On 01/01/2012 06:27 PM, Stefan Hajnoczi wrote:
On Sun, Jan 1, 2012 at 9:43 AM, Orit Wasserman<owasserm@xxxxxxxxxx> wrote:
On 12/30/2011 12:39 AM, Anthony Liguori wrote:
On 12/28/2011 07:25 PM, Isaku Yamahata wrote:
Intro
=====
This patch series implements postcopy live migration.[1]
As discussed at KVM forum 2011, dedicated character device is used for
distributed shared memory between migration source and destination.
Now we can discuss/benchmark/compare with precopy. I believe there are
much rooms for improvement.
[1] http://wiki.qemu.org/Features/PostCopyLiveMigration
Usage
=====
You need load umem character device on the host before starting migration.
Postcopy can be used for tcg and kvm accelarator. The implementation depend
on only linux umem character device. But the driver dependent code is split
into a file.
I tested only host page size == guest page size case, but the implementation
allows host page size != guest page size case.
The following options are added with this patch series.
- incoming part
command line options
-postcopy [-postcopy-flags<flags>]
where flags is for changing behavior for benchmark/debugging
Currently the following flags are available
0: default
1: enable touching page request
example:
qemu -postcopy -incoming tcp:0:4444 -monitor stdio -machine accel=kvm
- outging part
options for migrate command
migrate [-p [-n]] URI
-p: indicate postcopy migration
-n: disable background transferring pages: This is for benchmark/debugging
example:
migrate -p -n tcp:<dest ip address>:4444
TODO
====
- benchmark/evaluation. Especially how async page fault affects the result.
I'll review this series next week (Mike/Juan, please also review when you can).
But we really need to think hard about whether this is the right thing to take into the tree. I worry a lot about the fact that we don't test pre-copy migration nearly enough and adding a second form just introduces more things to test.
It's also not clear to me why post-copy is better. If you were going to sit down and explain to someone building a management tool when they should use pre-copy and when they should use post-copy, what would you tell them?
Start with pre-copy , if it doesn't converge switch to post-copy
Post-copy throttles the guest when page faults are encountered because
the destination machine waits for memory pages from the source
machine. Is there a reason this page fault-based throttling cannot be
done on the source machine with pre-copy migration? I'm not sure
post-copy provides new behavior in terms of convergence, we could do
the same with pre-copy migration.
There is different w/ these two approaches:
1. post-copy allows progress to vcpus that are not faulting at the
moment.
Assuming a subset of the guest vcpu can execute freely w/ their
memory already at the destination, they can get 100% cpu time.
The slowing down approach on the source host, slows down all vcpus.
2. Difference page access pattern
post-copy uses on-demand like paging, so the page that is really
required get transferred. The slow-down approach can just guess what
page to send first.
Post-copy has other advantages though, it immediately frees logical
CPUs on the source machine (though RAM and network bandwidth is still
required until migration completes).
W/ post-copy you can immediately free any page that got transferred to
the destination.
At the end of the day, it's performance testing using various scenarios
that can educate us whether post-copy worth the extra complexity over
slowing down the guest on the source.
Cheers,
Dor
Stefan
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
