Not sure what's going on here, but one connection per process seems... ancient.
Using threaded model is difficult, how about pre-fork? Some connections per process is a good improvement.
Using threaded model is difficult, how about pre-fork? Some connections per process is a good improvement.