On 07/23/2018 10:36 PM, Dr. David Alan Gilbert wrote:
* Michael S. Tsirkin (mst@xxxxxxxxxx) wrote:
On Fri, Jul 20, 2018 at 04:33:00PM +0800, Wei Wang wrote:
This patch series is separated from the previous "Virtio-balloon
Enhancement" series. The new feature, VIRTIO_BALLOON_F_FREE_PAGE_HINT,
implemented by this series enables the virtio-balloon driver to report
hints of guest free pages to the host. It can be used to accelerate live
migration of VMs. Here is an introduction of this usage:
Live migration needs to transfer the VM's memory from the source machine
to the destination round by round. For the 1st round, all the VM's memory
is transferred. From the 2nd round, only the pieces of memory that were
written by the guest (after the 1st round) are transferred. One method
that is popularly used by the hypervisor to track which part of memory is
written is to write-protect all the guest memory.
This feature enables the optimization by skipping the transfer of guest
free pages during VM live migration. It is not concerned that the memory
pages are used after they are given to the hypervisor as a hint of the
free pages, because they will be tracked by the hypervisor and transferred
in the subsequent round if they are used and written.
* Tests
- Test Environment
Host: Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz
Guest: 8G RAM, 4 vCPU
Migration setup: migrate_set_speed 100G, migrate_set_downtime 2 second
- Test Results
- Idle Guest Live Migration Time (results are averaged over 10 runs):
- Optimization v.s. Legacy = 409ms vs 1757ms --> ~77% reduction
(setting page poisoning zero and enabling ksm don't affect the
comparison result)
- Guest with Linux Compilation Workload (make bzImage -j4):
- Live Migration Time (average)
Optimization v.s. Legacy = 1407ms v.s. 2528ms --> ~44% reduction
- Linux Compilation Time
Optimization v.s. Legacy = 5min4s v.s. 5min12s
--> no obvious difference
I'd like to see dgilbert's take on whether this kind of gain
justifies adding a PV interfaces, and what kind of guest workload
is appropriate.
Cc'd.
Well, 44% is great ... although the measurement is a bit weird.
a) A 2 second downtime is very large; 300-500ms is more normal
b) I'm not sure what the 'average' is - is that just between a bunch of
repeated migrations?
c) What load was running in the guest during the live migration?
An interesting measurement to add would be to do the same test but
with a VM with a lot more RAM but the same load; you'd hope the gain
would be even better.
It would be interesting, especially because the users who are interested
are people creating VMs allocated with lots of extra memory (for the
worst case) but most of the time migrating when it's fairly idle.
Dave
Hi Dave,
The results of the added experiments have been shown in the v37 cover
letter.
Could you have a look at https://lkml.org/lkml/2018/8/27/29 . Thanks.
Best,
Wei
