Performance planning
Ensure that you plan before you install IBM® InfoSphere® Master Data Management Collaboration Server - Collaborative Edition. Your planning can greatly affect the performance of InfoSphere MDM Collaboration Server - Collaborative Edition.
Adhere to the following common guidelines so that you avoid
major performance problems:
- Testing, profiling and modifying the solution
- For each line item, ensure that you set aside 20% extra time to complete testing, profiling and modifying the solution as needed. Include the 20% extra time before there are any known performance problems.
- Define all of your use cases
- The use cases are performance sensitive. Ensure that you identify the requirements, dependencies, and required performance. Allow extra time for use cases that have a high potential for performance problems. For example, use cases that have a high potential for performance problems can include many specs, large amount of location data or large numbers of workflow steps. Test and profile the use cases as they are developed or as early as possible if there are any other dependencies. Ensure that you do not delay performance testing until the end of the project. You need to establish a baseline for the use cases and have them approved by the customer.
- Identify the hardware that is needed for testing
- For hardware, ensure that you identify the hardware that is needed for performance testing and have it available early in the project. The hardware for performance testing should be a replica of the hardware that is planned for production. Performance testing and user acceptance testing should always be done on hardware that is identical to production.
- Allocate the size of the hardware
- Allocating the correct hardware is critical to sustain performance
of the solution. The correct size of the hardware that is required
to effectively run the final solution depends on:
- The volume of the activity on the system
- The overall complexity of the solution
- Allocating the correct hardware is critical to sustain performance
of the solution. The correct size of the hardware that is required
to effectively run the final solution depends on:
- Tune the allocated hardware
- Correctly sized hardware is only effective when it is properly
tuned. The following two key areas that commonly appear as the cause
of performance problems are:
- Latency and bandwidth between the application server and database
- The latency should be under 0.25 ms between the application server and the database. It can be measured by running the traceroute command on most systems. The connection between the two should be a gigabit Ethernet capable of transferring large files at 25 MB/s through FTP.
- Number of open descriptors is too low
- Unexpected problems can be avoided by checking the number of open descriptors and verifying that they are set according to WebSphere® Application Server guidelines to 8000. The number of open descriptors can be checked using the ulimit –a command on most computers.
- Correctly sized hardware is only effective when it is properly
tuned. The following two key areas that commonly appear as the cause
of performance problems are:
- Balance the load and allow for failover
- An easy way to address potential overloading of the application server is to use a load balancer. Multiple instances of the scheduler can be started on one or more servers and various services of the scheduler load balance themselves automatically.