This topic provides an overview about dependent writes.
Think about the following typical sequence of write operations for a data base update transaction.
The database ensures correct ordering of these writes by waiting for each step to complete before starting the next. If, however, the database log (updates 1 and 3) and the database itself (update 2) are on different virtual disks and a FlashCopy mapping is started during this update, then the possibility that the database itself is copied slightly before the database log resulting in the target virtual disks seeing writes (1) and (3) but not (2) must be excluded. In this case, if the database were restarted from a backup made from the FlashCopy target disks, the data base log would indicate that the transaction had completed successfully when, in fact, that is not the case. The transaction would be lost and the integrity of the data base would be in question.
It may thus be the case that in order to create a consistent image of user data it is necessary to perform a flash copy operation on multiple virtual disks as an atomic operation. In order to meet this need, the SAN Volume Controller supports the concept of a consistency group. A consistency group contains a number of FlashCopy mappings. A consistency group can contain an arbitrary number of FlashCopy mappings up to the maximum number of FlashCopy mappings supported by a SAN Volume Controller cluster. The SAN Volume Controller allows the start command which causes the point in time copy to occur, to be directed at a consistency group. In this case all of the FlashCopy mappings in the consistency group are started at the same time, resulting in a point in time copy which is consistent across all of the FlashCopy mappings which are contained in the consistency group. The SAN Volume Controller supports 128 consistency groups per cluster.