|
From: Edde <edd...@gm...> - 2006-07-05 10:35:28
|
Hi Guys, We've been experiencing some delay problems with our FIX application and are currently investigating our FIX logs and comparing these with our counterparty. I'm currently using QuickFIX/J but we had the same problems before when using the _jni version and QuickFIX 1.11.1. To try and narrow down the problem I've been comparing the timestamp added to the FileLogs (incl. millisecs) with the OrigSendTime (Tag 52) added by our counterparty. Since our server (running QuickFIX) is located within the same network as our counterparties FIX gateway I'd expect these times to be almost identical (the clocks are synchronized on the two servers). After comparing about 30 random messages from the FileLog I was surprized to see that there was a mean time difference of about 0.750s with a max diff reaching 1.8s. Since both servers are running on the same network this seems a bit strange to me. However, I'm not really sure what exact time in the pricessing chain the timestamp in the FileLog represents? Is this time created as soon as possible when the message arrives on the socket or is there significant processing involved before the message is written to the log? The same goes for the counterparty and Tag 52. Is this time added to the message just before it's being sent or is there any processing involved after Tag 52 has been added to the message? Any suggestions would be helpful and if anyone has other suggestions on how to track down these delays feel free to share them with me. |