|
|
|
The CA Datacom MUF can use multiple TCBs in parallel to do user request processing. This multiple TCB ability allows sites that have a significant MUF workload to spread that workload across multiple available processors or engines, that is, GP CPUs.
In standard SMP mode, the MUF automatically has one main task TCB where MUF internal processing occurs. This main task TCB is also used to process user requests. In addition to the main task TCB, you can specify a number of additional TCBs that can be started when needed. In this SMP environment, when additional work is available to be processed and the main task TCB is already busy with work, the additional TCB is started as needed to process the latent demand.
Business Value
The user site with heavy MUF workloads can use this feature to spread the MUF workload across multiple TCBs which are able to use multiple processors or engines to complete the workload. While this mode does not typically reduce the total GP CPU consumed, it improves the request processing elapsed time and users’ response from the MUF.
The CA Datacom SMP implementation does not allow a specific user request to use more than one TCB. By having multiple TCBs available, a poorly constructed query that is keeping a TCB busy, does not stop other better performance requests from processing on the other TCBs. This protects sites from seeing major performance degradation when these heavy queries are processing.
Additional Considerations
Sites with minimal MUF workloads, should not use SMP processing because it incurs additional overhead required for parallel task locking that is not going to be useful unless multiple task processing is occurring. Also sites having only one available processor or engine would not typically benefit from SMP processing.
SMP processing is implemented in the MUF by specifying the startup option SMPTASK.
SMPTASK maxtcb,currenttcb,readytorun
Sets the high-water mark for additional TCBs that can be added to do request processing.
Sets the current number of additional TCBs that can be added to do request processing when the amount of demand is present. The MUF adds TCBs as the workload backlog increases until it has reached the value set in currenttcb.
Sets the amount of workload backload that is needed before an additional TCB is started. A workload backload is basically a request that is ready to run (that is, ready to be processed) that the MUF has not processed because it is busy on another request.
For example, a readytorun of 4, waits until there are four requests that are ready to run and are not currently being processed before it starts the additional TCB (TCB2). The workload backlog would have to grow to 8 readytorun requests before another TCB is started (TCB3). This continues until the number of additional TCBs equal the value set in currenttcb. Once the workload backlog drops back to zero, the additional TCBs are freed. If the workload backload redevelops again, the TCB add-on process repeats.
The user can dynamically adjust the currenttcb number and the readytorun number using the MUF console commands SMPTASK and SMPTASKR.
For more information about using SMP processing, see the CA Datacom/DB Database and System Administration Guide.
| Copyright © 2012 CA. All rights reserved. | Tell Technical Publications how we can improve this information |