Configuring persistent connections

HTTP persistent connection or HTTP keepalive is a TCP feature that is used to reduce network congestion and CPU usage, in particular on systems that have lower hardware specifications. The feature allows multiple requests over a single connection. Reusing a connection in this way is useful with secure HTTP because the same connection does not need to be renegotiated every time.

The following parameters in the httpd.conf file control persistent connections:

The KeepAliveTimeOut parameter controls the amount of time that the server waits for the next request on a persistent connection. If the value is too small, connections might not be reused efficiently. If the value is too large, web server threads might not be used efficiently. We recommend starting by setting the value to 2 and then adjust it if necessary. It is suggested that you keep this value lower than the average think time of the users of the system.

Recent performance tests highlighted that some disadvantages occur when persistent connections are used with Cúram application servers:

In addition, we found that the WebSphere® Application Server (WAS) setting WebServer Plug-in Properties-Connection Timeout affects the server’s ability to handle heavy loads. When the feature is turned on, issues that arise include huge numbers of objects are created when the server is unable to handle the load, extreme resource utilization occurs on the server, and server failures occur.

If you experience any of the issues when you test your system with heavy loads, we recommend disabling the persistent Connections/keepalive feature.