EXPERT RESPONSE
This is an interesting problem. As with all troubleshooting problems of this magnitude and impact, the creation of a true problem statement and investigation into that is essential to understanding how to resolve the issue.
To troubleshoot this issue, I would suggest loading up some packet analyzers to try to isolate and find the problem. For example, I would have network monitoring captures on the servers going on during the time that you have issues. This would help you understand if there are packets actually arriving at the servers and whether or not the servers are responding to those requests. This would allow you to see if it is the network or the server that is experiencing the problem. It would also show you if the server is being overloaded with backup or other types of traffic that might cause it to stop responding to other types of requests. You would see a lot of RST bits set if the servers are resetting sessions. This would be the first start. Alternatively, you can start looking at either getting a tap or doing a SPAN (port mirror) to determine what the traffic patterns look like. NetFlow and other IP accounting technology also would help determine where traffic is going and what protocols are consuming the traffic to impede response time.
|
