein the distributed hardware resources are selected from at least one of:
a number of processor cores available for distributed execution;
an amount of exclusive access to processor cores available for distributed execution;
an amount of processor multithreading available for distributed execution;
an amount of background execution available for distributed execution;
an amount of user mode execution available for distributed execution;
a processor frequency;
a processor architecture;
a processor cache architecture;
a processor power state;
a processor cache capacity available for distributed execution;
a processor power consumption available for distributed execution;
an instruction set architecture available for distributed execution;
a graphics processing unit available for distributed execution;
a memory capacity available for distributed execution;
a memory throughput available for distributed execution;
an amount of non-uniform memory access available for distributed execution;
a storage capacity available for distributed execution;
a storage throughput available for distributed execution;
a storage latency available for distributed execution;
a network capacity available for distributed execution;
a network throughput available for distributed execution;
a network latency available for distributed execution;
a system reboot associated with distributed execution; and
a system power down feature associated with distributed execution.

7. The method of claim 1 , further comprising:
receiving an indication specifying a change in first distributed resource information; and
based on the change, updating the confirmed allocation, including generating second distributed resource information indicative of the change.

8. At least one non-transitory computer readable medium, comprising computer readable instructions for optimizing distributed workloads on information handling systems, the instructions, when executed, cause a processor to:
receive a workload request from a first information handling system specifying a computing task for distributed execution;
based on the workload request, determine workload attributes of the computing task, the workload attributes including dependency of the computing task on hardware resources included in information handling systems;
identify distributed information handling systems accessible via a network for the distributed execution of the computing task, including instructions to identify distributed hardware resources present at the distributed information handling systems, wherein the distributed information handling systems include at least one information handling system;
based on the workload attributes determined and the distributed hardware resources identified, determine a preliminary allocation of the computing task among the distributed hardware resources;
send a first request to the distributed information handling systems requesting confirmation of the preliminary allocation;
responsive to the first request, receive a preliminary allocation response from at least some of the distributed information handling systems; and
based on the preliminary allocation response, determine a confirmed allocation of the computing task among the distributed hardware resources, wherein the computing task is executable using the confirmed allocation of the distributed hardw

Assignees

• Dell Products Lp

Inventors

• Farhan Munif M
• Morton Charles G

Classifications

• H04L67/30

  Profiles · CPC title

• G06F9/5083

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

• H04L67/1002

  Electricity · mapped topic

• G06F9/4881

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title

• G06F9/5011 Primary

  the resources being hardware resources other than CPUs, Servers and Terminals · CPC title