Performance optimisations in Cascaded FlashCopy Systems
Publication Date: 2010-Sep-17
The IP.com Prior Art Database
Performance enchancements in a system providing FlashCopy based automatic data tiering.
Performance optimisations in systems implementing 2008-0626 and follow on disclosures
In a system that uses incremental FlashCopy(TM) (point in time) copies, the configuration can be arranged in such a way as to provide a simplistic utilisation model. That is, newly read or written data can be separated from older (stale) data.
Such as system can provide the separation of frequently and less frequently accessed data for the purposes of tiering between different classes of storage. This is especially useful at providing a fine granularity (in the order of tens of KB) of progression of data between tiers over longer periods of time. Many other data progression or automated tiering systems have a much larger granularity of data - usually many MB.
However such a system requires additional storage controller bandwidth, each time a small write or read is performed (say 4KB) then a larger 64K (or 256K) read and write is required to move the data between the targets (tiers). This could result in increased, or overloaded bandwidth within a storage system - or on the storage devices themselves and defeat the purpose of the goal - to provide much faster data access by placing the data on the correct target/tier.
In order to maintain the performance benefits (mainly response time improvements) in a system such as described, the proposed solution is to take into account the current performance characteristics of the system (either theoretical or measured statistically) and make a decision before promoting or demoting data between the tiers (targets).
This could be done by associating a "band...