Input/output patterns and data pre-fetch
View Patent ↗Systems and methods for determining an access pattern in a computing system. Accesses to a file may contain random accesses and sequential accesses. The file may be divided into multiple regions and the accesses to each region are tracked. The access pattern for each region can then be determined independently of the access patterns of other regions of the file.
1. A method for determining an access pattern associated with accesses to a file stored in a computing system, the method comprising:
dividing the file into regions, wherein each region is associated with accesses;
tracking the accesses for each region in a context history, wherein the context history stores a history of the accesses for each of the regions;
determining an access pattern for each region based at least on each respective history;
implementing a caching strategy based on the determined access pattern of each region;
adding data to a cache when the access pattern is sequential, wherein the data added is data that is expected to be the subject of subsequent accesses; and
evicting data from the cache when the access pattern is random.
2. The method of claim 1 , further comprising determining a size of the regions.
3. The method of claim 1 , further comprising determining the access pattern for a particular region after each access to the particular region.
4. The method of claim 1 , wherein the context history includes an entry for each region of the file, wherein each entry in the context history identifies one or more of:
a region associated with the entry,
a data sequential threshold,
an access sequential threshold,
data sequentiality,
access sequentiality,
an index to current IO, and
a list of recorded IOs.
5. The method of claim 4 , wherein the access pattern is sequential when at least half or a majority of the recorded IOs have data sequentiality.
6. The method of claim 4 , wherein the access pattern is random when less than half of the recorded IOs have data sequentiality.
7. The method of claim 4 , wherein the recorded IOs are stored in a circular buffer such that the access pattern is determined based on a recent number of IOs.
8. The method of claim 4 , wherein an access has data sequentiality when an offset of the current IO is sequential to an offset of at least one other access in the context history.
9. The method of claim 8 , wherein the access has data sequentality when the offset is within a threshold number of bytes to at least one other access in the context history.
10. The method of claim 1 , further comprising evaluating the detected access pattern as additional accesses are recorded in the context history.
11. The method of claim 1 , wherein the access pattern is determined to be one of none, random, increasing, or sequential.
12. The method of claim 1 , wherein the regions include a first region and a second region, wherein the access pattern for the first region is determined to be random and the access pattern for the second region is determined to be sequential.
13. The method of claim 1 , further comprising determining that the access pattern for at least one of the regions changes from sequential to random or from random to sequential.
14. The method of claim 13 , wherein the access pattern for each region is re-evaluated after every access.
15. The method of claim 1 , further comprising determining that the file has a pre-determined access pattern that is based on an IP address, an administrator label, or application type.
16. The method of claim 15 , wherein the predetermined access pattern is subject to change and is re-evaluated as the accesses are tracked.
17. The method of claim 1 , wherein a caching strategy of pre-fetching data or metadata or evicting data or metadata changes in accordance with the currently determined access pattern for each of the regions.
18. A method for determining an access pattern associated with accesses to a file stored in a computing system, the method comprising:
dividing the file into regions, wherein each region is associated with accesses;
tracking the accesses for each region in a context history, wherein the context history stores a history of the accesses for each of the regions;
determining an access pattern for each region based at least on each respective history;
adding data to a cache when the access pattern is sequential, wherein the data added is data that is expected to be the subject of subsequent accesses; and
implementing a caching strategy of pre-fetching data or metadata or evicting data or metadata changes in accordance with the currently determined access pattern for each of the regions.