P
Paul Koufalis
Guest
Enable LOG-MANAGER logging on one client, specifically 4GLTrans and 4GLTrace. That will give you a good amount of information about timing on the client side. Look at the Degugging Guide in the documentation for details on how to do this - it's quite simple. It is very possible that when you are "finalizing a transaction", there is a very badly indexed FOR EACH occuring. On the server side you say "a 1 TB drive". Are you saying there is ONE hard disk of 1TB? If yes, you won't get much I/O throughput there unless it's an SSD. On the DB side, look specifically at checkpoints frequency and buffers-flushed-at-ckpt. If it's every 1-2-3 seconds then that's probably a problem. You can also see how much logical write vs physical write you are doing.
Continue reading...
Continue reading...