No subject

fault tolerance. Thus the purpose of implementing clustering was primarily to
enable the setup of a system with no single point of failure. So cable
failure, host crash, or even a dead network switch will not lead to any
downtime (except loosing a few existing calls currently running on affected
curcuits).

Thus, while the basic facilities are there for routing the signalling across
the interconnect, it was never intended for a setup with only one signalling
link, so currently that may not work very well.

I would also think that with ~900 phone lines, you _would_ want some kind of
redundancy? Otherwise just the need to switch cabling or reboot the host
carrying the link will take down _all_ 900 lines! With two links, one on each
host, you can just set the lines on one host to maintenance mode, wait for
existing calls to end, then reboot/recable/whatever with no noticable effect
on users. This also holds true for the other end, who might assume this
functionality (as it is a mandatory part of the SS7 specs).

Also, I think the clustering still only supports two hosts. So you could put 4
quad-span cards into each server (4 PCI slots should be possible I guess), but
I don't know if/how Asterisk will handle that many simultaneous lines.

Maybe if the SS7 boxes do nothing but route over IAX to a larger number of
Asterisks that do the real work (transcoding, SIP, DB auth, whatever) it would
increase the chance of Asterisk being able to cope with 450 lines? But that is
just speculation on my part.

 - Kristian.