Hypervisor-based flash cache space management in a multi-VM environment

• US 9,189,410 B2
• Filed: 05/17/2013
• Issued: 11/17/2015
• Est. Priority Date: 05/17/2013
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Luo, Tian, Liu, Deng
• Primary Examiner(s): Tseng, Cheng-Yuan
• Assistant Examiner(s): Ahmed, Zubair
• Application Number:
  US13/897,269
• Publication Number:
  US 20140344504A1
• Time in Patent Office: 914 Days
• Field of Search: 711/103
• US Class Current: 1/1
• CPC Class Codes: G06F 12/0842 for multiprocessing or multitasking
• CPC Class Codes: G06F 12/0871 Allocation or management of cache space
• CPC Class Codes: G06F 2212/152 Virtualized environment, e.g. logically partitioned system
• CPC Class Codes: G06F 2212/222 Non-volatile memory
• CPC Class Codes: G06F 2212/6042 Allocation of cache space to multiple users or processors

Abstract

Techniques for managing space in a flash storage-based cache are provided. In one embodiment, a computer system can calculate “ratio of effective cache space” (rECS) values for a plurality of VMs, where each VM has a cache allocation comprising a subset of a global pool of cache blocks in the flash storage-based cache, and where the rECS value for the VM indicates a proportion of the subset that has been populated with cached data and re-accessed by the VM within a current time window. The computer system can further determine a new cache allocation size for at least one VM in the plurality of VMs based on the rECS values. The computer system can then adjust the number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size.

Claims

1. A method for managing space in a flash storage-based cache, the method comprising:

calculating, by a computer system, ratio of effective cache space (rECS) values for a plurality of VMs, each VM in the plurality of VMs having a cache allocation comprising a subset of a global pool of cache blocks in the flash storage-based cache, the rECS value for the VM indicating a proportion of the subset that has been populated with cached data and re-accessed by the VM within a current time window;

determining, by the computer system, a new cache allocation size for at least one VM in the plurality of VMs based on the rECS values; and

adjusting, by the computer system, a number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size.

2. The method of claim 1 wherein determining the new cache allocation size for the at least one VM comprises, for each VM on the plurality of VMs:

generating a composite rECS value for the VM by combining the VM'"'"'s rECS value with a previous rECS value calculated for the VM during a previous time window; and

generating a score value for the VM by calculating a difference between the composite rECS value and the previous rECS value.

3. The method of claim 2 wherein determining the new cache allocation size for the at least one VM further comprises:

identifying a first VM in the plurality of VMs having the highest score value and a second VM in the plurality of VMs having the lowest score value;

increasing the first VM'"'"'s cache allocation size; and

if the second VM'"'"'s cache allocation size was not reduced during the previous time window, decreasing the second VM'"'"'s cache allocation size.

4. The method of claim 3 wherein the first VM'"'"'s cache allocation size is increased by an allocation value that is based on an average number of data blocks accessed by the first VM from a backend storage system over one or more recent time windows.

5. The method of claim 1 wherein the computer system maintains an array of metadata entries for the flash storage-based cache, each metadata entry corresponding to a cache block in the global pool of cache blocks and including an identifier that identifies a VM to which the cache block is allocated and a counter indicating whether the cache block has been re-accessed, and

wherein calculating the rECS values for the plurality of VMs comprises sampling a subset of the metadata entries during the current time window.

6. The method of claim 1 wherein adjusting the number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size comprises:

if the new cache allocation size is larger than a previous cache allocation size for the at least one VM, freeing a necessary amount of least recently used cache blocks in the VM'"'"'s cache allocation until the new cache allocation size is reached; and

if the new cache allocation size is smaller than the previous cache allocation size for the at least one VM, claiming free cache blocks in the global pool of cache blocks until the new cache allocation size is reached.

7. The method of claim 6 further comprising:

if a total number of free cache blocks in the global pool of cache blocks falls below a low watermark, freeing least recently used cache blocks until the number of free cache blocks in the global pool reaches a high watermark.

8. A non-transitory computer readable storage medium having stored thereon computer software executable by a processor, the computer software embodying a method for managing space in a flash storage-based cache, the method comprising:

calculating ratio of effective cache space (rECS) values for a plurality of VMs, each VM in the plurality of VMs having a cache allocation comprising a subset of a global pool of cache blocks in the flash storage-based cache, the rECS value for the VM indicating a proportion of the subset that has been populated with cached data and re-accessed by the VM within a current time window;

determining a new cache allocation size for at least one VM in the plurality of VMs based on the rECS values; and

adjusting a number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size.

9. The non-transitory computer readable storage medium of claim 8 wherein determining the new cache allocation size for the at least one VM comprises, for each VM in the plurality of VMs:

generating a composite rECS value for the VM by combining the VM'"'"'s rECS value with a previous rECS value calculated for the VM during a previous time window; and

generating a score value for the VM by calculating a difference between the composite rECS value and the previous rECS value.

10. The non-transitory computer readable storage medium of claim 9 wherein determining the new cache allocation size for the at least one VM further comprises:

identifying a first VM in the plurality of VMs having the highest score value and a second VM in the plurality of VMs having the lowest score value;

increasing the first VM'"'"'s cache allocation size; and

if the second VM'"'"'s cache allocation size was not reduced during the previous time window, decreasing the second VM'"'"'s cache allocation size.

11. The non-transitory computer readable storage medium of claim 10 wherein the first VM'"'"'s cache allocation size is increased by an allocation value that is based on an average number of data blocks accessed by the first VM from a backend storage system over one or more recent time windows.

12. The non-transitory computer readable storage medium of claim 8 wherein the computer system maintains an array of metadata entries for the flash storage-based cache, each metadata entry corresponding to a cache block in the global pool of cache blocks and including an identifier that identifies a VM to which the cache block is allocated and a counter indicating whether the cache block has been re-accessed, and

wherein calculating the rECS values for the plurality of VMs comprises sampling a subset of the metadata entries during the current time window.

13. The non-transitory computer readable storage medium of claim 8 wherein adjusting the number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size comprises:

if the new cache allocation size is larger than a previous cache allocation size for the at least one VM, freeing a necessary amount of least recently used cache blocks in the VM'"'"'s cache allocation until the new cache allocation size is reached; and

if the new cache allocation size is smaller than the previous cache allocation size for the at least one VM, claiming free cache blocks in the global pool of cache blocks until the new cache allocation size is reached.

14. The non-transitory computer readable storage medium of claim 13 further comprising:

if a total number of free cache blocks in the global pool of cache blocks falls below a low watermark, freeing least recently used cache blocks in the global pool until the number of free cache blocks reaches a high watermark.

15. A system for managing space in a flash storage-based cache, the system comprising:

a processor; and

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

calculate ratio of effective cache space (rECS) values for a plurality of VMs, each VM in the plurality of VMs having a cache allocation comprising a subset of a global pool of cache blocks in the flash storage-based cache, the rECS value for the VM indicating a proportion of the subset that has been populated with cached data and re-accessed by the VM within a current time window;

determine a new cache allocation size for at least one VM in the plurality of VMs based on the rECS values; and

adjust a number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size.

16. The system of claim 15 wherein the program code that causes the processor to determine the new cache allocation size for the at least one VM comprises program code that causes the processor to, for each VM in the plurality of VMs:

generate a composite rECS value for the VM by combining the VM'"'"'s rECS value with a previous rECS value calculated for the VM during a previous time window; and

generate a score value for the VM by calculating a difference between the composite rECS value and the previous rECS value.

17. The system of claim 16 wherein the program code that causes the processor to determine the new cache allocation size for the at least one VM further comprises program code that causes the processor to:

identify a first VM in the plurality of VMs having the highest score value and a second VM in the plurality of VMs having the lowest score value;

increase the first VM'"'"'s cache allocation size; and

if the second VM'"'"'s cache allocation size was not reduced during the previous time window, decrease the second VM'"'"'s cache allocation size.

18. The system of claim 17 wherein the first VM'"'"'s cache allocation size is increased by an allocation value that is based on an average number of data blocks accessed by the first VM from a backend storage system over one or more recent time windows.

19. The system of claim 15 wherein the computer system maintains an array of metadata entries for the flash storage-based cache, each metadata entry corresponding to a cache block in the global pool of cache blocks and including an identifier that identifies a VM to which the cache block is allocated and a counter indicating whether the cache block has been re-accessed, and

wherein the program code that causes the processor to calculate the rECS values for the plurality of VMs comprises program code that causes the processor to sample a subset of the metadata entries during the current time window.

20. The system of claim 15 wherein the program code that causes the processor to adjust the number of cache blocks in the at least one VM'"'"'s cache allocation based on the new cache allocation size comprises program code that causes the processor to:

if the new cache allocation size is larger than a previous cache allocation size for the at least one VM, free a necessary amount of least recently used cache blocks in the VM'"'"'s cache allocation until the new cache allocation size is reached; and

if the new cache allocation size is smaller than the previous cache allocation size for the at least one VM, claim free cache blocks in the global pool of cache blocks until the new cache allocation size is reached.

21. The system of claim 20 wherein the program code further causes the processor to:

if a total number of free cache blocks in the global pool of cache blocks falls below a low watermark, free least recently used cache blocks in the global pool until the number of free cache blocks reaches a high watermark.