I would look for kernel based profilers, something like oprofile, perhaps...

I assume you need this for Linux.

The project at:

https://github.com/Carrefour

Contains a modules that scans the pages of proccesses and collects statistics.

You might be able to modify it in order to get the dirty bits for each page.

They have published a paper with the same name that also covers the overall design,so it might help you.

can you explain bit more? today i have just seen...

actually, i am working on Xen live migration and p2m (physical to machine frame) is very crucial task so i think process migration is able to generate pages

I do not remember the technical details right now, and i am not sure what is the question of your last post. Regardless Carrefour uses IBS in order to gather page access statistics, i think this is quite close to what you were asking in your initial post.

I think section 3.3.1 of the paper would make it more clear if you can use Carrefour for your purposes.(The paper: http://dl.acm.org/citation.cfm?id=2451157)

Stelios Sir,

thanks for your quick reply. I have seen Paper and code  but at this stage it would be difficult for me to extract information. May i request to give i) if log files of such output because i don't know whether NUMA architecture would work on my platform (any linux) and/or x86 architecture ii) if any standard log details of process migration available, so i am able to  extract page info from those logs....

- what are the parameters to migrate process and what is the role of kernel at the same time? can we generate a tiny OS program given in Figure 4. (https://www.academia.edu/760613/Survey_of_Virtual_Machine_Migration_Techniques)