Last updated on NOVEMBER 29, 2016
Applies to:Oracle FS1-2 Flash Storage System - Version 6.2 to All Versions [Release 6.2 to All Releases]
Information in this document applies to any platform.
What is access skew ?
LUNs and Storage Domains are constructed from a collection of data blocks. Data access occurs across these data blocks and is characterized by a particular statistical pattern.
The degree of asymmetry in the statistical distribution of the block accesses for a LUN is called access skew.
The skew value for a LUN helps to determine the effectiveness of QOS plus on that particular LUN.
Volumes that have a large number of access requests that are skewed to a relatively small portion of the user data benefit the most from the features of QoS Plus.
Access skew range :
The FS1 calculates access skew in a range of 0 to 100 for comparative purposes.
When all the blocks in all the LUNs are accessed uniformly, auto-tiering has no hot data to place on the high performance storage, and no cold data to place on the low-cost storage.
Pure random access is an example of zero skew because all data blocks are accessed equally.
Pure sequential access is another example of zero skew as each data block is accessed only once.
An access skew of 100 indicates a limited number of data blocks accounting for all the data accesses.
This situation can be characterized as an unfair data access pattern, because a limited number of data blocks account for virtually all the accesses.
Auto-tiering is at its best when access patterns are unfair resulting in a skew value that is rewarded with the highest performing storage class.
How the skew value is determined :
The Oracle FS System monitors,collects and analyzes the data activity statistics of the LUN to determine if the initial properties make the best use of the available storage resources.
For a LUN that is using QoS Plus, the Oracle FS System gathers and stores statistics about the access requests as I/O is performed on the LUN.
When optimizing for system performance, QoS Plus moves frequently accessed data blocks of a LUN to a faster, more expensive Storage Class, such as highperformance solid-state drives (SSDs).
Conversely, when optimizing for cost, QoS Plus moves infrequently accessed data blocks to a slower, less expensive Storage Class, such as high-capacity hard disk drives (HDDs).
Sign In with your My Oracle Support account
Don't have a My Oracle Support account? Click to get started
Million Knowledge Articles and hundreds of Community platforms