Factors that affect remote journal performance

 

There are two main performance objectives for the remote journal function. To provide a timely delivery of journal entries to a target system and to minimize impacts to the journaling throughput on the source system.

Even though both aspects are very important for both synchronous and asynchronous delivery modes, each mode prioritizes the two in a different order. The top priority for synchronous delivery is to guarantee that the remote journal is always up to date with the source journal. For asynchronous delivery mode, the top priority is to minimize impacts to journaling throughput.

All performance considerations that are currently used for a local journal still apply and must continue to be employed. The following are additional factors that may affect the performance of the remote journal function. The factors are listed in the order of importance.

1. Transport method

   Your choice of transport depends on the rate of the journal activity in your environment. Make special consideration for using a fast transport method when you use synchronous delivery mode. Weigh the response time impacts of the synchronous delivery mode in your environment against the communications overhead of the transport method you choose.

   When replicating journal entries over a long distance, the most important performance factors regarding a communications transport method are the overall rated speed of the communications resource and any existing traffic already using the communications resource.

   For more information about transport methods, see the Networking topic.

2. Number of remote journals that are being maintained

   With respect to the job performing the journal entry deposit, the impact of the remote journal function for asynchronously maintained journals is not noticeable. For synchronously maintained journals, the impact depends on the slowest connection rather than number of remote journals.

   The impact to the job performing the journal entry deposit for an asynchronously maintained journal is significantly less than that for a synchronously maintained journal. Also, it is recommended that only one synchronous remote journal be maintained for a given local journal.

   With respect to the system performance impacts, the processor use typically increases by less than an equal factor for each additional remote journal.

3. Arrival rate of journal entries that are being deposited on the local system

   The higher the arrival rate of journal entries being deposited on the local system, the greater the chance journaling throughput will increase for synchronous or asynchronous delivery. A high arrival rate might cause asynchronous journaling to fall further behind.

4. Batch versus interactive

   In general, higher local and remote journal throughput can be maintained when many interactive jobs generate the journal throughput rather than a single-threaded batch job. Journal caching can also increase this throughput for batch processing regardless of the number of jobs.

5. Processor utilization on the source system

   The higher the processor utilization of the source system, the greater the chance of affecting journaling throughput for synchronous or asynchronous delivery. This may cause asynchronous journaling to fall further behind.

6. Processor utilization on the target system

   The higher the processor utilization of the target system, the greater the chance of affecting journaling throughput for synchronous or asynchronous delivery. This may cause asynchronous journaling to fall further behind.

7. The value set for the sending task priority when using the asynchronous delivery mode

   The larger the value, the smaller effect the remote journal function will have on the system, but the further the target system may lag behind the source system.

 

Performance considerations regarding the catch-up phase

Performance considerations regarding the catch-up phase when activating the remote journal function include the following in order of importance:

1. Total number of bytes for all of the journal entries that need to be caught up

   The larger the total size, the longer the catch-up phase will run.

2. Transport method

   Select a transport method that is appropriate for your remote journaling environment.

3. Disk protection on the target system

   At high data transfer rates, disk units with device parity protection in the ASP on the target system can limit the performance of the catch-up phase, unless the target system has sufficient write cache configured in the I/O adaptors servicing the disk units that house the journal receiver. One example of this is when you use the OptiConnect for i5/OS^® bus transport method. Having mirrored or unprotected disk units in the ASP on the target system would eliminate this effect.

4. Processor utilization on the source system

   The higher the processor utilization of the source system, the greater the chance of affecting the performance for the catch-up phase.

5. Processor utilization on the target system

   The higher the processor utilization of the target system, the greater the chance of affecting the performance for the catch-up phase.

6. Delivery mode

   The performance of the catch-up phase does not depend on the delivery mode that was specified, synchronous or asynchronous.

The catch-up processing that is performed by the remote journal function is the most efficient method of replicating the journal entries with the remote journal function.

 

How the journal attributes affect the remote journal performance

Reducing the size of the journal receivers on the source system will reduce the communications overhead of the remote journal function. Therefore, you may want to consider journaling *AFTER images and not journaling open, close, or force entries.

Some of the most common attributes you may want to use for auditing journal entries are the following:

• Maximum receiver size - RCVSIZOPT(*MAXOPT1, *MAXOPT2, or *MAXOPT3)

• Remove internal entries - RCVSIZOPT(*RMVINTENT)

• Minimized entry specific data - MINENTDTA(*FILE) or *FLDBDY

Attributes such as the FIXLENDTA can also cause minimal performance improvements.

Refer to the Remote journal attributes and Remote journals and auxiliary storage links below for more details about remote journal performance.

 

Parent topic:

Planning for remote journals

Related concepts
Journal management and system performance Methods to reduce the storage that journal receivers use Remote journal attributes Remote journals and auxiliary storage

Related information
Networking AS/400 Remote Journal Function for High Availability and Data Replication