Can ETW events be used to trace perf problems?

Question

How about it's latency? I mean how much time it will cost for one trace message? Is it in milliseconds level?

Answer 1

Certainly ETW can help identify perf issues. By default ETW measures every millisecond. ETW is built in to kernel and with least overhead you could identify bottle neck easily.

It especially true for VISTA and above client OS and Server 2008 because they have tracing capabilities, which provides the actual call stack where the time is being spent.

Here is an MSDN article and I also have bunch of posts on using ETW tracing with managed code