On 19 May 2009, at 12:16, Gary Mazz wrote:
You've got it...
I do like the idea of long life messages and events. I think events will be important just from the perspective of well know programming practices. I personally don't see unix signals going away for a long while.
Resources can change states for one reason or another. Networking and storage (another network) are often victims of accidental switch reconfigurations resulting in loss of connectivity to endpoints.
Configuration changes to resources outside the scope of the "active state" in the "life cycle model" may need to be logged or corrective action may be engaged.
Long life remote service dependencies are another area where events may be appropriate. Normally, I would suggest the applications handle occurrences between parties, but with migration as a normal circumstance, it may be out of control and scope of the parties. Either the infrastructure will handle a "pause for migration" event or the container will tell the app which will tell the other party. Queuing (a bridge) may be a way to avoid the issue as long as the service parameters are not violated. But it may be out of scope for interoperability.
Security or privilege revocation is another area. Hard execution breaks due to user or admin control is still another area.
Polling may be inappropriate in many cases, driving up CPU loads and network bandwidth consumption.
Going down the ATOM route, one does get a feed of events (cf. recent discussions related to holding the lifecycle of a resource as a collection of states in a feed). In addition, there are a number of approaches out there to alert interested parties to an updated feed, stopping relentless polling somewhat. I wonder if that is enough ? regards, Roger
I'll get them on the wiki in the later AM... I'm gmt -7
-gary
Sam Johnston wrote:
On Mon, May 18, 2009 at 7:55 PM, Gary Mazz <garymazzaferro@gmail.com <mailto:garymazzaferro@gmail.com>> wrote:
Hi,
From the current discussions, there has been no mention of events. Will the occi support events ? What will the model look like ?
This is something I had thought about a little but not considered a focus point for us. I did however consider the possibility of OCCI messages being long lived in that they can be queued and passed over transports like XMPP (this was one of a number of potential use cases for signatures).
The interest in doing this would be, for example, a client that has rendered a number of resources and wants to know about state changes and the like. They would just need to subscribe to the resource and when the state changes they would be sent the updated entry. Similarly an management system could ask for regular updates of e.g. performance monitor counters.
Does this sound adequate for your needs or do you have use case(s) that require more - and if so, why aren't they in the wiki ;)
Sam
_______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
Roger Menday (PhD) <roger.menday@uk.fujitsu.com> Senior Researcher, Fujitsu Laboratories of Europe Limited Hayes Park Central, Hayes End Road, Hayes, Middlesex, UB4 8FE, U.K. Tel: +44 (0) 208 606 4534 ______________________________________________________________________ Fujitsu Laboratories of Europe Limited Hayes Park Central, Hayes End Road, Hayes, Middlesex, UB4 8FE Registered No. 4153469 This e-mail and any attachments are for the sole use of addressee(s) and may contain information which is privileged and confidential. Unauthorised use or copying for disclosure is strictly prohibited. The fact that this e-mail has been scanned by Trendmicro Interscan and McAfee Groupshield does not guarantee that it has not been intercepted or amended nor that it is virus-free.