On 8/6/2015 6:51 AM, Vladimir Stackov wrote:
As fair as I understand Clear Linux is aiming to be a containerized applications
management system that was using KVM instead of namespaces/cgroups.
So I'm curious: what about guest memory reclaiming?
Memory reclaiming from virtualized OS is a pain and causes resource wasting,
KSM/ballooning had their own overhead.
I don't have to manage all of that things with namespaces/cgroups where I have single
running kernel that was managing available address space itself.
Does Clear Linux relies on KSM/ballooning or it did have their own tricks?
today we rely on sharing upfront (in the container setup, not VM setup yet),
but we have a plan to also do guest memory reclaiming... it just needs implementing.
(thanks for the poke; reminds me that I just need to code it ;-) )