[openstack-dev] Climate Incubation Application

Dina Belova dbelova at mirantis.com
Tue Mar 4 13:25:30 UTC 2014


Joe, thanks for discussion.


> I think nova should natively support booting an instance for a

> limited amount of time. I would use this all the time to boot up

> devstack instances (boot devstack instance for 5 hours)


Really nice idea, but to provide time based resource management for any
resource type in OS (instance, volume, compute host, Heat stack, etc.) that
needs to be implemented in every project. And even with that feature
implemented, without central leasing service, there should be some other
reservation connected opportunities like user notifications about close end
of lease / energy efficiency, etc. that do not really fit idea of some
already existing project / program.


> Reserved and Spot Instances. I like Amazon's concept of reserved and

> spot instances it would be cool if we could support something similar


AWS reserved instances look like your first idea with instances booted for
a limited amount of time - even that in Amazon use case that's *much* time.
As for spot instances, I believe this idea is more about some billing
service that counts current instance/host/whatever price due to current
compute capacity load, etc.


> Boot an instances for 4 hours every morning. This sounds like

> something that
https://wiki.openstack.org/wiki/Mistral#Tasks_Scheduling_-_Cloud_Cron

> can handle.


That's not really thing we've implemented in Climate - we have not
implemented periodic tasks like that - now lease might be not started,
started and ended - without any 'sleeping' periods. Although, that's quite
nice idea to implement this feature using Mistral.


> Give someone 100 CPU hours per time period of quota. Support quotas

> by overall usage not current usage. This sounds like something that

> each service should support natively.


Quotas (if we speak about time management) should be satisfied in any time
period. Now in Climate that's done by getting cloud resources from common
pool at the lease creation moment - but, as you guess, that does not allow
to have "resource reusage" at the time lease has not started yet. To
implement resource reusage advanced quota management is truly needed. That
idea was the first at the very beginning of Climate project and we
definitely need that in future.


> Reserved Volume: Not sure how that works.


Now we're in the process of investigating this moment too. Ideally that
should be some kind of volume state, that simply means only DB record
without real block storage created - and it'll be created only at the lease
start date. But that requires many changes to Cinder. Other idea is to do
the same as Climate does with compute hosts - consider cinder-volumes as
dedicated to Climate and Climate will manage them itself. Reserved volume
idea came from thoughts about 'reserved stack' - to have working group like
vm+volume+assigned_ip time you really need that.


> Virtual Private Cloud.  It would be great to see OpenStack support a

> hardware isolated virtual private cloud, but not sure what the best

> way to implement that is.


There was proposal with pclouds by Phil Day, that was changed after
Icehouse summit to something new. First idea was to use exactly pclouds,
but as they are not implemented now, Climate works directly with hosts
aggregates to imitate them. In future, when we'll have opportunity to use
pcloud (it does not matter how it'll be called really), we'll do it, of
course.


> Capacity Planning. Sure, but it would be nice to see a more fleshed

> out story for it.


Sure. I believe, that having resource reusage opportunity (when lease
creation and resource allocation steps won't be the same one) will help to
manage future capacity peak loads - because cloud provider will know about
future user needs before resources will be really used.


Cheers

Dina


On Tue, Mar 4, 2014 at 12:30 AM, Joe Gordon <joe.gordon0 at gmail.com> wrote:

> Overall I think Climate is trying to address some very real use cases,
> but its unclear to me where these solutions should live or how to
> solve them. Furthermore I understand what a reservation means for nova
> but I am not sure what it means in Cinder, Swift etc.
>
> To give a few examples:
> * I think nova should natively support booting an instance for a
> limited amount of time. I would use this all the time to boot up
> devstack instances (boot devstack instance for 5 hours)
> * Reserved and Spot Instances. I like Amazon's concept of reserved and
> spot instances it would be cool if we could support something similar
> * Boot an instances for 4 hours every morning. This sounds like
> something that
> https://wiki.openstack.org/wiki/Mistral#Tasks_Scheduling_-_Cloud_Cron
> can handle.
> * Give someone 100 CPU hours per time period of quota. Support quotas
> by overall usage not current usage. This sounds like something that
> each service should support natively.
> * Reserved Volume: Not sure how that works.
> * Virtual Private Cloud.  It would be great to see OpenStack support a
> hardware isolated virtual private cloud, but not sure what the best
> way to implement that is.
> * Capacity Planning. Sure, but it would be nice to see a more fleshed
> out story for it.
>
>
> It would be nice to see more of these use cases discussed.
>
>
> On Mon, Mar 3, 2014 at 11:16 AM, Joe Gordon <joe.gordon0 at gmail.com> wrote:
> > On Mon, Mar 3, 2014 at 10:43 AM, Sean Dague <sean at dague.net> wrote:
> >> On 03/03/2014 01:35 PM, Joe Gordon wrote:
> >>> On Mon, Mar 3, 2014 at 10:01 AM, Zane Bitter <zbitter at redhat.com>
> wrote:
> >>>> On 03/03/14 12:32, Joe Gordon wrote:
> >>>>>>
> >>>>>>> - if you're reserving resources far before you'll need them, it'll
> be
> >>>>>>> cheaper
> >>>>>
> >>>>> Why? How does this save a provider money?
> >>>>
> >>>>
> >>>> If an operator has zero information about the level of future demand,
> they
> >>>> will have to spend a *lot* of money on excess capacity or risk
> running out.
> >>>> If an operator has perfect information about future demand, then they
> need
> >>>> spend no money on excess capacity. Everywhere in between, the amount
> of
> >>>> extra money they need to spend is a non-increasing function of the
> amount of
> >>>> information they have. QED.
> >>>
> >>> Sure. if an operator has perfect information about future demand they
> >>> won't need any excess capacity. But assuming you know some future
> >>> demand, how do you figure out how much of the future demand you know?
> >>> But sure I can see this as a potential money saver, but unclear by how
> >>> much. The Amazon model for this is a reservation is at minimum a year,
> >>> I am not sure how useful short term reservations would be in
> >>> determining future demand.
> >>
> >> There are other useful things with reservations though. In a private
> >> context the classic one is running number for close of business. Or
> >> software team that's working towards a release might want to preallocate
> >> resources for longer scale runs during a particular week.
> >
> > Why can't they pre-allocate now?
> >
> >>
> >> Reservation can really be about global policy giving some tenants more
> >> priority in getting resources than others (because you pre-allocate
> them).
> >>
> >> I also know that with a lot of the HPC teams using OpenStack, this is a
> >> fundamental part of scheduling. Not just the when, but the how long.
> >> Having systems automatically get reaped after a certain amount of time
> >> is something they very much want.
> >
> > Agreed, I think this should either be part of Nova or Heat directly.
> >
> >>
> >> So I think the general idea have merrit. I just think we need to make
> >> sure it integrates well with the rest of OpenStack, which I believe
> >> means strong coupling to the scheduler.
> >>
> >>         -Sean
> >>
> >> --
> >> Sean Dague
> >> Samsung Research America
> >> sean at dague.net / sean.dague at samsung.com
> >> http://dague.net
> >>
> >>
> >> _______________________________________________
> >> OpenStack-dev mailing list
> >> OpenStack-dev at lists.openstack.org
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> >>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>



-- 

Best regards,

Dina Belova

Software Engineer

Mirantis Inc.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140304/46f2da7d/attachment.html>


More information about the OpenStack-dev mailing list