Mailing List Archive

Power saving
Hello,
I have tried to find answer myself but I have not found it.

I would like to build a cloud where several server are powered down and I
will power them on only on peak hours.

I need that virtual machines migrate to new servers and the migrate back
when
I power down them.

Is it possible?


Thanks in advance for interest!

Mario
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20110204/4cfdbdcf/attachment.html>
Power saving [ In reply to ]
Dear Mister Giammarco,

For your information :

http://www.asiafi.net/meeting/2010/summerschool/p/chu.pdf

(p. 16)

Best Regards,

Guillaume FORTAINETel : +33(0)631092519Mail :?gfortaine at gfortaine.biz

________________________________
> Date: Fri, 4 Feb 2011 13:38:59 +0100
> From: mgiammarco at gmail.com
> To: openstack-operators at lists.openstack.org
> Subject: [Openstack-operators] Power saving
>
> Hello,
> I have tried to find answer myself but I have not found it.
>
> I would like to build a cloud where several server are powered down and I
> will power them on only on peak hours.
>
> I need that virtual machines migrate to new servers and the migrate back when
> I power down them.
>
> Is it possible?
>
>
> Thanks in advance for interest!
>
> Mario
>
> _______________________________________________ Openstack-operators
> mailing list Openstack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
Power saving [ In reply to ]
On Feb 5, 2011, at 4:00 AM, openstack-operators-request at lists.openstack.org wrote:
> I have tried to find answer myself but I have not found it.
>
> I would like to build a cloud where several server are powered down and I
> will power them on only on peak hours.
>
> I need that virtual machines migrate to new servers and the migrate back
> when I power down them.
>
> Is it possible?

With the current OpenStack Bexar release, I don't think that is feasible.

Assuming that your power down algorithm was something like:
1. find all hosts with < x runnning VMs (where x is threshold of VMs feasible to migrate)
2. consolidate/migrate those VMs onto other hosts (chosen by load)
3. alert the scheduler that these VM-less hosts are powering down
4. power down the hosts

We would need the live migration feature (https://blueprints.launchpad.net/nova/+spec/cactus-migration-live) that is coming in Cactus (April timeframe) and probably some changes to the scheduler (or maybe not ?) and instance monitor.

And, of course, we would need actually code the algorithm that I have above :)

/k
Power saving [ In reply to ]
This should be turned in to a blueprint... we would need some BMC/IPMI
interfaces coded in to do this but once we have live migration in OpenStack
it is a totally feasible feature (pending algorithms below but those should
be modular anyways). If you want to create a blueprint for it Mario/Ken feel
free, otherwise let me know and we can discuss it and I can put a blueprint
together. It might not make cactus but it is a good enterprise level feature
to have on the roadmap.

-----Original Message-----
From: openstack-operators-bounces@lists.openstack.org
[mailto:openstack-operators-bounces at lists.openstack.org] On Behalf Of Ken
Pepple
Sent: Saturday, February 05, 2011 1:27 PM
To: openstack-operators at lists.openstack.org
Subject: Re: [Openstack-operators] Power saving

On Feb 5, 2011, at 4:00 AM, openstack-operators-request at lists.openstack.org
wrote:
> I have tried to find answer myself but I have not found it.
>
> I would like to build a cloud where several server are powered down
> and I will power them on only on peak hours.
>
> I need that virtual machines migrate to new servers and the migrate
> back when I power down them.
>
> Is it possible?

With the current OpenStack Bexar release, I don't think that is feasible.

Assuming that your power down algorithm was something like:
1. find all hosts with < x runnning VMs (where x is threshold of VMs
feasible to migrate)
2. consolidate/migrate those VMs onto other hosts (chosen by load)
3. alert the scheduler that these VM-less hosts are powering down
4. power down the hosts

We would need the live migration feature
(https://blueprints.launchpad.net/nova/+spec/cactus-migration-live) that is
coming in Cactus (April timeframe) and probably some changes to the
scheduler (or maybe not ?) and instance monitor.

And, of course, we would need actually code the algorithm that I have above
:)

/k
_______________________________________________
Openstack-operators mailing list
Openstack-operators at lists.openstack.org
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators