[occi-wg] Cluster kind or such

[Gary Mazz]

Andy,

We keep on bumping up against this question and the topic of extensions. 
I think we may need to specify an extension best practices or a frame work.

-gary

Edmonds, AndrewX wrote:
> Hey Gyula!
>
> Great to have another OCCI adopter on board! 
>
> Yes currently there is no _explicit_ representation of a cluster or a "system of compute entities" in the specification. What I would suggest is two-fold. First, have a look at the section on Categories. I've used categories to group multiple compute resources into "services" (could easily be cluster depending on your context) and then perform operations across those compute entities that share the same service category. Second, if you find that categories are either not rich enough or that they are simply not appropriate then, the group would sincerely appreciate a proposal on how best you would see this specified in the specification. This brings me on to your question of peer review. If you do have something that you would like to see brought into the specification, describe it as best you can and present it to the mailing list. It's here were the OCCI community can then discuss the suggestion and see how best to accommodate the suggestion, whether it be an OCCI extension 
>  or part of the core specification. There is also the weekly OCCI confcall [1] and its facilities available to discuss suggested modifications and additions to the specification
>
> HTH,
>
> Andy
>
> [1] http://forge.ogf.org/sf/wiki/do/viewPage/projects.occi-wg/wiki/TeleconferenceCalls
>
> -----Original Message-----
> From: occi-wg-bounces at ogf.org [mailto:occi-wg-bounces at ogf.org] On Behalf Of Csom Gyula
> Sent: Tuesday, May 11, 2010 3:38 PM
> To: occi-wg at ogf.org
> Subject: [occi-wg] Cluster kind or such
>
> Hi!
>
> We decided to go with the OCCI interface within our cloud implementation (and just to be specific we chose OpenNebula as our CMS platform). We hope that our selection will prove to be a long term one, thus we'd like to adhere to the OCCI specification as much as possible. Meanwhile we have to deal with clusters, too, nost just single compute entities. My main question is whether you have any plan to incorporate compute systems (like clusters) within the OCCI specification?  
>
> ---
>
> In our context we'll have to deal with different types of clusters such as compute clusters (with one frontend, nas and many compute nodes) and load balancing clusters (with one load balancer, many application servers and backend storage). In many scenarios these systems must be handled as one single entity (eg. must be acquired, deployed at once, must be deleted at once, etc.). After all one might consider them as resouces, just like any others (compute, storage, and network).
>
> So the question is whether you plan to introduce compute systems or such into the specification?  
>
> ---
>
> If yes, what will be the shape model? main entities and relations?  
>
> For instance in our context all clusters will share the same simple structure where a system can be characterized through roles, and node counts (per roles). This might be handled by 1 additional main entity (system), and 3 pseudo entities (role, compute link and system link).  
> See the attached UML diagram: cluster.png. Of course more complex scenarios are also possible (peer-to-peer compute relation, system of systems, etc.) but such systems are out of our current scope...
>
> ---
>
> In the spec you put the following: "A simple peer review process is available for extending the registries which should be used where possible." My last question is regarding this
> process:  
>
> How does the above peer review work? When and where should someone go to make his request?:)
>
> Cheers,
> Gyula
> -------------------------------------------------------------
> Intel Ireland Limited (Branch)
> Collinstown Industrial Park, Leixlip, County Kildare, Ireland
> Registered Number: E902934
>
> This e-mail and any attachments may contain confidential material for
> the sole use of the intended recipient(s). Any review or distribution
> by others is strictly prohibited. If you are not the intended
> recipient, please contact the sender and delete all copies.
>
> _______________________________________________
> occi-wg mailing list
> occi-wg at ogf.org
> http://www.ogf.org/mailman/listinfo/occi-wg
>
>