When servers start, they start the number of processes specified by StartServers. If each of the processes is handling a request, incoming requests are queued. If the parent process detects that there are items in the queue, it starts a new child process after one second. After two seconds, if there are still requests in the queue, it starts two more child processes. After three seconds, it starts four more child processes. After four seconds, eight processes, and so on until the MaxClients limit is reached. It can take several seconds for the MaxClients limit to be reached.

Suppose you have 50 users who all synchronize their devices once per day, when they start work at 8:00 AM. Your sync server process is idle throughout most of the day but has a huge spike of activity at 8:00 AM. If the StartServers and MinSpareServers are set to 5 and MaxClients 50, it takes about 5 seconds for the server to spawn enough children to handle all of the incoming requests each morning.

To accommodate large spikes in server access, set the StartServers and MinSpareServers values to high numbers. Doing so starts up a high number of processes and leaves them idle until they are needed, which saves several seconds during start-up.