How successful have you been with setting up an Elastix server (or other opensource distro) in a datacenter / main site environment (perhaps a HA cluster as we have discussed) and used it in conjunction with a 3rd party SIP provider to provide phone service to multiple remote sites? My main interest is call quality.
As many know, I have my Elastix server running as a VM at our main site with little problem (and there will be less once the pipe gets to be 17 Mbps in a few weeks) using Broadvox for SIP trunks.
I was thinking I could use this solution to power the phones at my 3 remote sites (soon to be 4 of them in a couple months) as well since they are all on site-to-site VPN links with the main site. So I would of course make sure there is enough internet bandwidth at each site to support the number of concurrent calls I would need (knowing that extension to extension calls would go back to the main site first before connecting if one / both parties is at a different site), prioritize the VOIP traffic at the main site's gateway (already done) and at all other sites.
Will the call quality be considerably worse in this scenario than if I were to just put an Elastix box at each of the remote sites, still use SIP trunking with Broadvox at each site, and then trunk them together with IAX trunks in Elastix so we have 4-digit dialing between sites? I understand call quality cannot be guaranteed with a 3rd party SIP provider, but I am worried about the calls that have to go across the site-to-site tunnel to my main site, then out to Broadvox, and then back again.
Obviously making the Elastix server (or HA cluster eventually) at my main site the one phone server for all sites makes it a bit more critical in terms of uptime, but 2-4 hours would still be an acceptable amount of downtime I think (probably closer to 2 hours).
Here are the DR scenarios:
One centralized PBX
- 1. Internet failure at main site - all phone service for all sites goes down - at the SLA mercy of the ISP
- 2. Hardware failure at main site - back up and running in 2-4 hours no problem
PBX at each site trunked together
- 1. Internet failure at any site - only phone service for that site goes down - at the SLA mercy of the ISP
- 2. Hardware failure at any site - back up and running in 2-4 hours
I think call quality issues at a remote site would be much more difficult to troubleshoot with only one centralized PBX, but centralized management would sure be nice. In the event of hardware failure, I think restoring the one PBX at the main site would be easier since I have a failover ESXi host (which might not be the case at a remote site).
But, I don't need a failover ESXi host at the remote sites if I'm putting a PBX at each site because it's only servicing 10-15 people at each remote site (as opposed to 60+ at the main site).
I'm trying to let business requirements dictate my choice of solution here and feel like with either one I can hit my required SLA (unless my ISP tanks on me, which I can't control and does not happen often). I just need to know what would be best for call quality.
How do hosted VOIP providers do it?