The fundamental idea behind sizing a connection pool is that you should have enough connections to use all of the available resources, but not significantly more than that. The right size to saturate the server in most cases depends on the number of CPU cores, how much of the database is cached in memory, and the speed of the underlying disks. Once you've moved beyond the point where the server is busy all the time, adding more connections only serves to reduce efficiency, forcing the server to swap among multiple tasks when it would be better served with a smaller number.
It's hard to predict how many pooled connections to the database are necessary to keep it fully loaded without overloading it. Many users report ...
