How does the number of Method Servers improve performance

[avis]

Hi,

Still new to Windchill, so forgive my ignorance.

I am tuning my Windchill test server and am just curious about how exactly does the number of Method Servers effects the performance of the system? 

When would I know to increase the number of method servers and not just the heap size on those servers?

Also when would you ever want to have a different number of method servers than the recommended amount given in Configurator.properties? Is that even a realistic situation?

I guess maybe I do not fully understand what the method server does. I understand that the method server is responsible for the interaction of objects, and enforces their behavior and business rules,  but from the clients end is there specific performance issues that they would encounter?

For example, say I have the recommended number of method servers (based on my number of CPU cores), is there some event(s) XYZ that a client would have, or I could have visibility to, that would signal to me that I need to increase the number of CPU cores and therfore the number of Method Servers? 

Is it more complicated than just some XYZ event(s)?

Thank you in advance.

[BenLoosli · Accepted solution]

A lot depends on the number of users.

Each Method Server handles the transactions between the user and the Windchill database. If you have too few Method Servers, the requests that a user submits will wait in queue until a Method Server becomes available to server their request. The wait may be very small, but it is there. If your system has the CPU cores and memory, it does not hurt to have additional Method Servers. 

On my 12 core machines, I run 3 Method Servers and a Background Method Server for 6 users of Creo and an additional 10 Windchill users.