Can someone give me pointers where to look
Check the Event Logs. Log files are the first place any administrator of any device or operating system should look. Twice. Always. Forever. No exceptions.
Server 2008 offers great filtering features for the event logs, so you can search by criticality, application source, event ID and etc. If you spend a few good hours crunching data, you should be able to recreate the history of crashes and have a great idea of what's going wrong.
If not, then blame stray alpha particles.
EDIT
Of course, I was remiss in delving deeper into the concept of event logs. I focused on the operating system. However, most enterprise-grade hardware also has event logs. If the OS is seemingly unaware of any problem on its hands, and yet the server reset itself, then perhaps you have faulty hardware that's tripping a restart response. I would suggest looking through any hardware logs that may exist for your server.
For example, in HP hardware that has an ILO card, you can sift through hardware logs for any events that might have occurred. Perhaps there was some PSU problem.
Going backwords even further, perhaps there was an issue with the PDU that your server is plugged into. Sort through those logs to see if there was some kind of power cycle that was tripped.
Trace the problem back from a top down perspective. Application -> Services -> Operating System -> Server Hardware -> Power Distribution. Each link in that chain will likely have some reporting mechanism that you can sift through to see a history of what has happened.
EDIT 2
Egads! I am a fool's fool! I left out the other most important place to look. When picking up the pieces after an OS crash, memory dumps are can lead you to the scene of the crime, the motive and the murder weapon. Once you learn how to analyze Windows crash dump files, you'll be a master detective.