[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Memory Trace Project Help

On Tue, Aug 30, 2011 at 10:06 AM, Todd Deshane <todd.deshane@xxxxxxx> wrote:
> On Mon, Aug 29, 2011 at 5:25 AM, Sameer Pramod Niphadkar
> <spniphadkar@xxxxxxxxx> wrote:
>> Please feel free to comment and suggest for any new insights
> You may find this research on Satori interesting and perhaps related:
> http://www.cl.cam.ac.uk/~dgm36/publications/2009-milos2009satori.pdf
> Hope that helps.
> Thanks,
> Todd
> --
> Todd Deshane
> http://www.linkedin.com/in/deshantm
> http://www.xen.org/products/cloudxen.html
> http://runningxen.com/

Thanks Todd...

I knew about the Satori research before and basically got the idea for
my "heat zones" from their "enlightenment factor" - which is a means
of devising the sharing element to be used by the guest VMs. It has
also been mentioned in the paper that

"When an operating system loads data from disk, it is stored in the
page cache, and other researchers have noted that between 63.8% and
93.0% of shareable pages between VMs are part of the page cache.  For
example, VMs based on the same operating system will load identical
program binaries, configuration files and data files. In these
systems, the kernel text will also be identical, but this is loaded by
Xen domain builder (bootloader), and does not appear in the page

So the basic idea of my project can be summarized as   :

1. Find out about the processes running most frequently at a
particular time interval on different systems (this may be an easier

 2. Go deeper to the physical memory(PM) trace and find the
relationship between the PM addresses and most frequent access per
universal time clock per system.

I understand that with address space randomized mappings and with
different systems running different processes it might be hard to find
any suitable pattern emerging from this study. But as most of us know
that identical systems belonging in a particular network and during a
time frame might end up accessing similar PM blocks. (A block here
being groups of pages) I intend to  find if there is any kind of
correlation between this time frame and the access. According to the
working set model of a process, there exits a temporal and spatial
locality of memory page access and hence we end up using the
appropriate page replacement algorithms. Now I intent to see if this
same analogy can be applied to the entire memory address space for
access in a system. I mean if there exists some sort of a pattern
emerging for physical memory access based on time and space.

I hope to know if there has been any similar work done before with
memory traces or if there are any other areas which I need to look
into before I can begin this study.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.