IBM Atlas Extensions 6.0.3.2 Performance and Scalability

Content

Many of the IBM Atlas Policy Suite features depend on a replication of the company’s business hierarchy, including lists of employees, department organization, and matter definitions. In most companies, this data is represented in an existing database. Atlas Extensions provides a way to automatically import people, organization, and matter information from an external source. It is without the use of an intermediate storage medium, with changes to the original source automatically propagated to Atlas. This paper presents the scalability characteristics and performance tuning guidelines for the Atlas Extensions 6.0.3.2 HRLoader and OULoader on WebSphere Application Server and DB2.
• HRLoader tests demonstrate linear scalability as the number of imported people increases. The complexity of the organization tree will affect the loading performance. Optimal performance can be found by varying the number of HRLoader threads.
• OULoader tests show that performance is affected by the complexity of the organization tree structure. OULoader can use incremental mode to upload delta changes to an organization structure as it changes over time, and demonstrates linear scalability as the amount of updated data increases.
• HRLoader can finish importing 1 million people in less than 4 hours. With best tuning on the lab's environment, it took less than 14 minutes to import 100 thousand people. OULoader can finish importing 100 thousand people with depth tree structure around 1 hour.

The performance results that are reported in this paper represent data and workloads that are run on an isolated network on specific operating environments and system configurations. Actual performance in real customer environments with production workloads might vary depending on many factors such as system configurations, workload characteristics, and data volume. The presented results here are not guaranteed to be repeatable in other systems.