Machine-learning-based aggregation of activation prescriptions for scalable computing resource scheduling

Patent metadata

Field	Value
Publication number	US-10999212-B2
Application number	US-201916423720-A
Country	US
Kind code	B2
Filing date	May 28, 2019
Priority date	May 28, 2019
Publication date	May 4, 2021
Grant date	May 4, 2021

Abstract

A multi-layer resource aggregation (RA) stack may generate prescriptive activation timetables for controlling activation states for computing resources. To facilitate operator control and adjustment, the RA stack may, at an aggregation engine layer, aggregate the computing resource into one or more resource aggregates. The computing resources within the resource aggregates may have similar individual activation prescription patterns. Machine learning techniques may be used by the RA stack to identify these related individual activation prescription patterns and aggregate the computing resources accordingly. Once aggregated, the RA stack may make a uniform activation determination for the aggregates as single units. Therefore, the computing resources within the aggregate may be controlled and/or adjust together. Thus, the RA stack increases the scalability of implementation of prescriptive computing resource activation state determinations.

First claim

What is claimed is:

1. A system including:
network interface circuitry configured to:
receive historical utilization data for multiple computing resources; and
receive activation schedule data for the multiple computing resources; and

aggregation circuitry in data communication with the network interface circuitry, the aggregation circuitry configured to execute a resource aggregation (RA) stack,
the RA stack executable to:
parse the historical utilization data to generate interval-scaled data for a selected interval;
generate, for each of the multiple computing resources, a set of activation determinations corresponding to multiple timeslots, each of the multiple timeslots with a length equal to the selected interval;
generate distance metrics by comparing the sets of activation determinations to determine spacings among individual ones of the multiple computing resources;
based on the distance metrics, determine a set of summary vectors detailing the spacings among the individual ones of the multiple computing resources;
based on the distance metrics and the summary vectors, aggregate the multiple computing resources in to a selected number of resource aggregates, the selected number being less than a total number of the multiple computing resources, such that each computing resource aggregated into a resource aggregate of the selected number of resource aggregates is controllable via aggregate-level management of the aggregated computing resources including uniform activation and deactivation along with each other computing resource included the same resource aggregate; and
for each of the resource aggregates:
determine a uniform activation determination for a corresponding future timeslot, the future timeslot having the length equal to selected interval; and
generate entries in an activation timetable configured to requisition application of the uniform activation determination uniformly to each of the multiple computing resources within the resource aggregate, such that activations for individual ones of the multiple computing resources are reviewable together by reviewing the uniform activation determination.

2. The system of claim 1 , where the set of summary vectors characterize periods above and below an activation recommendation threshold for the multiple computing resources within the multiple timeslots.

3. The system of claim 1 , where the distance metrics are based on:
grouping the summary vectors into groups; and
determining distances between centers for the groups.

4. The system of claim 3 , where the RA stack is further executable to aggregate the multiple computing resources by selecting the resource aggregates to achieve a base least-squares value.

5. The system of claim 1 , where the RA stack is further executable to regroup the multiple computing resource into a specific number of resource aggregates after making uniform activation determinations for each of the selected number of resource aggregates, where:
the specific number is different from the selected number; and
the specific number is less than the total number of the multiple computing resources.

6. The system of claim 1 , where the RA stack is further executable to, for each of the resource aggregates, determine the uniform activation determination based on a first aggregate activation threshold.

7. The system of claim 6 , where the RA stack is further executable to, for each of the resource aggregates:
determine to make a uniform activation determination that activates each computing resource in the resource aggregate when a number of activation recommendations for the computing resources in the resource aggregate exceeds the first aggregate activation threshold; and
determine to make a uniform activation determination that deactivates each computing resource in the resource aggregate when the number of activation recommendations for the computing resources in the resource aggregate is below the first aggregate activation threshold.

8. The system of claim 6 , where the RA stack is further executable to, for each of the resource aggregates, re-determine the uniform activation determination based on a second aggregate activation threshold different that the first aggregate activation threshold to simulate uniform activation determinations for multiple different activation thresholds.

9. The system of claim 1 , where the RA stack is further executable to generate an RA control interface configured to accept operator selections uniform activation determination parameters.

10. The system of claim 9 , where the RA control interface includes an RA-window presentation configured to display information on uniform activation determinations, uniform activation determination parameters, or both.

11. A method including:
at network interface circuitry:
receiving historical utilization data for multiple computing resources; and
receiving activation schedule data for the multiple computing resources; and

at a resource aggregation (RA) stack aggregation circuitry in data communication with the network interface circuitry:
parsing the historical utilization data to generate interval-scaled data for a selected interval;
generating, for each of the multiple computing resources, a set of activation determinations corresponding to multiple timeslots, each of the multiple timeslots with a length equal to the selected interval;
generating distance metrics by comparing the sets of activation determinations to determine spacings among individual ones of the multiple computing resources;
based on the distance metrics, determine a set of summary vectors detailing the spacings among the individual ones of the multiple computing resources;
based on the distance metrics and the summary vectors, aggregating the multiple computing resources in to a selected number of resource aggregates, the selected number being less than a total number of the multiple computing resources, such that each computing resource aggregated into a resource aggregate of the selected number of resource aggregates is controllable via aggregate-level management of the aggregated computing resources including uniform activation and deactivation along with each other computing resource included the same resource aggregate; and
for each of the resource aggregates:
determining a uniform activation determination for a corresponding future timeslot, the future timeslot having the length equal to selected interval; and
generating entries in an activation timetable configured to requisition application of the uniform activation determination uniformly to each of the multiple computing resources within the resource aggregate, such that activations for individual ones of the multiple computing resources are reviewable together by reviewing the uniform activation determination.

12. The method of claim 11 , where determining a uniform activation determination for each of the resource aggregates includes determining the uniform activation determination based on a first aggregate activation threshold.

13. The method of claim 12 , where determining a uniform activation determination for each of the resource aggregates includes:
determining to make a uniform activation determination that activates each computing resource in the resource aggregate when a number of activation recommendations for the computing resources in the resource aggregate exceeds the first aggregate activation threshold; and
determining to make a uniform activation determination that deactivates each computing resource in the resource aggregate when the number of activation recommendations for the computing resources in the resource agg

Assignees

• Accenture Global Solutions Ltd

Inventors

• Srinivasan Madhan Kumar
• Purushothaman Arun
• Kolachalam Manish Sharma

Classifications

• H04L47/83

  based on usage prediction · CPC title

• G06F9/5077 Primary

  Logical partitioning of resources; Management or configuration of virtualized resources (specific details on emulation or internal functioning of virtual machines G06F9/455) · CPC title

• G06N20/00

  Machine learning · CPC title

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• G06F9/5027

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title