METHOD AND SYSTEM FOR INSERTING CACHE BLOCKS

• US 20120054445A1
• Filed: 01/14/2011
• Published: 03/01/2012
• Est. Priority Date: 08/31/2010
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Swart, Garret Frederick, Vengerov, David
• Granted Patent: US 8,601,217 B2
• Time in Patent Office: Days
• US Class Current: 711/133
• CPC Class Codes: G06F 12/0871 Allocation or management of cache space
• CPC Class Codes: G06F 12/0888 using selective caching, e.g. bypass
• CPC Class Codes: G06F 12/121 using replacement algorithms
• CPC Class Codes: G06F 12/123 with age lists, e.g. queue, most recently used [MRU] list or least recently used [LRU] list
• CPC Class Codes: G06F 2212/461 Sector or disk block
• CPC Class Codes: G06F 2212/465 Structured object, e.g. database record

Abstract

A method of inserting cache blocks into a cache queue includes detecting a first cache miss for the cache queue, identifying a storage block receiving an access in response to the cache miss, calculating a first estimated cache miss cost for a first storage container that includes the storage block, calculating an insertion probability for the first storage container based on a mathematical formula of the first estimated cache miss cost, randomly selecting an insertion probability number from a uniform distribution, and inserting, in response to the insertion probability exceeding the insertion probability number, a new cache block corresponding to the storage block into the cache queue.

Claims

1. A method of inserting cache blocks into a cache queue, comprising:

detecting, by a processor, a first cache miss for the cache queue;

identifying, by the processor, a storage block receiving an access in response to the cache miss;

calculating, by the processor, a first estimated cache miss cost for a first storage container comprising the storage block;

calculating, by the processor, an insertion probability for the first storage container based on a mathematical formula of the first estimated cache miss cost;

randomly selecting an insertion probability number from a uniform distribution, wherein the insertion probability exceeds the insertion probability number; and

inserting, in response to the insertion probability exceeding the insertion probability number, a new cache block corresponding to the storage block into the cache queue.

2. The method of claim 1, wherein calculating the first estimated cache miss cost comprises:

calculating an average number of cache hits received by a plurality of cache blocks corresponding to the first storage container in a first passage through a probationary segment of the cache queue;

identifying a latency of the first storage container; and

calculating, based on the average number of cache hits and the latency, the first estimated cache miss cost.

3. The method of claim 1, further comprising:

calculating a plurality of estimated cache miss costs for a plurality of storage containers having cache blocks in the cache queue, wherein the plurality of estimated cache miss costs comprises the first estimated cache miss cost and the plurality of storage containers comprises the first storage container; and

identifying a maximum estimated cache miss cost of the plurality of estimated cache miss costs corresponding to a second storage container of the plurality of storage containers, wherein the insertion probability is further based on the maximum estimated cache miss cost.

4. The method of claim 3, further comprising:

identifying a first arrival rate of a plurality of new cache blocks for the first storage container, wherein the insertion probability is further based on the first arrival rate; and

identifying a second arrival rate of a plurality of new cache blocks corresponding to the second storage container, wherein the first arrival rate exceeds the second arrival rate.

5. The method of claim 4, wherein the new cache block is inserted at a beginning of a probationary segment of the cache queue, and wherein the probationary segment comprises a contiguous set of cache blocks at an end of the cache queue.

6. The method of claim 1, further comprising:

detecting, after detecting the first cache miss and based on an access rate of an old storage container, a workload change for the old storage container;

identifying, within a victim segment of the cache queue, an old cache block corresponding to the old storage container having a latest access time which precedes the workload change; and

evicting, prior to inserting the new cache block and based on the workload change, the old cache block from the cache queue.

7. The method of claim 1, further comprising:

identifying, within the cache queue, a previous new cache block storing a value of a previous storage block;

calculating, by the processor, a second estimated cache miss cost for a second storage container comprising the previous storage block;

calculating, based on a mathematical formula of the second estimated cache miss cost, a removal probability for the second storage container;

randomly selecting a removal probability number from the uniform distribution, wherein the removal probability exceeds the removal probability number; and

evicting, in response to the removal probability exceeding the removal probability number and prior to inserting the new cache block into the cache queue, the previous new cache block from the cache queue.

8. The method of claim 1, wherein the mathematical formula represents the insertion probability as an increasing function of the first estimated cache miss cost.

9. A method of inserting cache blocks into a cache queue, comprising:

calculating an estimated cache miss cost for a cache block within the cache queue;

evicting the cache block from the cache queue;

inserting an entry for a storage block corresponding to the cache block into a shadow list corresponding to the cache queue;

detecting, by a processor, a cache miss for the cache queue referencing the storage block;

accessing, in response to the cache miss, the entry within the shadow list;

calculating, based on a mathematical formula of a plurality of estimated old cache miss costs for cache blocks evicted from the cache queue, an estimated cache miss cost threshold; and

inserting, in response to the estimated cache miss cost exceeding the estimated cache miss cost threshold, a new cache block corresponding to the storage block into the cache queue.

10. The method of claim 9, wherein calculating the estimated cache miss cost threshold comprises:

calculating the plurality of estimated old cache miss costs;

identifying a maximum estimated old cache miss cost of the plurality of estimated old cache miss costs;

calculating a standard deviation of the plurality of estimated old cache miss costs; and

calculating, based on the maximum estimated old cache miss cost and the standard deviation, the estimated cache miss cost threshold.

11. The method of claim 9, wherein calculating the estimated cache miss cost for the cache block comprises:

calculating, for the cache block, an estimated access rate based on a number of cache hits received since insertion into the cache queue divided by a time elapsed since insertion into the cache queue; and

calculating the estimated cache miss cost based upon a latency of the cache block multiplied by the estimated access rate.

12. A computer readable storage medium storing a plurality of instructions for inserting cache blocks into a cache queue, the plurality of instructions comprising functionality to:

detect a cache miss for the cache queue;

identify a storage block receiving an access in response to the cache miss;

calculate a first estimated cache miss cost for a first storage container comprising the storage block;

calculate, based on a mathematical formula of the first estimated cache miss cost, an insertion probability for the first storage container;

randomly select an insertion probability number from a uniform distribution, wherein the insertion probability exceeds the insertion probability number; and

insert, in response to the insertion probability exceeding the insertion probability number, a new cache block corresponding to the storage block into the cache queue.

13. The computer readable storage medium of claim 12, wherein calculating the first estimated cache miss cost comprises:

calculating, for the first storage container, an average number of cache hits received by a plurality of cache blocks corresponding to the first storage container in a first passage through a probationary segment of the cache queue;

identifying a latency of the first storage container; and

calculating, based on the average number of cache hits and the latency, the first estimated cache miss cost.

14. The computer readable storage medium of claim 12, wherein the plurality of instructions further comprise functionality to:

calculate a plurality of estimated cache miss costs for a plurality of storage containers having cache blocks in the cache queue, wherein the plurality of estimated cache miss costs comprises the first estimated cache miss cost and the plurality of storage containers comprises the first storage container; and

identify a maximum estimated cache miss cost of the plurality of estimated cache miss costs, wherein the insertion probability is further based on the maximum estimated cache miss cost.

15. The computer readable storage medium of claim 12, wherein the plurality of instructions further comprise functionality to:

identify, within the cache queue, a previous new cache block storing a value of a previous storage block;

calculate a second estimated cache miss cost for a second storage container comprising the previous storage block;

calculate, based on a mathematical formula of the second estimated cache miss cost, a removal probability for the second storage container;

randomly select a removal probability number from the uniform distribution, wherein the removal probability exceeds the removal probability number; and

evict, in response to the removal probability exceeding the removal probability number and prior to inserting the new cache block into the cache queue, the previous new cache block from the cache queue.

16. The computer readable storage medium of claim 12, wherein the plurality of instructions further comprise functionality to:

calculate an estimated arrival rate of new cache blocks for the first storage container, wherein the insertion probability is further based on the estimated arrival rate.

17. A system for inserting cache blocks, comprising:

a cache queue, comprising;

a probationary segment at an end of the cache queue, and

a protected segment adjacent to the probationary segment; and

a cache manager executing on a processor and configured to;

detect a cache miss for the cache queue;

identify a storage block receiving an access in response to the cache miss;

calculate an estimated cache miss cost for a storage container comprising the storage block;

calculate, based on a mathematical formula of the estimated cache miss cost, an insertion probability for the storage container;

randomly select a probability number from a uniform distribution, wherein the insertion probability exceeds the probability number; and

insert, in response to the insertion probability exceeding the probability number, a new cache block corresponding to the storage block into the cache queue at a beginning of the probationary segment.

18. The system of claim 17, further comprising:

a shadow list corresponding to the cache queue, comprising;

a plurality of entries for evicted cache blocks;

wherein the cache manager is further configured to;

evict, prior to inserting the new cache block, a victim cache block from the cache queue; and

insert an entry for the victim cache block into the shadow list.

19. The system of claim 17, further comprising:

a storage device comprising the storage container, wherein the storage container comprises a plurality of storage blocks, and wherein the plurality of storage blocks comprises the storage block.

20. The system of claim 17, wherein calculating the estimated cache miss cost comprises:

calculating an average number of cache hits received by a plurality of cache blocks corresponding to the first storage container in a first passage through a probationary segment of the cache queue;

identifying a latency of the first storage container; and

calculating, based on the average number of cache hits and the latency, the first estimated cache miss cost.