Workload prediction based CPU frequency scaling

Patent metadata

Field	Value
Publication number	US-11119529-B2
Application number	US-202016748486-A
Country	US
Kind code	B2
Filing date	Jan 21, 2020
Priority date	Mar 1, 2018
Publication date	Sep 14, 2021
Grant date	Sep 14, 2021

Abstract

A method comprises predicting, based on corresponding historical workload data, a change in virtual network function demand during a future workload period, wherein the virtual network function is supported by a node. The method further comprises determining a target clock speed of one or more physical CPU cores of one or more processors of one or more servers in the node corresponding to the change in the virtual network function demand and adjusting the CPU CORE of the node to the target clock speed corresponding to the change in the virtual network function demand for the future workload period.

First claim

What is claimed is:

1. An apparatus comprising:
a processor; and
a memory coupled with the processor, the memory storing executable instructions that when executed by the processor cause the processor to effectuate operations comprising:
predicting, based on historical workload data, a predicted change in virtual network function demand during a future workload period, wherein the virtual network function is supported by a node;
determining a target clock speed of a physical central processing unit (CPU) core in the node corresponding to the predicted change in the virtual network function demand;
calculating a latency, wherein the latency is based on:
an adjustment delay associated with changing a clock speed of the physical CPU core from an initial clock speed to the target clock speed, and
a network delay; and

adjusting the physical CPU core of the node to the target clock speed corresponding to the predicted change in the virtual network function demand for the future workload period, wherein adjusting the physical CPU core of the node is coordinated to the latency.

2. The apparatus of claim 1 , the operations further comprising collecting the historical workload data.

3. The apparatus of claim 1 , the operations further comprising collecting the historical workload data, wherein the historical workload data is based on a time period.

4. The apparatus of claim 1 , the operations further comprising collecting the historical workload data, wherein the historical workload data is based on a time period, wherein the time period is a date and time.

5. The apparatus of claim 1 , the operations further comprising comparing a recent virtual network function demand to the historical workload data to determine a difference between the recent virtual network function demand and the historical workload data, wherein predicting the change in virtual network function demand is further based on the difference.

6. The apparatus of claim 5 , wherein the comparing the recent virtual network function demand to the historical workload data is based on a date or a time.

7. The apparatus of claim 5 , wherein the comparing the recent virtual network function demand to the historical workload data is based on a usage pattern.

8. The apparatus of claim 5 , wherein the comparing the recent virtual network function demand to the historical workload data identifies an anomalous event.

9. A method comprising:
predicting by a server, based on historical workload data, a future demand during a future workload interval, wherein the future demand supports a virtual network function or container, and wherein the future demand is met by a server;
determining by the server, a target clock speed of at least one hyper-thread of a central processing unit (CPU) core corresponding to the future demand, wherein two or more hyper-threads are supported by the CPU core corresponding to the future demand; and
causing by the server, the at least one hyper-thread of the CPU core to adjust to the target clock speed corresponding to the future demand for the future workload interval.

10. The method of claim 9 , further comprising grouping cores for common scaling.

11. The method of claim 9 , further comprising grouping hyper-threads for common scaling.

12. An apparatus comprising:
a processor; and
a memory coupled with the processor, the memory storing executable instructions that when executed by the processor cause the processor to effectuate operations comprising:
comparing a recent virtual network function demand to historical workload data to determine a difference between the recent virtual network function demand and the historical workload data, wherein the comparison identifies an anomalous event;
predicting, based on historical workload data and the difference, a predicted change in virtual network function demand during a future workload period, wherein the virtual network function is supported by a node;
determining a target clock speed of a physical central processing unit (CPU) core in the node corresponding to the predicted change in the virtual network function demand; and
adjusting the physical CPU core of the node to the target clock speed corresponding to the predicted change in the virtual network function demand for the future workload period.

13. The apparatus of claim 12 , wherein adjusting the physical CPU core of the node is coordinated to a latency.

14. The apparatus of claim 13 , the operations further comprising calculating the latency, wherein the latency is based on an adjustment delay associated with changing a clock speed of the physical CPU core from an initial clock speed to the target clock speed.

15. The apparatus of claim 14 , wherein the latency is based on a network delay.

16. The apparatus of claim 12 , the operations further comprising collecting the historical workload data.

17. The apparatus of claim 16 , wherein the historical workload data is based on a time period.

18. The apparatus of claim 17 , wherein the time period is a date and time.

19. The apparatus of claim 12 , wherein the comparing the recent virtual network function demand to the historical workload data is based on a date or a time.

20. The apparatus of claim 12 , wherein the comparing the recent virtual network function demand to the historical workload data is based on a usage pattern.

Assignees

• At & T Ip I Lp

Inventors

• Seenappa Vikram
• Mhatre Vivek

Classifications

• G06F1/08 Primary

  Clock generators with changeable or programmable clock frequency · CPC title

• G06F2009/45595

  Network integration; Enabling network access in virtual machine instances · CPC title

• G06F9/45558

  Hypervisor-specific management and integration aspects · CPC title

• G06F1/324 Primary

  by lowering clock frequency · CPC title

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title