These slides are very interesting. There are a number of issues that this prompts that should be discussed, most of them noted in the slides. 1) handling of errors when there are failures in the middle of a sequence. This will have to be handled for the error case. If it is handled for that it might well be used in other cases - I think this should be considered. 2) the distinction between times -- what is required to set up vs what is needed for use seems very important. One way to think about this might be the time a resource needs to be reserved vs the time the resource needs to be available to the user. 3) set up time is likely known only by the NRM,and then possibly only statistically. A request for usable time might result in reservation for usable + setup. In this case having setup initiated from a user several steps away means that all resources in the chain will be reserved at least as long as the one with the longest setuptime. we can talk about this -- I am still thinking about it -- John
Peoples, I was interrupted before the end of the discussion on the maximum service duration that could be requested for schedules with no end time. In afterthought, would this not just be a general case for all schedule requests including the now request. I assume that if I have a policy in my network that no service can exceed 6 hours then this will apply to both unbounded requests and requests that have an end time exceeding the maximum policy. This can get extremely messy for the protocol if we have different policy values in the network since each NRM may be allowing different maximum durations that need to be communicated to the end requester. Comments? John.
participants (2)
-
John MacAuley
-
John Vollbrecht