Writeback overhead reduction for workloads

Patent metadata

Field	Value
Publication number	US-11914512-B2
Application number	US-202117449217-A
Country	US
Kind code	B2
Filing date	Sep 28, 2021
Priority date	Sep 28, 2021
Publication date	Feb 27, 2024
Grant date	Feb 27, 2024

Abstract

An example system includes a memory, a processor in communication with the memory, and a supervisor. The supervisor is configured to allocate a memory space in the memory to a workload executing on the processor. The supervisor is configured to store data written by the workload as dirty memory in the memory space at least until the data is written back to a data storage. Based on a type of the workload being a first type, the supervisor is configured to trigger write back of at least a portion of the dirty memory into the data storage in response to the dirty memory exceeding a threshold level. Based on the type of the workload being a second type, the supervisor is configured to delay write back of the dirty memory into the data storage in response to the dirty memory exceeding the threshold level.

First claim

What is claimed is:

1. A system comprising:
a memory;
a processor in communication with the memory; and
a supervisor, wherein the supervisor is configured to:
allocate a memory space in the memory to a workload executing on the processor;
store, as dirty memory in the memory space allocated to the workload, data written by the workload at least until the data is written back into a data storage;
based on a type of the workload being a first type, trigger write back of at least a portion of the dirty memory into the data storage in response to the dirty memory exceeding a first threshold level; and
based on the type of the workload being a second type,
delay write back of the dirty memory into the data storage in response to the dirty memory exceeding the first threshold level, and
continue execution of at least one thread of the workload on the processor in response to the dirty memory exceeding a second threshold level greater than the first threshold level,

wherein the workload of the first type is more sensitive to data loss and data consistency risks than the workload of the second type.

2. The system of claim 1 , wherein the first and second threshold levels are threshold ratios of amounts of the dirty memory of the workload relative to a total amount of the memory space allocated to the workload.

3. The system of claim 1 , further comprising:
a node, wherein the processor and the memory are disposed in the node, wherein the workload includes a container deployed at the node, and wherein the type of the workload is based on a configuration parameter assigned to the container prior to deployment of the container at the node.

4. The system of claim 1 , wherein the supervisor is configured to:
retrieve metadata associated with the workload, wherein the metadata includes a configuration parameter selected for the workload prior to dispatch of the workload for execution on the processor; and
determine the type of the workload based on the configuration parameter in the metadata.

5. The system of claim 1 , wherein the memory is a random access memory (RAM).

6. The system of claim 5 , wherein the data storage is a disk, and wherein triggering the write back causes the supervisor to move at least a portion of the data from the RAM to the disk.

7. The system of claim 1 , wherein the supervisor is a hypervisor.

8. The system of claim 1 , wherein the supervisor is configured to:
based on the type of the workload being the first type, delay execution of the at least one thread of the workload on the processor until the at least portion of the dirty memory is written back into the data storage in response to the dirty memory exceeding the second threshold level.

9. The system of claim 1 , wherein the supervisor is configured to:
assign, to the workload, a third threshold level greater than the first threshold level based on the type of the workload being the second type; and
in response to the dirty memory of the workload exceeding the third threshold level and the type of the workload being the second type, trigger the write back of the at least portion of the dirty memory of the workload into the data storage.

10. A method comprising:
allocating memory space in a memory to a workload executing on a processor, wherein the processor is in communication with the memory;
storing, as dirty memory in the memory space allocated to the workload, data processed by the workload at least until the data is written back to a data storage;
based on a type of the workload being a first type, writing out at least a portion of the dirty memory to the data storage in response to the dirty memory exceeding a first threshold level; and
based on the type of the workload being a second type,
delaying write out of the dirty memory to the data storage in response to the dirty memory exceeding the first threshold level, and
continue execution of at least one thread of the workload on the processor in response to the dirty memory exceeding a second threshold level greater than the first threshold level,

wherein the workload of the first type is more sensitive to data loss and data consistency risks than the workload of the second type.

11. The method of claim 10 , wherein the first and second threshold levels are threshold ratios of amounts of the dirty memory of the workload relative to a total amount of the memory space allocated to the workload.

12. The method of claim 10 , further comprising:
retrieving metadata associated with the workload, wherein the metadata includes a configuration parameter assigned to the workload; and
determining the type of the workload based on the configuration parameter in the metadata.

13. The method of claim 10 , wherein the memory is a random access memory (RAM).

14. The method of claim 13 , wherein the data storage is a disk, and wherein writing out the at least portion of the dirty memory includes transferring at least a portion of the data from the RAM to the disk.

15. The method of claim 10 , further comprising:
based on the type of the workload being the first type, delaying execution of the at least one thread of the workload in response to the dirty memory exceeding the second threshold level.

16. The method of claim 15 , wherein delaying the execution of the at least one thread of the workload includes delaying the execution until completion of the writing back of the at least portion of the dirty memory to the data storage.

17. The method of claim 10 , the method further comprising:
assigning, to the workload, a third threshold level greater than the first threshold level based on the type of the workload being the second type; and
in response to the dirty memory of the workload exceeding the third threshold level and the type of the workload being the second type, writing back the at least portion of the dirty memory of the workload to the data storage.

18. A non-transitory machine-readable medium storing instructions that, when executed by at least one processor of a computer system, cause the computer system to:
allocate memory space in a memory to a workload executing on a processor, wherein the processor is in communication with the memory;
store, as dirty memory pages in the memory space allocated to the workload, data from the workload at least until the data is written out to a data storage;
based on a type of the workload being a first type, write at least a portion of the dirty memory pages to the data storage in response to the dirty memory pages exceeding a first threshold level; and
based on the type of the workload being a second type,
delay writing the dirty memory pages to the data storage in response to the dirty memory pages exceeding the first threshold level, and
continue execution of at least one thread of the workload on the processor in response to the dirty memory exceeding a second threshold level greater than the first threshold level,

wherein the workload of the first type is more sensitive to data loss and data consistency risks than the workload of the second type.

Assignees

• Red Hat Inc

Inventors

• Tsirkin Michael
• Arcangeli Andrea
• Scrivano Giuseppe

Classifications

• G06F12/0804 Primary

  with main memory updating (G06F12/0806 takes precedence) · CPC title

• G06F9/45558 Primary

  Hypervisor-specific management and integration aspects · CPC title

• G06F9/5016

  the resource being the memory · CPC title

• G06F2009/45583

  Memory management, e.g. access or allocation · CPC title

• G06F12/0868

  Data transfer between cache memory and other subsystems, e.g. storage devices or host systems · CPC title