Migration of graphics processing unit (gpu) states

Patent metadata

Field	Value
Publication number	US-2016247248-A1
Application number	US-201514630830-A
Country	US
Kind code	A1
Filing date	Feb 25, 2015
Priority date	Feb 25, 2015
Publication date	Aug 25, 2016
Grant date	—

Abstract

The claimed subject matter includes techniques for live migration of a graphics processing unit (GPU) state. An example method includes receiving recorded GPU commands from a relay at a destination GPU. The method also includes replaying the recorded GPU commands at the destination GPU. The method also includes detecting a downtime for the GPU commands. The method further includes establishing a connection between the destination GPU and the client during the detected downtime.

First claim

What is claimed is:

1 . A system for live migration of a graphics processing unit (GPU) state, comprising:
a first module to record GPU commands at a source GPU, the first module to send the GPU commands to a destination GPU, the destination GPU to replay the recorded GPU commands; a second module to predict a downtime for the GPU commands at the source GPU that is below a threshold time; and a third module to connect a client to the destination GPU at the detected downtime.

2 . The system of claim 1 , the first module to further record a snapshot of a GPU state of the source GPU, the first module to also record the GPU commands at the source GPU in response to recording the snapshot.

3 . The system of claim 2 , further comprising a fourth module to generate a map linking GPU resource identifiers and logical resource identifiers, each of the GPU commands comprising an object with a GPU resource identifier that is to be mapped to a logical resource identifier.

4 . The system of claim 3 , the third module to utilize the map to schedule a plurality of threads for execution, each thread comprising one or more logical resource identifiers.

5 . The system of claim 1 , the third module comprising a relay encapsulated in a virtual machine, the relay communicatively coupled to a guest virtual machine (VM) via a Transmission Control Protocol (TCP) link, the relay to create a new TCP connection to a new renderer and send a GPU state log to the new renderer before switching the TCP link of the guest virtual machine to the new renderer, the new renderer communicatively coupled to the destination GPU.

6 . The system of claim 1 , the destination GPU to be located in a machine that is closer in proximity to a client device than the source GPU, the client device to produce the GPU commands.

7 . The system of claim 1 , the guest VM comprising a guest operating system (OS) that is to run one or more applications or games.

8 . The system of claim 1 , the GPU commands to be recorded to a log file with corresponding timestamps.

9 . A method for live migration of a graphics processing unit (GPU) state, the method comprising:
receiving recorded GPU commands from a relay at a destination GPU; replaying the recorded GPU commands at the destination GPU; detecting a downtime for the GPU commands; establishing a connection between the destination GPU and the client during the detected downtime.

10 . The method of claim 9 , further comprising:
receiving a snapshot of a GPU state of the source GPU, the recorded GPU commands comprising commands received at the source GPU after the snapshot was created; and restoring the snapshot on the destination GPU before replaying the recorded GPU commands.

11 . The method of claim 9 , further comprising receiving a map linking GPU resource identifiers to logical resource identifiers, each of the GPU commands comprising an object with a GPU resource identifier that is to be mapped to a particular logical resource identifier.

12 . The method of claim 11 , further comprising utilizing the map to order GPU commands in a thread for execution, each thread containing one or more logical resource identifiers.

13 . The method of claim 9 , the destination GPU located in a machine that is closer in proximity to the client than the source GPU.

14 . The method of claim 9 , further comprising processing GPU commands from a plurality of clients at the destination GPU.

15 . The method of claim 14 , further comprising dynamically allocating resources by detecting an underutilized GPU and migrating the client to the underutilized GPU.

16 . The method of claim 14 , further comprising dynamically allocating resources between a software renderer and a hardware renderer by migrating the client between the software renderer and the hardware renderer.

17 . The method of claim 14 , further comprising dynamically allocating resources by processing GPU commands from a plurality of clients at the destination GPU.

18 . One or more computer-readable memory storage devices for storing computer readable instructions that, when executed by one or more processing devices, instruct live migration of a graphics processing unit (GPU) state, the computer-readable instructions comprising code to:
record a snapshot of a GPU state of a source GPU; record GPU commands from a virtual machine at the source GPU and store the recorded GPU commands to a log file; send the log file and the snapshot to a destination GPU; predict a downtime for the GPU commands at the source GPU; switch a relay to connect the client to the destination GPU during the predicted downtime.

19 . The one or more computer-readable memory storage devices of claim 18 , the computer-readable instructions comprising code to generate a map linking GPU resource identifiers and logical resource identifiers, each of the GPU commands comprising an object with a GPU resource identifier that is to be mapped to a logical resource identifier.

20 . The one or more computer-readable memory storage devices of claim 18 , the computer-readable instructions comprising code to create a new set of Transmission Control Protocol (TCP) links and restore a GPU state of the source GPU on a destination GPU before the relay switches to the destination GPU.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Ha Kiryong
• Bahl Paramvir
• Chu David Chiyuan
• Cuervo Eduardo
• Sivalingam Lenin Ravindranath
• Wolman Alastair

Classifications

• G06T1/20 Primary

  Processor architectures; Processor configuration, e.g. pipelining · CPC title

• G06F9/48

  Program initiating; Program switching, e.g. by interrupt · CPC title

• G06F9/455

  Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines · CPC title

• G06F9/4856

  resumption being on a different machine, e.g. task migration, virtual machine migration (G06F9/5088 takes precedence) · CPC title

• G06F2009/4557

  Distribution of virtual machine instances; Migration and load balancing · CPC title