It's important for a cloud broker because I want to
stipulate, e.g., "I need n VMs that can be suspended and do not cost
money when in that state".
Enumerating the state machine is one way to support that. It's
expensive, though, isn't it? And advertising the state transitions only
works for an actual resource, and it only allows you to see "one step"
ahead, not the entire state transition diagram - unless I'm willing to
push the buttons and change my resource's state.
Anyway I think we're into extension territory here.
.. Shlomo
On Thu, Sep 3, 2009 at 8:36 PM, Sam Johnston
<samj@samj.net> wrote:
Will we also provide an API to discover the
lifecycle capabilities of the provider? Or the lifecycle that is
applicable for a specific abstract category of resource, or a specific
actual resource?
These capabilities are important in order to support the use case of a
"cloud broker".
I have just added a discovery capability last week which covers the
resource types or "kinds", categories, [open]search, etc. - we may be
able to enumerate the state machine if it makes sense to do so (maybe
it does - I'm not sure - maybe it makes more sense as an extension).
I'm not sure how this is critical for the "cloud broker" use case, but
regardless we will be advertising possible state transitions via link
relations.
Sam
On Thu, Sep 3, 2009 at 8:13 PM, Sam
Johnston
<samj@samj.net>
wrote:
If you feel that an "occi life cycle" recommendation is needed at this
point, we should discuss it. For now, I would prefer to report provider
life cycle states mapped to occi enumerated state "values".
Right, that's more or less what I had in mind - I'll be happy if we're
not locking anyone down to any particular approach, because you can be
sure the second we do someone will want something else (the ability to
"destroy" an active resource being an obvious one). The fact that
"destroy" maps directly to the "DELETE" verb rather than a state
transition is another example of a reason to avoid overspecifying it.
I do think we could afford to give guidance in this department though -
both in the form of a normative state table and an informative state
diagram (that is, a "may" or "should" requirement level).
Cheers,
Sam
On Thu, Sep 3, 2009 at 10:19 AM, Sam
Johnston
<samj@samj.net>
wrote:
Gary,
While I still don't think we need nor want to lock down a state
diagram, if we were to recommend one this looks fairly sensible.
Reality is that if we mandate anything it won't mesh with all
implementations so we'll exclude people unnecessarily.
Sam
Hi,
I put together a simplified version of the OCCI life cycle (state)
diagram. I'll be using this as an overview and for discussion
purposes. I'll provide a description in a couple of days.
-gary
_______________________________________________
occi-wg mailing list
occi-wg@ogf.org
http://www.ogf.org/mailman/listinfo/occi-wg
_______________________________________________
occi-wg mailing list
occi-wg@ogf.org
http://www.ogf.org/mailman/listinfo/occi-wg