You can assign priority levels to Teradata Database system recovery or table rebuild by using the PRIORITY command feature of rcvmanager. Control over job priority is assigned by selecting one of three priority levels:
For information about these priority levels, see REBUILD/RECOVERY PRIORITY.
Teradata Database system recovery and disk rebuild are primarily I/O intensive tasks. The major portion of the time taken by a given task is involved in setting up for or waiting for completions of a disk or a BYNET message traffic operation. Very little manipulation or computation is required on the data once it is available.
If the competing work load of the Teradata Database system can be reasonably characterized, then you can gauge the impact of the priority changes to recovery and rebuild operations. In general, use the following guidelines when assigning priorities:
|No work load competition||If there is no competition for resources as in a COLDWAIT restart for recovery, the priority setting of the recovery and rebuild jobs will have no practical effect.|
|Compute-intensive work load||If the online Teradata Database system work load is heavily compute intensive, raising the priority of the I/O intensive recovery operations can dramatically improve the recovery (including rebuild).
The high recovery will have a relatively minor impact on the online Teradata Database system operations. However, it will provide a better resource utilization and result in better Teradata Database system throughput. Similarly, a low priority of recovery in this work load will dramatically slow down recovery with only a moderate gain in online throughput.
|Moderate or heavy disk work load||If a moderate or heavy amount of disk and/or BYNET usage occurs by the online system, then recovery will show moderate throughput changes by controlling the priority setting but with a larger impact against the Teradata Database system throughput. Memory contention becomes a major component of operation in these cases.|
|I/O saturation||As the I/O utilization approaches saturation, there are fewer opportunities to improve throughput or execution time of either the online Teradata Database system or the recovery job. In this case, we are competing for the same resource and that resource is not amenable to manipulation by control over the scheduling priority.|