On Wed, 2009-02-18 at 14:45 -0800, J Coatline wrote: > > Thanks everyone for your answers. > > > To mranga: The fact that I can only have an ITSP registration on only > one of the bridges is okay. I just want to have the US ITSP connected > via the US sipXbridge, and the UK ITSP connected via the UK > sipXbridge, which will match where most of the users using those line > will be. > > > > To milosz: At present, I have US users' calls tunneling back to the > UK, out onto the internet to the US ITSP. The latency and jitter here > is double what it would otherwise be. Although the Cisco 7965 phones > we use are really very good at handling latency, packet loss and > jitter, you basically end up with a 2 second one-way delay, which is > not really acceptable. > > > > To Michael: I would definitely trade a bit of latency on voicemail and > IVR calls into the US office for the benefits and simplicity of HA. > For example, by having only one cluster I can have UK users answer > inbound US calls and vice-versa, which I presume would be a lot more > hassle to set up if I use 2 separate installations. > > > > Is it a goal of sipXecs to become HA by default, where a single server > instance is just a special non-redundant case of this? I guess there > are a few services such a voicemail and IVR which are single instance > only right now, but I suppose voicemails could be replicated between > redundant servers, and could IVR not be multi-homed? Regarding > sipXbridge, once it is multi-instance, if the "system" detected a > server running sipXbridge had gone down, could it not have one of the > other instances of sipXbridge re-register against those ITSPs that > were running on the crashed server. (The superuser could select the > default sipXbridge server for ITSPs based on his view of the optimal > path, but at least if that server went down, calls could still come in > and out through that ITSP). If the system was told a little about the > topology of the network, it could presumably work out to allow > voicemails to be recorded on a server near the sipXbridge for incoming > calls, but let the end user connect to the voicemail server closest to > their phone to pick them up. > > > > I'm new to sipXecs so maybe these ideas are rubbish or they've already > been thought of and rejected as no good. >
The ideas aren't rubbish at all. HA was originally used for a single site and redundant routing. I had really hoped that with the move to the new Clustered configuration we would see redundant media services (VM, AA/IVR). But that's probably too much to handle in one upgrade cycle, plus they have a lot of other goodies on their plate (SBC, SIPTrunking, Conferencing Server, Location based Dial Plans, Emergency call notification, etc.). Plus, in the near future, I think the plan is to move to a freeswitch back-end for the media services (at least that's how it looks to me from the outside looking in). So it makes sense to make that move before working on the redundancy for those services. Most of the elements seem to be in place with the 4.0 release towards what you are envisioning. Mike _______________________________________________ sipx-users mailing list sipx-users@list.sipfoundry.org List Archive: http://list.sipfoundry.org/archive/sipx-users Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-users
