Page compressibility checker

• US 9,600,317 B2
• Filed: 04/16/2014
• Issued: 03/21/2017
• Est. Priority Date: 04/16/2014
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Guo, Fei, Yedlapalli, Praveen
• Primary Examiner(s): Bertram, Ryan
• Assistant Examiner(s): TA, TRANG KHANH
• Application Number:
  US14/254,696
• Publication Number:
  US 20150301946A1
• Time in Patent Office: 1,070 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 2009/45583 Memory management, e.g. access or allocation
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects

Abstract

Techniques for checking the compressibility of a memory page that is allocated to a virtual machine (VM) running on a host system are provided. In one embodiment, the host system can determine a compression ratio for the memory page by compressing the memory page using a first compression algorithm. The host system can then compare the compression ratio to a threshold. If the compression ratio does not exceed the threshold, the host system can predict that the memory page is compressible by a second compression algorithm that is distinct from the first compression algorithm. On average, the second compression algorithm can be slower, but achieve a lower compression ratio, than the first compression algorithm.

Claims

1. A method for checking the compressibility of virtual machine (VM) memory pages for facilitating host swapping on a host system, the method comprising:

identifying a swap target VM that a hypervisor of the host system has determined to be a target for host swapping;

determining, by the host system, a target number of swap candidate pages to be swapped to disk from an allocated memory space of the swap target VM;

selecting, by the host system, a memory page at random from the allocated memory space of the swap target VM;

determining, by the host system, a compression ratio for the memory page by compressing the memory page using a first compression algorithm;

comparing, by the host system, the compression ratio to a threshold;

if the compression ratio does not exceed the threshold;

predicting, by the host system, that the memory page is compressible by a second compression algorithm that is distinct from the first compression algorithm, wherein on average the second compression algorithm is slower, but achieves a lower compression ratio, than the first compression algorithm; and

adding the memory page to a list of swap candidate pages for the swap target VM;

determining, by the host system subsequently to adding the memory page to the list of swap candidate pages, whether the target number of swap candidate pages has been reached;

if the target number of swap candidate pages has not been reached, determining, by the host system, whether a maximum number of memory pages that the host system may check for compressibility has been reached; and

if the maximum number of memory pages that the host system may check for compressibility has been reached, adding, by the host system, randomly selected memory pages from the allocated memory space of the swap target VM to the list of swap candidate pages until the target number of swap candidate pages is reached.

2. The method of claim 1 wherein the memory page is compressible by the second compression algorithm if the second compression algorithm can achieve a compression ratio for the memory page that is less than or equal to a predetermined target compression ratio.

3. The method of claim 1 wherein, if the host system predicts that the memory page is compressible by the second compression algorithm, the method further comprises:

setting a compressibility bit in metadata for the memory page.

4. The method of claim 3 further comprising dynamically adjusting the threshold during runtime of the host system.

5. The method of claim 4 wherein dynamically adjusting the threshold comprises, at a time of processing the list of swap candidate pages, for each swap candidate page:

compressing the swap candidate page using the second compression algorithm;

determining, based on the compressing, whether the swap candidate page is compressible by the second compression algorithm;

if the swap candidate page is not compressible by the second compression algorithm and the compressibility bit in the swap candidate page is set, detecting a false positive result; and

if the swap candidate page is compressible by the second compression algorithm and the compressibility bit in the swap candidate page is not set, detecting a false negative result.

6. The method of claim 5 wherein dynamically adjusting the threshold further comprises:

calculating a percentage of false positive results and a percentage of false negative results;

if the percentage of false positive results exceeds a false positive threshold, decreasing the threshold by a predetermined amount; and

if the percentage of false negative results exceeds a false negative threshold, increasing the threshold by the predetermined amount.

7. The method of claim 1 wherein the first compression algorithm is configured to, at the time of compressing the memory page:

track compression output size for the memory page, without writing any compression output to an output buffer; and

abort the compressing before completion if the first compression algorithm determines, based on a current compression output size and a remaining uncompressed data size, that the compression ratio will not exceed the threshold.

22. The method of claim 2 wherein the predetermined target compression ratio is based on a compression factor that is needed to fit the memory page into a compression cache block used by the hypervisor.

8. A non-transitory computer readable storage medium having stored thereon software executable by a host system, the software embodying a method for checking the compressibility of virtual machine (VM) memory pages for facilitating host swapping on the host system, the method comprising:

identifying a swap target VM that a hypervisor of the host system has determined to be a target for host swapping;

determining, by the host system, a target number of swap candidate pages to be swapped to disk from an allocated memory space of the swap target VM;

selecting, by the host system, a memory page at random from the allocated memory space of the swap target VM;

determining a compression ratio for the memory page by compressing the memory page using a first compression algorithm;

comparing the compression ratio to a threshold;

if the compression ratio does not exceed the threshold;

predicting that the memory page is compressible by a second compression algorithm that is distinct from the first compression algorithm, wherein on average the second compression algorithm is slower, but achieves a lower compression ratio, than the first compression algorithm; and

adding the memory page to a list of swap candidate pages for the swap target VM;

determining, subsequently to adding the memory page to the list of swap candidate pages, whether the target number of swap candidate pages has been reached;

if the target number of swap candidate pages has not been reached, determining whether a maximum number of memory pages that the host system may check for compressibility has been reached; and

if the maximum number of memory pages that the host system may check for compressibility has been reached, adding randomly selected memory pages from the allocated memory space of the swap target VM to the list of swap candidate pages until the target number of swap candidate pages is reached.

9. The non-transitory computer readable storage medium of claim 8 wherein the memory page is compressible by the second compression algorithm if the second compression algorithm can achieve a compression ratio for the memory page that is less than or equal to a predetermined target compression ratio.

10. The non-transitory computer readable storage medium of claim 8 wherein, if the host system predicts that the memory page is compressible by the second compression algorithm, the method further comprises:

setting a compressibility bit in metadata for the memory page.

11. The non-transitory computer readable storage medium of claim 10 wherein the method further comprises dynamically adjusting the threshold during runtime of the host system.

12. The non-transitory computer readable storage medium of claim 11 wherein dynamically adjusting the threshold comprises, at a time of processing the list of swap candidate pages, for each swap candidate page:

compressing the swap candidate page using the second compression algorithm;

determining, based on the compressing, whether the swap candidate page is compressible by the second compression algorithm;

if the swap candidate page is not compressible by the second compression algorithm and the compressibility bit in the swap candidate page is set, detecting a false positive result; and

if the swap candidate page is compressible by the second compression algorithm and the compressibility bit in the swap candidate page is not set, detecting a false negative result.

13. The non-transitory computer readable storage medium of claim 12 wherein dynamically adjusting the threshold further comprises:

calculating a percentage of false positive results and a percentage of false negative results;

if the percentage of false positive results exceeds a false positive threshold, decreasing the threshold by a predetermined amount; and

if the percentage of false negative results exceeds a false negative threshold, increasing the threshold by the predetermined amount.

14. The non-transitory computer readable storage medium of claim 8 wherein the first compression algorithm is configured to, at the time of compressing the memory page:

track compression output size for the memory page, without writing any compression output to an output buffer; and

abort the compressing before completion if the first compression algorithm determines, based on a current compression output size and a remaining uncompressed data size, that the compression ratio will not exceed the threshold.

23. The non-transitory computer readable storage medium of claim 9 wherein the predetermined target compression ratio is based on a compression factor that is needed to fit the memory page into a compression cache block used by the hypervisor.

15. A computer system comprising:

a processor; and

a non-transitory computer readable medium having stored thereon program code that causes the processor to, upon being executed;

identify a swap target VM that a hypervisor of the computer system has determined to be a target for host swapping;

determine a target number of swap candidate pages to be swapped to disk from an allocated memory space of the swap target VM;

select a memory page at random from the allocated memory space of the swap target VM;

determine a compression ratio for the memory page by compressing the memory page using a first compression algorithm;

compare the compression ratio to a threshold;

if the compression ratio does not exceed the threshold;

predict that the memory page is compressible by a second compression algorithm that is distinct from the first compression algorithm, wherein on average the second compression algorithm is slower, but achieves a lower compression ratio, than the first compression algorithm; and

add the memory page to a list of swap candidate pages for the swap target VM;

determine, subsequently to adding the memory page to the list of swap candidate pages, whether the target number of swap candidate pages has been reached;

if the target number of swap candidate pages has not been reached, determine whether a maximum number of memory pages that the host system may check for compressibility has been reached; and

if the maximum number of memory pages that the host system may check for compressibility has been reached, add randomly selected memory pages from the allocated memory space of the swap target VM to the list of swap candidate pages until the target number of swap candidate pages is reached.

16. The computer system of claim 15 wherein the memory page is compressible by the second compression algorithm if the second compression algorithm can achieve a compression ratio for the memory page that is less than or equal to a predetermined target compression ratio.

17. The computer system of claim 15 wherein, if the processor predicts that the memory page is compressible by the second compression algorithm, the program code further causes the processor to:

set a compressibility bit in metadata for the memory page.

18. The computer system of claim 17 wherein the program code further causes the processor to dynamically adjust the threshold during runtime of the computer system.

19. The computer system of claim 18 wherein the program code that causes the processor to dynamically adjust the threshold comprises program code that causes the processor to, at a time of processing the list of swap candidate pages, for each swap candidate page:

compress the swap candidate page using the second compression algorithm;

determine, based on the compressing, whether the swap candidate page is compressible by the second compression algorithm;

if the swap candidate page is not compressible by the second compression algorithm and the compressibility bit in the swap candidate page is set, detect a false positive result; and

if the swap candidate page is compressible by the second compression algorithm and the compressibility bit in the swap candidate page is not set, detect a false negative result.

20. The computer system of claim 19 wherein the program code that causes the processor to dynamically adjust the threshold further comprises program code that causes the processor to:

calculate a percentage of false positive results and a percentage of false negative results;

if the percentage of false positive results exceeds a false positive threshold, decrease the threshold by a predetermined amount; and

if the percentage of false negative results exceeds a false negative threshold, increase the threshold by the predetermined amount.

21. The computer system of claim 15 wherein the first compression algorithm is configured to, at the time of compressing the memory page:

track compression output size for the memory page, without writing any compression output to an output buffer; and

abort the compressing before completion if the first compression algorithm determines, based on a current compression output size and a remaining uncompressed data size, that the compression ratio will not exceed the threshold.

24. The computer system of claim 16 wherein the predetermined target compression ratio is based on a compression factor that is needed to fit the memory page into a compression cache block used by the hypervisor.