Method for using service level objectives to dynamically allocate cache resources among competing workloads

Patent metadata

Field	Value
Publication number	US-9916250-B2
Application number	US-201715437635-A
Country	US
Kind code	B2
Filing date	Feb 21, 2017
Priority date	Oct 23, 2014
Publication date	Mar 13, 2018
Grant date	Mar 13, 2018

Abstract

A method, device, and non-transitory computer readable medium that dynamically allocates cache resources includes monitoring a hit or miss rate of a service level objective for each of a plurality of prior workloads and a performance of each of a plurality of cache storage resources. At least one configuration for the cache storage resources for one or more current workloads is determined based at least on a service level objective for each of the current workloads, the monitored hit or miss rate for each of the plurality of prior workloads and the monitored performance of each of the plurality of cache storage resources. The cache storage resources are dynamically partitioned among each of the current workloads based on the determined configuration.

First claim

What is claimed is:

1. A method comprising:
monitoring, by a computing device, a performance rate for each of a plurality of prior workloads and a latency of each of a plurality of cache storage resources;
determining, by the computing device, a plurality of configurations for the plurality of cache storage resources for one or more current workloads based at least on a service level objective for each of the one or more current workloads, the monitored performance rate for each of the plurality of prior workloads, and the monitored latency of each of the plurality of cache storage resources;
determining, by the computing device, a cost for each of the determined plurality of configurations;
selecting, by the computing device, one of the determined plurality of configurations for the dynamic partitioning based on the determined cost for each of the determined plurality of configurations; and
dynamically partitioning, by the computing device, the plurality of cache storage resources among each of the one or more current workloads based on the selected one of the determined plurality of configurations.

2. The method as set forth in claim 1 wherein the plurality of cache storage resources comprises an in-memory buffer cache and a flash cache.

3. The method as set forth in claim 1 further comprising modeling, by the computing device, the observed performance rate for each of the plurality of prior workloads over a range of sizes of the plurality of cache storage resources, wherein determining the configuration for the plurality of cache storage resources for the one or more current workloads is further based on the modeling.

4. The method as set forth in claim 1 further comprising:
identifying, by the computing device, when the service level objective for at least one of the one or more current workloads is not met by the determined configuration for the plurality of cache storage resources; and
modifying, by the computing device, the determined configuration for the plurality of cache storage resources when the service level objective for at least one of the one or more current workloads is not met.

5. The method as set forth in claim 1 wherein determining the total cost for each of the determined plurality of configurations comprises:
obtaining, by the computing apparatus, a cache resource cost assigned to each of the plurality of cache storage resources; and
determining, by the computing apparatus, a value of a portion of each of the plurality of cache storage resources consumed in each of the determined plurality of configurations based on the obtained cache resource cost assigned to each of the plurality of cache storage resources, wherein the determined total cost for each of the determined plurality of configurations is a sum of the determined values that configuration.

6. The method as set forth in claim 1 wherein determining the total cost for each of the determined plurality of configurations comprises determining an error sensitivity cost for each of the determined plurality of configurations, and wherein selecting one of the determined plurality of configurations for the dynamic partitioning comprises selecting one of the determined plurality of configurations that has a lowest error sensitivity cost for dynamic partitioning.

7. A computing device comprising:
a memory containing a machine readable medium comprising machine executable code having stored thereon instructions for performing a method of dynamically allocating cache storage resources; and
a processor coupled to the memory, the processor configured to execute the machine executable code to cause the processor to:
monitor a performance rate for each of a plurality of prior workloads and a latency of each of a plurality of cache storage resources over consecutive and partially overlapping period of times;
determine a plurality of configurations for the plurality of cache storage resources for one or more current workloads based at least on a service level objective for each of the one or more current workloads, the monitored performance rate for each of the plurality of prior workloads, and the monitored latency of each of the plurality of cache storage resources;
determine a cost for each of the determined plurality of configurations;
select one of the determined plurality of configurations for the dynamic partitioning based on the determined cost for each of the determined plurality of configurations; and
dynamically partition the plurality of cache storage resources among each of the one or more current workloads based on the selected one of the determined plurality of configurations.

8. The device as set forth in claim 7 wherein the plurality of cache storage resources comprises an in-memory buffer cache and a flash cache.

9. The device as set forth in claim 7 wherein the machine executable code further causes the processor to model the performance rate for each of the plurality of prior workloads over a range of sizes of the plurality of cache storage resources, and wherein determining the configuration for the plurality of cache storage resources for the one or more current workloads is further based on the modeling.

10. The device as set forth in claim 7 wherein the machine executable code further causes the processor to:
identify when the service level objective for at least one of the one or more current workloads is not met by the determined configuration for the plurality of cache storage resources; and
modify the determined configuration for the plurality of cache storage resources when the service level objective for at least one of the one or more current workloads is not met.

11. The device as set forth in claim 7 wherein determining the total cost for each of the determined plurality of configurations comprises:
obtaining a cache resource cost assigned to each of the plurality of cache storage resources; and
determining a value of a portion of each of the plurality of cache storage resources consumed in each of the determined plurality of configurations based on the obtained cache resource cost assigned to each of the plurality of cache storage resources, wherein the determined total cost for each of the determined plurality of configurations is a sum of the determined values that configuration.

12. The device as set forth in claim 7 wherein determining the total cost for each of the determined plurality of configurations comprises determining an error sensitivity cost for each of the determined plurality of configurations and wherein selecting one of the determined plurality of configurations for the dynamic partitioning comprises selecting one of the determined plurality of configurations that has a lowest error sensitivity cost for the dynamic partitioning.

13. A non-transitory machine readable medium having stored thereon instructions for performing a method comprising machine executable code which, when executed by at least one machine, causes the at least one machine to:
monitor a performance rate for each of a plurality of prior workloads and a latency of each of a plurality of cache storage resources;
determine a plurality of configurations for the plurality of cache storage resources for one or more current workloads based at least on a service level objective for each of the one or more current workloads, the monitored performance rate for each of the plurality of prior workloads, and the monitored latency of each of the plurality of cache storage resources;
determine a cost for each of the determined plurality of configurations;
select one of the determined plurality of configurations for the dynamic partitioning based on the determi

Assignees

• Netapp Inc

Inventors

• Shah Peter
• Smith Keith

Classifications

• G06F12/0895

  of parts of caches, e.g. directory or tag array · CPC title

• G06F12/0871 Primary

  Allocation or management of cache space · CPC title

• G06F2212/282

  Partitioned cache · CPC title

• G06F2212/222

  Non-volatile memory · CPC title

• G06F2212/1021

  Hit rate improvement · CPC title