Background copy FlashCopy FlashCopy applications

This topic provides an overview about background copy.

A FlashCopy mapping has a property background copy rate. This is a value between 1 and 100. The background copy rate can be changed when the FlashCopy mapping is in any state.

If "NOCOPY" is specified, then background copy is disabled. One use of this is for short-lived FlashCopy mappings which are to be used for backup purposes only. Since the source data set is not expected to change much during the lifetime of the FlashCopy mapping, it is more efficient in terms of managed disk I/Os not to perform a background copy.

The relationship of the background copy rate value to the attempted number of grains to be split per second is given by the following table. (A grain is the unit of data represented by a single bit, which is 256K.)

Table 3. Background copy
User-specified value KB/sec Grains/sec
1 - 10 128 0.5
11 - 20 256 1
21 - 30 512 2
41 - 50 2048 8
91 - 100 64 MB 256

The grains/sec numbers represent goals that the code tries to achieve. The SAN Volume Controller will be unable to achieve these goals if insufficient bandwidth is available from the SAN Volume Controller nodes to the physical disks making up the managed disks after taking into account the requirements of foreground I/O. If this situation arises, then background copy I/O will contend for resources on an equal basis with I/O arriving from hosts. Both will tend to see an increase in latency and consequential reduction in throughput with respect to the situation had the bandwidth not been limited.

Degradation will be graceful. Both background copy and foreground I/O will continue to make forward progress, and will not stop, hang or cause the node to fail.

The background copy is performed by one of the nodes belonging to the I/O group in which the source virtual disk resides. This responsibility is failed over to the other node in the I/O group in the event of the failure of the node performing the background copy.

The background copy is performed backward; that is, it starts with the grain containing the highest logical block numbers (LBAs) and works backwards towards the grain containing LBA 0. This is done to avoid any unwanted interactions with sequential write streams from the using application.

(C) Copyright International Business Machines Corporation 2003, 2004