Open Stack

On 01/18/2013 01:59 PM, Brian Lamar wrote:
> Woah there, why are we talking about scaling?

Sorry, I probably should not have brought up scaling, as I was referring
more to how zones and regions are organized within a global deployment.

> I'm not 100% confident in my knowledge of availability zones, but aren't
> they just a lightweight concept that is basically nothing more than a
> property on compute nodes? How does scaling come in to play?

Availability zones and regions are merely ways of grouping resources
together -- not just compute resources, but block storage resource and
any other kind of resource. At least, that's the way I think of them. :)

> IMO AZs seem to be very much nova-centric and that data belongs to Nova
> or *potentially* some sort of orchestration service. The same goes for
> host aggregate groups, which I still very much confuse with AZs.

Host aggregates are not just ways of grouping host resources, but ways
of describing the capabilities of those groups of hosts. An AZ and a
region are not as specific, in my mind, and are merely tiers of grouping.

> Once again, my terminology may be off but I feel like you're talking
> about cells?

Cells come into play within a very large compute deployment. They
neither are useful nor applicable to non-compute resources (from what I
can tell...) or to deployments that do not have hundreds or thousands of
compute nodes in an AZ.

Again, the main point of my email is that the single source of truth
regarding what regions and availability zones are available in a global
deployment should be in the service that most naturally exposes global,
unified information. And IMHO, that is Keystone.

Best,
-jay