Open Stack

Excerpts from Marcus White's message of 2014-08-23 09:55:24 -0700:
> Hello,
> I had some keystone and Horizon questions, would apprecite it if
> anyone can help:-)
> 
> a. Does Keystone run as a service scaled across multiple nodes? Since
> it would need access to shared data, I dont understand how it scales.
> 

It stores data in various backends. The default is SQL, and many use
LDAP to link with existing ID stores.

It runs as a REST API in front of those data backends, and scales out
quite nicely like any other web application.

> b. If it does not run on multiple nodes as a service, how can it
> handle a large number of concurrent requests?
> 
> c. Is it expected to be a service that is exposed to the clients on a
> public cloud? Does Rackspace today do it or any other public cloud?
> 

Yes it is expected to be exposed. Users authenticate to keystone and are
given a token which can be used to grant access to the other OpenStack
services.

> Some horizon questions also
> 
> a. Does Horizon allow customization for specific services? Is there
> any way to add value on top of it for my own Nova or Swift or Cinder
> view, can I do that or is it whatever is done by the community the
> only thing available?
> 

It's open source. Go wild. ;) I believe there are hooks for
skins/themes, and one can add plugins for other functionality.

> b. It seems like Horizon is basic right now, do vendors provide their
> own GUI for customers?
> 

See horizon.hpcloud.net for a public cloud provider (disclosure: my
employer) using Horizon.

> c. Is there a hardware monitoring also as part of Horizon? For
> servers? For storage arrays? Is there a standardized agent or
> something that runs on hardware?
> 

Horizon isn't really about "servers" and "arrays". It is meant as the
interface for the users of the cloud, not the operators.