CA Datacom Version 15.0 provides Database Administrators with faster restart processing. Faster restart processing is achieved by using parallel processing that incorporates multiple tasks. The user tasks provided are used (up to the first 500). All except the last are used for maintenance command reprocessing. The last is reserved for any required OPEN commands.
Taking advantage of faster restart processing in CA Datacom/DB Version 15.0 is transparent to users. Parallel processing functionality requires no Multi-User Facility (MUF) startup option specifications.
If a MUF fails or is cancelled, a wide variety of data changes that were in-flight can be left unresolved. The MUF restart process is automatically initiated when the MUF is started. If the MUF had an abnormal end, the restart process detects it. In previous releases, when MUF had an abnormal end, the restart process ran in a serial order. In-flight transactions were reprocessed in a time-sequence order to return the database to stability. In Version 15.0, parallel processing techniques that use multiple tasks complete restart processing in less time.
During MUF restart, incomplete requests are reprocessed. The tasks that are used in parallel are provided by the first parameter of the MUF startup option TASKS. They are used and reused as the restart process sees fit to minimize elapsed time during the reprocessing.
Reprocessing is done in the main task TCB and in any of the additionally allocated TCBs or SRBs which were specified with the MUF startup option SMPTASKS.
The parallel processing in the provided task areas occurs with two infrequent exceptions. Required openings of data sets occur sequentially in the last task area. Maintenance to the same row that is already in process in a task area waits on its completion before starting.
In Version 15.0, a new subtask is used for controlling and driving the main TCB. The new subtask is named DBMFSPR. The activity of the DBMFSPR subtask is reflected in the MUF EOJ report section for all TCBs and SRBs.
The primary reason for faster restart processing is to support 24x7 environments. If there is a MUF outage, faster restart processing means that MUF is more quickly enabled. Outages are rare, and the savings of faster restart processing are important. In addition, however, faster reprocessing means that users can increase the value of the MUF startup option LOGPEND. The LOGPEND option defines the size of the pipeline. The size of the pipeline significantly impacts MUF performance. Increasing the LOGPEND value improves the performance of MUF in general with less risk of time expenditures for reprocessing due to an outage.
In Version 15.0, the restart report includes two new sections that report parallel processing. The first new section contains aggregate statistics that are related to reprocessing. The second new section contains task-level statistics that are related to reprocessing.
|
Copyright © 2014 CA.
All rights reserved.
|
|