Folks, Here is a proposal for scenarios to use at the resource model discussion at the face-to-face. (Chris, thanks for the feedback). Starting by what how to do the discussion, we should project the EMS picture (i.e., the one with all its services) on a whiteboard, then submit "job(s)" to the Job Manager. Then we follow all the information related to the resource model while we go through through the steps of discovery (which happens before job submission), resource selection, deployment, execution, etc.. I'm proposing two scenarios below, which put different emphasis on different points (queues vs online, dynamic vs static information, etc.). On each example I'm cutting details that would be redundant (storage in the first one, I/O bandwidth in the second, etc.). We probably will only have time to tackle one of them, and the first one should be complete enough. (1) Commercial Data Center (see OGSA use cases) We want to allocate two Web servers, two application servers and one database server in a data center, and run a Web site on it. This is an online system so resources are scheduled (i.e., the Web site comes live on a given date). The whole Web site is a single "job submission" which specifies the whole system structure and applications (including versions). The emphasis here is the system structure. [We might add monitoring the load on the resources to re-allocate resources according to the workload.] (2) Consolidation A data center has lots of 4-way machines, and lots of single-CPU jobs. Jobs are queued. You know how much memory each job uses. The objective here is to submit jobs in a way that maximizes utilization, so you want to select a machine that has enough memory and CPU available available to add a given job without thrashing it. I'm assuming that you are using dynamic information on load and memory with reasonably good "freshness" to make the selection. Regards, Fred Maciel Hitachi America R&D