Wednesday, January 6, 2010

UC@GMC - Connecting to PSTN

Next to come is the connection of your internal OCS environment to the world.


I am puzzled every time when see folks working hard to integrate existing VoIP platforms with OCS. For me this does not make sense – adding one complex system on top of another, thus doubling the chance of disruption the normal business operations big time if (or when) something goes wrong. Indeed, one cannot “pull the plug” of the old system at once since migration to new hardware (endpoints that is) for a big number of users can’t happen overnight, and yet… I see how some colleagues want to have 100% working test environment for demonstration and justification purposes (indeed full blown UC is a powerful convincing tool), but planning to run it in production is insane. Not that it will not work; just the overhead is too much…

Actually, if you have an existing hosted PBX (analog or digital incl. VoIP), by carefully planning the migration steps, you can move your users to Microsoft UC platform working your way from department to building to campus to DLC. As I mentioned before, GMC had PBX (sort of), hosted on our provider premise. You might find interesting the fact that due to the stubbornness of our provider, (refused to release the phone numbers for porting), we changed ALL phone numbers throughout the State of Georgia. You see, reducing the MRC up to 80% could be a very effective argument of otherwise No-No in the EDU sector.

Those familiar with OCS know already that the platform is very tight in meaning of what can and cannot connect to it. Plus, Microsoft adds many proprietary SIP messages (needed for different parts of UC) and so, we have native conflict(s) when comes to SIP protocol - not that MS does not comply, sometimes follows the RFC “too strict” I might say. Long story short, there is role called Mediation Server – a role necessary to make the connection between “standard” SIP (device, provider etc.) and the internal OCS roles. Careful examination if the traffic in and out of the “external” interface facing the gateway shows just a standard SIP and nothing else. This is how I got the idea about SIP Trunking with VoIP provider, BTW…

There are not so many options when comes to connecting OCS to PSTN. We can use Gateway or Trunk (although some use the term “trunk” to describe the physical connection between Mediation and Gateway). In any case, some sort of phone service – analog (POTS), BRI (ISDN), T1 (PRI), E1 (Euro PRI) and so on, must terminate on your premise to a gateway (analog or digital) and this gateway will convert it to SIP and RTP thus making it “understandable” for Mediation and further more OCS. Since we did not have any PBX or other device, GMC had the leverage to test and consider any scenario. So, I was playing with Audiocodes analog gateway and at some point I was like “Wait a second – this traffic looks a lot like my VoIP phone’s traffic at home”! Quick call to CallCentric (VoIP provider for my test account), revealed an ugly truth - that they do not support Sip over TCP. Took me about a week to find a US based provider who agreed to test SIP over TCP trunk with GMC – www.boadvox.com

Here I need to say something – I considered Microsoft certified partner first (no names) and even ran quick interop test which worked beautifully. One problem though – even they offered “local” DID’s i.e. Milledgeville numbering scheme, the billing Local Calling Area was Atlanta and so, when someone from Milledgeville calls (478) 387-xxxx (local number), the caller would be billed LONG DISTANCE charges because the termination of the trunk is actually Atlanta. Bad idea! The local phone companies fall from the band wagon right away - $1,300 for a single T1 PRI (23 voice channels) or $56.52. What were they thinking!!!

Back to Broadvox… a concurrent call (single trunk) cost ~ $13 - $15 (depends of the type of the contract), and so our Milledgeville campus trunk (40 concurrent calls) cost us now… a little over $500. Can’t beat this. Basically, we accept SIP and RTP traffic from set of IP addresses (a distributed failover) and send traffic to FQDN (load balancer). Of course, one might say “SIP trunk is a single point of failure” and this is correct. However, we presented the Pros and Cons of every option to the Boss and the ultimate decision was to achieve maximum savings while recognize the risks.

No comments: