COLLABORATIVE CACHING

• US 20140325161A1
• Filed: 04/27/2012
• Published: 10/30/2014
• Est. Priority Date: 04/27/2012
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Inventors: Voigt, Douglas L.
• Granted Patent: US 9,372,810 B2
• Time in Patent Office: Days
• US Class Current: 711/136
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3466 Performance evaluation by tracing or monitoring
• CPC Class Codes: G06F 12/0866 for peripheral storage systems, e.g. disk cache
• CPC Class Codes: G06F 12/122 of the least frequently used [LFU] type, e.g. with individual count value
• CPC Class Codes: G06F 2201/81 Threshold
• CPC Class Codes: G06F 2201/88 Monitoring involving counting
• CPC Class Codes: G06F 2201/885 Monitoring specific for caches
• CPC Class Codes: G06F 2212/311 In host system
• CPC Class Codes: G06F 2212/313 In storage device
• CPC Class Codes: G06F 2212/502 using adaptive policy

Abstract

A method is provided for collaborative caching between a server cache (104) of a server computer (102) and an array cache (112) of a storage array (110) coupled to the server computer. The method includes collecting instrumentation data on the server cache and the array cache of the storage array and, based on the instrumentation data, adjusting the operation of at least one of the server cache and the array cache.

Claims

1. An method (200) for collaborative caching between a server cache (104) of a server computer (102) and an array cache (112) of a storage array (110) coupled to the server computer, comprising:

collecting instrumentation data on the server cache and the array cache of the storage array; and

based on the instrumentation data, adjusting the operation of at least one of the server cache and the array cache.

2. The method of claim 1, wherein the instrumentation data include server cache page hits, array cache page hits, and server resource utilization, and adjusting the operation of at least one of the server cache and the array cache occurs in real-time.

3. The method (400) of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises adjusting a distribution of server cache capacity.

4. The method (400) of claim 3, wherein adjusting the distribution of server cache capacity comprises:

for each page in either the server cache or the array cache, summing cache hits for the cache page over a time period of interest to determine an aggregate hit rate for the cache page;

sorting pages in a list according to their aggregate hit rates;

truncating the list by a total capacity of the server cache;

calculating a ratio of number of cache pages in the truncated list per volume;

comparing per-volume cache sizes in the server cache to the calculated ratio; and

adjusting the per-volume cache sizes to align with the calculated ratio when they are different.

5. The method of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises causing the server cache to evict cache pages based on time the cache pages spent in the server cache and the array cache.

6. The method (500) of claim 5, wherein evicting the cache pages based on time the cache pages spent in the server cache and the array cache comprises:

for each cache hit of an array cache page in the array cache, initializing a page eviction metric for a newly allocated server cache page in the server cache that corresponds to the array cache page, the page eviction metric representing a number of cache hits of the array cache page for a time period of interest;

when the server cache page is about to be evicted from an eviction list based on a least recently used (LRU) policy, determining if the page eviction metric is greater than a threshold;

when the page eviction metric is greater than the threshold, decrementing the page eviction metric and select a next server cache page on the eviction list as a candidate to be evicted; and

when the page eviction metric is not greater than the threshold, evicting the server cache page.

7. The method (500) of claim 5, wherein evicting the cache pages based on time the cache pages spent in the server cache and the array cache comprises:

for each cache hit of an array cache page in the array cache, initializing a page eviction metric for a newly allocated server cache page in the server cache that corresponds to the array cache page, the page eviction metric being set equal to a number of cache hits of the array cache page for a time period of interest;

when the server cache page is about to be evicted from an eviction list based on a least recently used (LRU) policy, determining if the page eviction metric is greater than a threshold;

when the page eviction metric is greater than the threshold, moving the server cache page to a head of an alternate list and select another eviction candidate from the eviction list; and

when the page eviction metric is not greater than the threshold;

decrementing a last server cache page on the alternate;

when the last server cache page is greater than the threshold, move the last server cache page to the head of the alternative list; and

when the last server cache page is not greater than the threshold, move the last server cache page back to the eviction list.

8. The method (600) of claim 5, wherein evicting the cache pages based on time the cache pages spend in the server cache and the array cache comprises:

for each cache page in the server cache, summing hit counts of the cache page in the server cache and the array cache to determine a running hit rate average of the cache pages; and

evicting cache pages from the server cache using a least frequency used (LFU) policy based on running hit rate averages of cache pages in the server cache.

9. The method (700) of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises:

detecting high server processor utilization;

determining a volume on the server computer with low server cache hit rates; and

turning off server caching for the volume.

10. The method (800) of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises:

detecting a process relocating from the server computer to an other server computer coupled to the storage array;

determining data needed by the relocated process; and

filling at least one of an other server cache of the other server computer and the array cache with the needed data.

11. The method (900) of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises:

for each cache read of blocks in the array cache, determining if the array cache has sequential blocks contiguous with the just read blocks; and

when the array cache has the sequential blocks contiguous with the just read blocks, reading the sequential blocks from the array cache into the server cache.

12. The method of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises:

determining a workload indicating that evicted cache pages from the server cache should be retained in the array cache; and

causing the server cache to send a cache eviction message for each evicted cache page to the array cache, wherein the array cache caches the evicted page.

13. The method of claim 1, wherein adjusting the operation of at least one of the server cache and the array cache comprises:

detecting data cached in the server cache is frequently invalidated by writes from other server computers; and

configuring the server cache not to retain any data the next time it is read.

14. A non-transitory computer readable medium encoded with executable instructions for execution by a processor to:

collect instrumentation data on a server cache (104) of a server computer (102) and an array cache (112) of a storage array (110); and

based on the instrumentation data, adjust the operation of at least one of the server cache and the array cache.

15. A computer system (100), comprising:

a server computer (102) comprising a server cache (104);

a storage array (110) comprising an array cache (112), the storage array being coupled to the server computer; and

an analytics system (108) to;

collect instrumentation data on the server cache and the array cache of the storage array; and

based on the instrumentation data, adjust the operation of at least one of the server cache and the array cache.