Forum Moderators: phranque
We are running IIS6 on Windows server 2003 and use classic asp with buffering on to serve regular html and gif images (no fancy audio/video etc). We use the custom error page (404.asp) method to catch and interpret html extesnion files by doing Server.Transfer (In case this matters).
Really scratching our head to find out where the bandwidth leak is occurring and stop this overage. Any tips?
you might have to turn on the compression in IIS, that will reduce a ton of bandwidth ( but eat some CPU at first ) - this is common, and it's a quick fix
you might have someone hot linking your images or content ( frame )-- seen that and cost me a bundle
you might want to check your database look up and make sure that it's local and not pulling the data from one server to the other server then posting it to the user IE: query database , pull data to local, post/print to local -- seen this more than once where a person has a few servers that communicate at the switch, it would just be a misconfiguration but a very costly one ( costing you 2 to 3 times your bandwidth )
e-mail issues, very doubtful, but it could happen
OOOHHHH - check for a real big file I bet that might be the issue, it would never load fully but it would get requested all the time. -- had that happen to me, my backup log files were being read all the time by search engines until I killed the access
Mojomike
I agree stats packages like analog may not be accurate, but multiple stats packages are all reporting < 4 Gig a day and the host is reporting 20 Gig a day!
Ya, I had looked into compression and delaying that path since all of our files are generated dynamically via a 404 catch, so I think we wont be able to utilize the caching benefit leading to performance issues.
Hot-linking (images and frames) seems possible, but these would still be logged by IIS and read by stats packages, right?
Databases are local, e-mail is minimal, but anyway the huge chunk in question is reported clearly as port 80, outgoing traffic by the host control panel.
The big file theory seems interesting. Took a quick look, but couldn't find anything. Would these escape the web server log file? Also, the host reports bandwidth granularly to a 5 minute interval and the high number is uniform throughout the day. So, it almost looks like there is some big TCP overhead or something like that not caught at the application (web server) logging level?
Everything else was working out at this host, but guess we will have to move hosts, once again, sigh ;)
