Setting the PROCESS_CREATION_DELAY parameter

Tip Siemens Digital Industries Software recommends that you start your tuning process start by specifying an optimal setting for the PROCESS_CREATION_DELAY parameter. The PROCESS_CREATION_DELAY parameter is the most important pool server parameter for preventing CPU overload due to starting new server processes. It is critical that creating new Teamcenter server processes in response to logon demand does not overwhelm the processing resources available on the pool server machine. When logon demand requires that the pool manager create a new server process and add it to the pool, the server pool PROCESS_CREATION_DELAY parameter value list determines the interval of time (in milliseconds) the pool manager waits before attempting to create a new TcServer process. The default value for the parameter is the list 2000 2000 8000 16000 30000 60000. Example If the most recent attempt to create a Teamcenter server processes succeeded, then the pool manager waits the first interval in the list before attempting to create a new server process. If the attempt to create a new process fails, then the manager increments a failure count, waits the second interval in the list, and then again attempts to create a new server process. If a second attempt fails, then the pool manager increments the failure count and waits the third interval in the list (default value is 8000 milliseconds) before a third attempt. When an attempt to start a server process succeeds, the failure count resets to zero, and for the next creation attempt, the pool manager waits the first value in the list. Out of the box, the PROCESS_CREATION_DELAY parameter value is not explicitly present in the serverPool.properties file. To override the default, use the following procedure to find your own optimal first value and explicitly add the parameter to the file. Note After you complete deployment of the Server Manager, future changes to Server Manager parameters must be made through the Teamcenter management console. Changes are not retrieved from the serverPooldatabase-name.properties file after initial deployment. After you make changes to Server Manager parameters, you must restart server pools for changes to take effect. Procedure For your specific installation configuration and machine, measure the amount of operating system processing resources consumed by a Teamcenter server process started and added to the warm pool. This can be done simply by looking at the processing resources consumed by a server in a newly started up (but unused) pool. This amount can be expressed as [warmCPUSec] in units of CPU seconds. Determine what fraction of the machine can be dedicated to creating a new server process in response to logon demand during the peak usage of the system. Siemens Digital Industries Software recommends you choose a fraction between 0.4 and 0.7. This fraction can be expressed as [warmFraction] (unitless). A lower value for this fraction reserves more of the machine processing resources for concurrent nonstartup activity. Determine how many CPUs are on the machine ([numCPUs]). Calculate the optimum process creation delay in units of milliseconds (mSec) as follows: Copy [optimumDelay]=([warmCPUSec][1000 mSec/Sec])/([warmFraction][numCPUs]) Example Copy [warmCPUSec] = 8 seconds [warmFraction] = 0.6 [num_CPUs] = 2 [optimumDelay] = (8 * 1000) / (0.6 * 2) = 6667 Add the PROCESS_CREATION_DELAY parameter with a value list of increasing delays to the serverPool.properties file, starting with the optimum (or slightly under) delay, and additional delays ramping up to about 60000 milliseconds. Example Copy PROCESS_CREATION_DELAY = 6667 7000 12000 24000 40000 60000 In our example of a low-powered machine where each server launch occupies a core for eight seconds, the machine is overloaded if the pool manager needs to start at least two Teamcenter server processes at the default rate of one every four seconds, even with no actual request processing. Therefore, setting the PROCESS_CREATION_DELAY parameter to greater than the default values list is critical to avoid overloading the machine.