Steps to debugging web site latency and timeout issues

Question

I have a client who has multiple offices around the country, all of which share the same Internet connection via their WAN. One specific office for this client is experiencing severe latency and timeout issues with my web site. Most, but not all, of the latency occurs on a specific ASPX page where multiple postbacks are made while populating cascading dropdown lists (rapid form submits). The latency is sporadic and can be anywhere from a few seconds to a full timeout. There is no indication that the timeouts are occurring on the server's end.

The IT guy for this client is having trouble narrowing down the problem. Since it is affecting only one location for one client, I am led to believe it is not something with my site but something specific to that location. He's measured ping times while using the site and has noticed no real variance in ping times even when the page has timed out. I believe this may be being caused by some sort of Internet filter that doesn't like rapid form submits, but beyond a hunch I haven't a clue.

My question is what things should I tell the IT guy to look for? While I'm not trying to provide active tech support for this issue, I would at least like to glean an understanding of what is going on and try to offer some sort of advice.

Thank you.

Answer 1

Assuming some of the basics have been done:

1. Determine it is an issue, and get away to reproduce it.
2. Try and reproduce timeouts, on different workstations at the affected site.
3. Try and reproduce timeouts, at a different branch site.

There are several things you can use to debug timeouts on the client side:

1. Use browser debugging tool kits (firebug/fiddler) to 'catch' timeouts or errors.
2. Setup network/bandwidth monitoring.
3. Check firewall/filter logs and investigate HTTP options.

Hope that gives you a starting point.

Answer 2

The more data that you can gather, the better. I highly recommend getting a network trace of the problem in action, maybe by running Wireshark on one of the computers remotely. A full packet capture like this could show you exactly what is happening at the network layer between the two systems.

Also, consider setting up a nagios instance (or whatever your monitoring tool of choice is) at the remote location. It can be configured to monitor ICMP latency / packet loss as well as perform basic HTTP checks against the site at a regular interval. This might give you some insight into a potential intermittent network issue. If you can identify a problem, he can then right custom checks to help you zero in on the exact issue.

Hope this helps!