.NET Debugging Demos Lab 4: High CPU Hang - Review

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

Tess1 — Thu, 12 Feb 2009 00:06:56 GMT

Nitin,

Have a look at some of the other posts in the post index, there are various different high cpu posts there that should probably give you the start that you need.

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

Nitin — Wed, 11 Feb 2009 19:26:32 GMT

Good article. But it is only focused on high CPU due to memory issues and garbage collection. I am running into a scenario when the CPU spikes to 100% for w3wp process but the system is doing fine for memory. I do not see more than 2% GC time and the ratio of Gen 0, Gen 1 and Gen 2 is perfectly fine.

What else could be wrong? How to debug that?

Using VSTS Test and Profilers to troubleshoot a High CPU in GC issue

If broken it is, fix it you should — Wed, 08 Oct 2008 11:24:21 GMT

Following my post about how to troubleshoot a low CPU performance issue with VSTS Test, I will now do

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

zantoro — Thu, 21 Aug 2008 14:05:23 GMT

you saved my ass!

Thank you

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

Nakuldev Patel — Tue, 04 Mar 2008 17:01:16 GMT

Thanks for sharing an amazing article.

/nakul.

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

charju — Thu, 28 Feb 2008 16:46:35 GMT

I love it!!! Very cool!!!

===========================================

Q: What is the ratio between Gen 0, Gen 1, and Gen 2 collections? What should it be in an optimal scenario and why?

A: Gen 0 collections are basically free. Gen 1 collections (which includes Gen 0 collections) are also pretty cheap since we are still dealing with a low amount of memory that needs to be collected, relocated etc... Gen 2 collections on the other hand can get really really expensive since it deals with the whole .net GC heap. An optimal ratio would be to see 100 Gen 0 collections, per 10 Gen 1 collections and 1 Gen 2 collection, so 100:10:1. In my case the numbers are

5853 Gen 0 Collections

5830 Gen 1 Collections

5430 Gen 2 Collections

That means that pretty much every single collection that was done was a Gen 2 collection (since this also triggers a 1 and 0 collection) which is really bad since it means that we take no advantage of having a generational GC, we basically scavenge through all memory all the time.

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

Tess1 — Thu, 28 Feb 2008 12:31:26 GMT

I gather this from the fact that the stack suggests that we are concatenating strings (i.e. potential for allocating a new string) and it is currently doin wstrcpy(char* destination, char* source, int charCount) which takes a couple of pointers to char* so it is likely allocating a char array to copy data from source to dest...

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

baal — Thu, 28 Feb 2008 12:26:53 GMT

can you put some pictures of the performance counter log on the page?

re: .NET Debugging Demos Lab 4: High CPU Hang - Review

baal — Thu, 28 Feb 2008 12:20:12 GMT

we got GCHeap::GarbageCollectGeneration in native stacks

Q: What type of object was it allocating?

A: Based on the managed stack it looks like it could be either a char[] or a string ----how do you get this answer？

.NET Debugging Demos

Paul Mooney — Wed, 27 Feb 2008 14:47:15 GMT

Lab 1: HangLab 2: CrashLab 3: MemoryLab 4: High CPU Hang