Power management system and method implementing interconnect controls

Patent metadata

Field	Value
Publication number	US-10877542-B2
Application number	US-201916295992-A
Country	US
Kind code	B2
Filing date	Mar 7, 2019
Priority date	Mar 7, 2019
Publication date	Dec 29, 2020
Grant date	Dec 29, 2020

Abstract

The present disclosure provides a system and a method for power management of accelerators using interconnect configuration. The method comprises receiving a power management command comprising a power budget and a designated accelerator, identifying a port associated with the designated accelerator from a plurality of ports of an interconnect, determining a target data transmission parameter for the designated accelerator according to the power budget, and controlling a data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator. The present disclosure further provides a non-transitory computer-readable medium that stores a set of instructions that are executable by one or more processors of an apparatus to perform the method for power management of accelerators using interconnect configuration.

First claim

What is claimed is:

1. A method for power management, comprising:
receiving a power management command, the power management command comprising a power budget and a designated accelerator;
identifying, from a plurality of ports of an interconnect, a port associated with the designated accelerator;
determining a target data transmission parameter for the designated accelerator according to the power budget; and
controlling a data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator.

2. The method of claim 1 , wherein identifying, from the plurality of ports of an interconnect, the port associated with the designated accelerator further comprising traversing through the interconnect to locate a port connected to the designated accelerator.

3. The method of claim 1 , wherein controlling the data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator further comprising:
determining whether the designated accelerator has exceeded the power budget; and
pausing sending data to the designated accelerator through the interconnect in response to the determination that the power budget has been exceeded; or
sending data to the designated accelerator through the interconnect in response to the determination that the power budget has not been exceeded.

4. The method of claim 1 , wherein the power management command further comprising multiple designated accelerators.

5. The method of claim 4 , further comprising determining different target data transmission parameters for different ports on the interconnect that are associated with the designated accelerators.

6. The method of claim 1 , wherein the data comprising training weights in an algorithm of artificial neural network.

7. The method of claim 1 , wherein controlling the data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator further comprising configuring the interconnect to operate relative to the target data transmission parameter on the port.

8. The method of claim 1 , wherein the interconnect comprises a peripheral component interconnect express (PCIe) bus or a mesh network.

9. A computer system, comprising:
an interconnect;
one or more accelerators;
one or more processors coupled to the one or one or more accelerators through the interconnect, the one or more processors configured to:
receive a power management command, the power management command comprising a power budget and a designated accelerator from one or more accelerators;
identify, from a plurality of ports of an interconnect, a port associated with the designated accelerator;
determine a target data transmission parameter for the designated accelerator according to the power budget; and
control a data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator.

10. The computer system of claim 9 , wherein the processor is further configured to traverse through the interconnect to locate a port connected to the designated accelerator.

11. The computer system of claim 9 , wherein the processor is further configured to:
determine whether the designated accelerator has exceeded the power budget; and
pause sending data to the designated accelerator through the interconnect in response to the determination that the power budget has been exceeded; or
send data to the designated accelerator through the interconnect in response to the determination that the power budget has not been exceeded.

12. The computer system of claim 9 , wherein the power management command further comprising multiple designated accelerators.

13. The computer system of claim 12 , wherein the processor is further configured to determine different target data transmission parameters for different ports on the interconnect that are associated with the designated accelerators.

14. The computer system of claim 9 , wherein the computer system is configured to be implemented in a cloud environment.

15. The computer system of claim 14 , wherein the processor is configured to operate by following instructions from an operating system of a data center in the cloud environment.

16. The computer system of claim 15 , wherein the operating system is Linux kernel and the operating system comprises a kernel-based virtual machine (KVM).

17. The computer system of claim 11 , wherein the data comprising training weights in an algorithm of artificial neural network.

18. The computer system of claim 9 , wherein the designated accelerator is one or more graphics processing units (GPUs).

19. The computer system of claim 9 , wherein the processor is further configured to configure the interconnect to operate at a data transmission parameter relative to the target data transmission parameter on the port.

20. The computer system of claim 9 , wherein the interconnect is a peripheral component interconnect express (PCIe) bus or a mesh network).

21. A non-transitory computer readable medium that stores a set of instructions that are executable by one or more processors of an apparatus to perform a method comprising:
receiving a power management command, the power management command comprising a power budget and a designated accelerator;
identifying, from a plurality of ports of an interconnect, a port associated with the designated accelerator;
determining a target data transmission parameter for the designated accelerator according to the power budget; and
controlling a data transmission parameter relative to the target data transmission parameter through the port associated with the designated accelerator.

22. The medium of claim 21 , wherein the set of instructions that are executable by one or more processors of the apparatus to further perform traversing through the interconnect to locate a port connected to the designated accelerator.

23. The medium of claim 21 , wherein the set of instructions that are executable by one or more processors of the apparatus to further perform:
determining whether the designated accelerator has exceeded the power budget; and
pausing sending data to the designated accelerator through the interconnect in response to the determination that the power budget has been exceeded; or
sending data to the designated accelerator through the interconnect in response to the determination that the power budget has not been exceeded.

24. The medium of claim 21 , wherein the power management command further comprising multiple designated accelerators.

25. The medium of claim 24 , wherein the set of instructions that are executable by one or more processors of the apparatus to further perform determining different target data transmission parameters for different ports on the interconnect that are associated with the designated accelerators.

26. The medium of claim 21 , wherein the data comprising training weights in an algorithm of artificial neural network.

27. The medium of claim 21 , wherein the set of instructions that are executable by one or more processors of the apparatus to further perform configuring the interconnect to operate at a data transmission parameter relative to the target data transmission parameter on the port.

28. The medium

Assignees

• Alibaba Group Holding Ltd

Inventors

• Jin Lingling

Classifications

• Y02D10/00

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

• G06N3/08

  Learning methods · CPC title

• G06F1/3253

  Power saving in bus · CPC title

• G06F1/329

  by task scheduling · CPC title

• G06F1/3206 Primary

  Monitoring of events, devices or parameters that trigger a change in power modality · CPC title