Re: RFC 4666 M3UA - Reg Request Message
"Brian F. G. Bidulock" <[email protected]>
| Newsgroups | gmane.ietf.sigtran |
|---|---|
| Organization | http://www.openss7.org/ |
| Message-ID | <[email protected]> |
Ilie, It does not alter the fact that DRN/TID labels have no notification procedures and cannot support fail-over. It is in this regard that DRN/TID labels are fundamentally broken. The fact that DRN/TID labels in RFC 3868 are not registered means that errors occurring at activation time, as dictated by 4.7.2.1, can cripple an entire AS. Registering the load selection as described in LOADSEL has the advantage that errors are detected at registration time instead of at activation time. The inability of multiple ASPs to activate for the same DRN/TID label range makes it impossible to fail-over or smoothly transition from one to another as is possible for the non-label model. For example, if an ASP registered for a label range fails and this is detected by another ASP, the other ASP cannot activate for the label range for the duration of time that the SG considers the failed ASP still active. None of these problems or issues exist for the LOADSEL approach. LOADSEL provides a notification procedure for fail-over; it detects load selection errors at registration (or configuration) time instead of at activation time; and it permits multiple ASPs to serve the same load selection, even in Override mode, permitting smooth ASP-controlled fail-over. No, DRN/TID labels were rushed in at the end, did not receive sufficient inspection or review, and are largely impractical when compared even to LOADSEL. --brian Ilie Glib wrote: (Tue, 07 Nov 2006 12:21:11) > Hello Brian, > > multiple SG as STP scenario in SUA is a dead end, at least with the > current SUA status. > > I think the difference between CIC and DRN or destination TID is > essential. CIC is a "resource" owned by both peers of the signalling > relation, while TID/DRN is owned only by one side of the signalling > relation. > More over DRNs and TIDs are software resources, while CICs are > physical resources. > > For the reasons above, static handling of DRN/TID labels is > acceptable, whereas CICs require a rather dynamic handling. That is > why I find loadsel or similar way as something mandatory for M3UA to > support ISUP, while it is rather nice to have for SUA. > > Regards > > /Ilie > -- Brian F. G. Bidulock [email protected] http://www.openss7.org/