Hello All,

Based on recent discussions I see two competing models for solving immediate type connection requests.

Option 1:

In the first solution (based on a proposal by Inder and described by Kudoh-san) a start-time of Now+GuardTime is used, where GuardTime is a hold-off time to prevent race conditions.  We allow provisioning to be signalled either by the requester NSA and locally by the provider NSA.  The guard-time is chosen by policy and is the (worst-case) time taken for reservation to complete.   In this model, there is no distinction between near-immediate and advanced, the difference is only in the start time provided by the requester.

The problem with option 1 is that it could be very slow – in large systems the guard-time could be several minutes.

Option 2:

An alternative solution to this problem (proposed by Jerry) would be to use the 'now' string in the start time.  To prevent race conditions, instigation of provision must be signalled from the Requester.  This would ensure that the provisioning phase is only begun once all reservations are completed.  The down-side with this proposal is that it would preclude support for locally instigated provisioning based on the Providers timer.  In this model the immediate reservation is distinguished by putting the ‘now’ string in the start time field.

If you believe that we have a requirement to support local instigation of provisioning, then this will make option 2 unworkable.

Too quickly gather feedback on these options, I have prepared the following web survey, please complete if you have an opinion on this topic:

http://www.surveymonkey.com/s/R8KGD9F

Guy

------------------------------------------------------------------------------------------------------------------

Guy  Roberts,  Ph.D

Network Engineering & Planning

DANTE - www.dante.net

Tel: +44 (0)1223 371 316

City House, 126-130 Hills Road 

Cambridge, CB2 1PQ, UK

------------------------------------------------------------------------------------------------------------------