Hi Avadh,
This works fine. But, the memory usage with cache=writeback or cache=unsafe 
option is huge. Like when I run after restoring the checkpoint with these drive 
cache options, resident memory is ~9GB.
Without this, earlier it was 1-2G. Did you also see the problem?

Thanks
Ishwar
From: avadh patel [mailto:[email protected]]
Sent: Monday, June 27, 2011 7:32 PM
To: Ishwar Singh Bhati
Cc: [email protected]
Subject: Re: qemu-0.14 checkpoint slow...

Since Qemu-0.13 version they have changed default disk cache settings so 
creating checkpoint and loading checkpoint with default parameters take too 
long. You have to use 'cache=unsafe' option to create checkpoints faster. I 
have just pushed few fixes to 'qemu-0.14' and 'core-models' that will make sure 
that checkpoint is created correctly when using 'cache=unsafe' option.

new command : $ qemu/qemu-system-x86_64 -drive 
file=/your/qcow2/image,cache=unsafe

Also when you load the checkpoint use the same 'cache' option.

- Avadh
On Mon, Jun 27, 2011 at 1:25 PM, Ishwar Singh Bhati 
<[email protected]<mailto:[email protected]>> wrote:
Hi All,

Is anybody using the "qemu-0.14" branch of marss with checkpoints?

I am calling the function "ptlcall_checkpoint_and_shutdown(str_checkname)" at 
the region of interest of my benchmark to get the checkpoint.
This works well in master branch, but it takes either very long time (~12 
hours) or never finishes checkpointing in qemu-0.14 branch.
Did somebody find any reason/solution for this?

Thanks
Ishwar

_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to