AT & T case study
AT&T’s (Legacy Bellsouth) Operations Services Department (OSD) manages the DSL Network Management System (NMS) for Southeastern USA. NMS interfaces with a dozen independent applications in provisioning DSL connections to home and trouble shooting customer problems.
Performance of NMS and other applications it interfaces was of concern during peak traffic of the day and during a big outage. Semafor’s performance team worked with AT&T to study the situation on a per project basis to scope the requirements, prepare a strategy, conduct tests and submit detailed analysis reports.
Semafor’s expertise with HP (Mercury) LoadRunner and Diagnostics was of great value in this effort. Defining the requirements involved talking with customers to understand the expectations and analyzing production logs to understand the performance characteristics of the application from production usage. Once the requirements were gathered a detailed requirements document was generated to document the current system usage and also with predictions for future growth.
Based on the requirements document a detailed test strategy was prepared. The strategy document described in depth about the different types of performance tests to be performed, architecture of the test lab, comparison of the test lab with production and scaling of requirements to match test lab.
Mercury LoadRunner was used to prepare scripts simulating users under JAVA RMI, SOAP/XML and HTTP environments. Load, stress and soak testing was conducted using LoadRunner to test the application performance. Inbuilt monitors in LoadRunner along with Mercury Diagnostics for J2EE was used to capture the performance metrics for analysis.
Many applications within NMS was found to exceed allocated heap. A heap snapshot revealed memory leaks within some JVMs and the code was sent to developers for profiling. Semafor also recommended an increase in initial weblogic thread count and JDBC pool size. Also Semafor found the most expensive SQL statements in Oracle and tuned them.
By fixing slower server methods, slower SQL queries and memory problems, Semafor helped AT&T improve overall application performance during peak traffic and also save money that would have been spent in buying additional hardware. |
