DRV increases the availability of information stored on disk by recording the data on 2 physical disk drives. The duplication takes place in the IO system of BS2000. For a synchronous write operation, first of all an asynchronous output to the first disk is initiated, then a synchronous output to the second disk drive.
The disk drive with the shorter queue is selected when a read operation is to be performed, thus resulting in an unsymmetrical load on the original disk and mirror disk. Selecting the least heavily used disk drive reduces the I/O time (software duration) for read accesses in comparison to SRV (Single Recording by Volume) mode.
Approximately the same performance behavior can be expected as for SRV mode.
Copy speed
The copy speed can be set dynamically in three steps for each pubset using the operand COPY-SPEED=*LOW/*MEDIUM/*HIGH of /SET-DRV-PARAMETER at both restoration and equalization.
With COPY-SPEED=*LOW, the DRV copy operations are slowed down considerably so that the pubset for the I/O operations of applications is more accessible. The default is *MEDIUM.
Extent of copying
Not all the disk's blocks are copied at restoration; only the blocks identified as occupied in the F5 label are copied.
After a system failure or an abnormally terminated export of a pubset, DRV carries out equalization in order to match the disk pairs again.
In addition to the existing variants of copying either all the blocks or only the system data, DRV also allows only the system data and the open files to be equalized. This option is only possible for SF pubsets or volume sets of an SM pubset. It can only be set by the following command before the pubset or volume set is imported:
/SET-DRV-PARAMETER UNIT=*PUBSET(...)/*VOLUME-SET(...), EQUALIZE-DATA= *OPEN-FILES
Rapid restoration after separate processing
In the case of separate processing (e.g. backup of the original data, batch processing of the mirror data), DRV remembers the changed areas. This allows the restoration time to be dramatically reduced when only the changes since the beginning of separate processing have to be copied (since this can be measured in terms of Mbytes) rather than all the files (Gbytes of data).