Jumping straight to Profiler is probably the wrong first step. First try checking the Perfmon statistics on the server. I have a tutorial online here:
http://www.brentozar.com/perfmon
Start taking these figures, and then, after he has experienced one of these slowdowns, stop collecting. Look at the performance indicators at this time, and a bottleneck will appear. If you want to send me csv output from Perfmon to the address brento@brentozar.com , I can give you some idea of ββwhat is going on.
You may still need to run Profiler, but I would have ruled out the OS and hardware first. Also, just a thought - did you check the system and server application event logs to make sure that nothing is happening at this time? I have seen cases where, say, an anti-virus client downloads new patches too often and performs an easy scan after each update.
source share