[Xen-devel] Question about distributed checkpointing with Xen

[Siddharth Aggarwal <saggarwa@xxxxxxxxxxx>]

Hi,

I am doing some research on time-travel in distributed systems. We have
implemented a disk checkpointer on FreeBSD and thinking about taking the
next step to checkpointing complete system state.

I also saw on the research roadmap for Xen, the statement

"Deterministic replay. Running in a virtual machine, it becomes possible
to record lightweight checkpoints of guests OSes as they run, and log
incoming events and interrupts. Thus, execution can be rolled back and
replayed in a deterministic fashion. This is great for debugging, and can
also be used to run two systems in lock-step to provide software
fail-over."

My debate is whether it is better to implement checkpointing at the
physical machine level, or at the virtual machine level. I figure it would
be easier to do it with a VMM, with the added overhead of needing a VMM.
Is there native OS support on BSD/Lunux to save and restore system state
or something that does it partially?

Has there been any discussion on pros and cons of checkpointing at the
physical machine versus VMM level? Can anybody point me to some
information about it?

Thanks,
Sid.



-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues
Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek.
It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/xen-devel