23:33:26 <n0ano> #startmeeting kvm 23:33:26 <collabot> Meeting started Thu Aug 20 23:33:26 2015 UTC. The chair is n0ano. Information about MeetBot at http://wiki.debian.org/MeetBot. 23:33:26 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic. 23:33:26 <collabot> The meeting name has been set to 'kvm' 00:31:24 <n0ano> #topic planning etherpad 00:31:29 <n0ano> Don D. to send an email to the list to describe what needs to be 00:31:34 <n0ano> done to the etherpad and to encourage people to self signup for 00:31:40 <n0ano> tasks. If they do sign up for a task we will look for dates 00:31:48 <n0ano> when things will be done. 00:31:57 <n0ano> #topic qmeu tree 00:32:04 <n0ano> Should we keep a copy of the latest qemu stable release in the OPNFV 00:32:09 <n0ano> git repo. Don D. feels this is important as it makes a stable base 00:32:12 <n0ano> that everyone can easily get and also matches what we are doing for 00:32:16 <n0ano> the kernel. 00:32:28 <n0ano> #topic Live migration 00:32:40 <n0ano> China Mobile is using a FPA device in it's environment. The FPA is a 00:32:44 <n0ano> passthru to the guest so it's state has to be passed to the new 00:32:47 <n0ano> machine also. Downtime vs. total migration time. Downtime is the 00:32:50 <n0ano> time when the service cannot respond to a request, total migration 00:32:55 <n0ano> time can be much greater as long as the downtime is small. 00:32:59 <n0ano> What about CPU & Network overhead during migration. Not an issue 00:33:03 <n0ano> for CM because they only run 2 VMs with 4 cores per VM, since the 00:33:08 <n0ano> host has 16 cores they have plenty of CPU resources available to 00:33:12 <n0ano> do the migration overhead. Under subscription, especially for 00:33:15 <n0ano> NFV use cases, would be a good discussion to start on the ML. 00:33:19 <n0ano> CM is also using a dedicated 10G nic for management, including 00:33:22 <n0ano> migration, so migration network BW will not affect client access 00:33:28 <n0ano> to the server. 00:33:36 <n0ano> #endmeeting