TXSeries transaction need additional 400 milliseconds suddenly
This document applies only to the following language version(s):
When conducting a TXSeries performance test a sudden increase was encountered in the ECI client response time by an additional 400 milliseconds.
In ECI application client, client encountered an increase in response time which became an issue. Sometimes, the performance recovered to normal automatically, and sometimes it become worse again within in minutes or hours.
In this case, the response time grew by an additional 400 milliseconds when the issue occurred. For example, the normal response time is 20ms, when the issue occurs, the response time will be 420ms.
From the TCPIP trace, it shows network adapter can not send the TCPIP message as soon as possible. Sometimes it will hold the message up to 200ms before send it out. If both the message send and the message receive encounter the issue, the transaction will be more than 400 milliseconds( 200+200ms).
The structure is CTG->Front End Region->Back End Region->database. Or the simple structure is CTG->Region.
Resolving the problem
If TCPIP parameter tcp_nodelayack is 0, even in high TPS(Transaction Per Second) environment can encounter the issue.
The issue can be fixed by setting the tcp_nodelayack=1. Please work with your AIX System Administrator for modifying the setting.
Translate this page: