There is a misunderstanding in the documentation of appldata/interval, which specifiies the interval at which Linux kernel sample data are collected for the CP monitor stream. The "Performance Toolkit Guide", SC24-6156-02, says it is specified in seconds and in an example it has:
"... to set the interval time: echo 1 > /proc/sys/appldata/interval For this command the '1' represents the sample interval, expressed in seconds of virtual CPU time to be used by the Linux system before refreshing the data collected by the CP monitor – in this case, 1 second." In fact it is specified in milliseconds, the default is 10.000. echo 1 > /proc/sys/appldata/interval might bring your system to a grinding halt. The "Virtualization Cookbook for SLES10 SP2", SG24-7493, also has this example, although it correctly states that the time is in milliseconds. Mit freundlichen Gruessen/Regards Klaus Bergmann
