Hi all!

As Ganeti is going to take part in GSoC 2016, I would like to propose a
couple of ideas for the ideas list:

- Migration Speed Accounting in htools (Haskell, Python)
  Hbal usually performs complex sequence of moves during cluster
  balancing in order to achieve local optimal cluster state. Unfortunately,
  some moves may take significant amount of time. Usually moves
  involving disk moves are significantly slower than just migrations and
  that's the reason for the --avoid-disk-moves option introduced in
  ganeti 2.17 (
http://docs.ganeti.org/ganeti/2.17/html/design-migration-speed-hbal.html).
  However, even migrations might be slow e.g. because of active memory
  modification. Thus, I want to propose the following:
  - implement move time estimation in HTools;
  - introduce time limits for cluster balancing steps;
  - introduce the overall balancing sequence time limit;
  - add --avoid-long-moves option, the --avoid-disks-moves analog.
  Note, that moves involving disk moves may take significantly different
  times, so disk move time should be somehow accounted as well.

- Better Containers Support (Python, a bit of Haskell)
  Currently Ganeti has a restricted support for LXC containers:
    - it's impossible to attach several disks to container;
    - container migration is not supported.
  Both are not LXC limitations any more. It seems that the most simple
  way to implement LXC migrations is to use LXD containers. With LXD
  containers instead of LXC some existing parts of current LXC support
  code might be simplified.

What do you think about the ideas? (Note, that I am unable to participate
as a student this year but I am ready to participate as a mentor).

Sincerely,
Oleg

Reply via email to