On 05/13/2013 11:16 AM, Michael S. Tsirkin wrote:
However, there's a big question mark: host specifies inflate, guest says deflate, who wins?
If we're dealing with a NUMA guest, they could both win :) The host could see reduced memory use of the guest in one place, while the guest could see increased memory availability in another place... I also suspect that having some "churn" could help sort out exactly what the working set is.
At some point Google sent patches that gave guest complete control over the balloon. This has the advantage that management isn't involved.
I believe the Google patches still included some way for the host to initiate balloon inflation on the guest side, because the guest internal state alone is not enough to see when the host is under memory pressure. I discussed the project with the Google developers in question a little over a year ago, but I do not remember whether their pressure notification went through qemu, or directly from the host kernel to the guest kernel...
And at some level it seems to make sense: why set an upper limit on size of the balloon? The bigger it is, the better.
Response time. If too much of a guest's memory has been removed, it can take too long for the guest to react to user requests, be it over the web or ssh or something else... -- All rights reversed -- 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
