Admin guide Chapter: Partitioning your cloud

[Lorin Hochstein]

There are multiple ways to partition your compute cloud(s):

 * regions (multiple clouds tracked by one service catalog, exposed to end-user)
 * cells (hierarchically decomposing a large compute cloud for scalability, not exposed to end-user)
 * availability zones (simple decomposition of a compute cloud, exposed to end-user)
 * host aggregates (grouping of compute nodes, typically to achieve some additional functionality such as Xen storage pools or trusted computing nodes. May or may not be exposed to end-user)

I think we should have a chapter that's specifically dedicated to partitioning, with an overview that describes the differences between the different strategies and when to use which one, and then each section describes the different partition techniques.

[Tom Fifield (fifieldt)]

comment: cells is not just about scalability - it also works really well for geographical distribution, and actually it can be exposed to the end user

[koolhead17 (koolhead17)]

Is concept of availability zones still exists in Grizzly?

[Anne Gentle (annegentle)]

Atul - yes. See http://www.cloudcomp.ch/2013/02/how-do-you-organise-your-openstack-deployment

[Lorin Hochstein (lorinh)]

The confusion is likely due to the fact that nova used to have something called "zones", which worked a little bit like cells (allowed you to partition your cloud), and no longer exists. But the old "zones" was always a separate thing from "availability zones", even though they both have the world "zones" in them.

[Tom Fifield (fifieldt)]

Covered by the Ops guide: https://blueprints.launchpad.net/openstack-manuals/+spec/openstack-operations-manual