What UCaaS deployment model suits enterprises with hybrid infrastructure?

Suppose an enterprise has three kinds of site. Its 400-person headquarters has a phone system (PBX) and a carrier contract that runs past the migration date. Its distribution center needs phones that work through an internet outage. Its four branch offices, with 20 people each, have nothing to preserve. The company can't move everything at once, so the deployment has to let the headquarters PBX and carrier keep working beside the cloud service until it's ready to retire them.

Microsoft Teams Phone makes a useful worked example, because its documentation treats each of these choices as a separate feature. On coexistence, Microsoft's overview of PSTN connectivity options says one tenant can use several connections to the public telephone network at once, serving different users, locations and regions. In our example, the branches can take numbers from Microsoft or from a certified operator, and they need no telephony gateway or SBC. Headquarters has something to preserve, so it uses Direct Routing. We compare the operator routes in the difference between Direct Routing and Operator Connect.

Direct Routing connects Teams to your own telephony through a session border controller (SBC), a device that sits between the cloud service and your carrier trunks or phone systems. Microsoft's Direct Routing planning guide names the cases it fits, among them an existing carrier relationship, a third-party PBX and analog devices. It also describes a common mixed pattern. Ordinary outside calling goes through a Microsoft Calling Plan or Operator Connect, and Direct Routing handles calls to the old PBX and the analog devices.

Hybrid doesn't oblige you to own that SBC. Microsoft says it can be procured, installed and managed by you, by your integrator, or by a provider that sells Direct Routing as a service. So the first question for each site is what has to keep working. Who runs the equipment is a second choice you make afterward.

Coexistence can also run the other way, with the old phone system kept in place and cloud applications added to it. Atlantech Online's UCaaS page describes this. Its SIP trunking extends a softphone to the users of an existing PBX, so a business can start using unified communications before it replaces the phone system. Whether that works with the PBX you have is something the provider confirms against your system. We cover the general case in whether SIP trunking works with an existing phone system.

Atlantech's hosted phone system is a different service from the Teams example, and it's priced by the user. Atlantech runs the phone system in its own cloud and advertises two lines: a Business Line starting at $12 per user per month and a Premium Line starting at $18. Suppose our enterprise chose this service, kept headquarters on its PBX for now and moved the four branches onto Premium lines. At the starting rate, 80 users come to $1,440 a month in per-user charges, if every branch user is on that line and the quote comes in at that rate. The quote also sets the contract term. The SIP trunking for headquarters needs its own quote.

In this example the subscription grows with each site that moves. The old system's support and carrier costs continue for as long as the contracts and equipment you've kept require them. We walk through the sequence in how to migrate from an on-premises PBX to UCaaS.

A call has two parts. Signaling is the exchange of messages that sets the call up and ends it. Media is the audio. In a cloud deployment the signaling goes to the provider, but the audio doesn't have to follow it.

Microsoft's media bypass guide gives an example. A user in a Frankfurt building calls an outside number through an SBC in the same building. Without media bypass, the audio travels to a Microsoft datacenter in Amsterdam or Dublin and back to the SBC in Frankfurt. With media bypass turned on, the audio goes directly between the user's Teams client and the SBC. Microsoft says the longer route doesn't hurt call quality in most geographies, but it calls the loop unnecessary.

Two limits keep this decision in its place. Signaling always flows through Microsoft's cloud, so in ordinary operation a site with media bypass still depends on the cloud to set up its calls. And Microsoft states that turning on media bypass doesn't mean all media stays inside the corporate network. A user outside the network who can't reach the SBC has their audio relayed through Microsoft. Calls handled by voice applications such as auto attendants and call queues always pass through Microsoft's media processors.

So media routing is a choice about the audio path for people who sit on the same network as the SBC. We'd turn it on where many users share a building with the SBC, because that's where the shorter path applies to the most calls. It isn't an outage plan, although the outage plan below requires it.

Cloud call control has a consequence. When a site's internet connection fails, devices that reach the provider only through that connection can't set up calls. The fallback is either another path to the cloud, such as a second connection or a mobile phone on cellular, or equipment on site that takes over.

For Direct Routing, the on-site option is Microsoft's Survivable Branch Appliance. It's software Microsoft supplies to SBC vendors, who build it into the SBC's firmware or ship it to run on a separate machine. The situation it covers is specific. The branch can't reach Microsoft's cloud, but its internal network still works and users can still reach the SBC that connects to the carrier. When that happens, the Teams client switches to the appliance on its own and calls in progress continue.

The requirements are just as specific. The SBC has to be configured for media bypass, and each user needs a branch survivability policy assigned ahead of time. The appliance works with the Teams desktop app on Windows and macOS and with Teams Phones. Desktop apps on virtual machines and Teams in a browser aren't supported.

What users get during the outage is a reduced phone. They can make and receive outside calls, put a call on hold, pass it to someone else with a blind transfer, and forward calls. A banner in the app tells users they're in appliance mode. Microsoft's troubleshooting note is that a user who doesn't see the banner isn't in that mode and can't call.

Emergency calls need attention before the outage. Microsoft says a user in appliance mode can still place an emergency call, but the caller's location isn't shared. The appliance also doesn't support emergency call routing policies. The emergency number has to match a pattern in the user's regular voice routing policy, or the call fails. We explain what location normally adds in what Enhanced 911 is.

One more dependency follows from how the appliance works. It sends calls out through the local SBC's carrier connection. If that carrier trunk rides the same internet circuit that just failed, the appliance has nowhere to send them. So a site that needs survivability also needs a carrier connection on a separate path.

These choices get made one site at a time, so take our enterprise site by site.

Its headquarters and distribution center follow the Teams Direct Routing design, since media bypass and the Survivable Branch Appliance are Teams features. Headquarters gets an SBC so the PBX and the carrier keep working while users move in groups. Its 400 users share a network with that SBC, so we'd turn on media bypass there. The distribution center gets the survivability appliance, a carrier connection on its own path, supported clients, and emergency dialing set up for appliance mode.

The four branches get plain cloud calling with no local call-control appliance, on whichever service carries them. Suppose each branch accepts mobile calling as its fallback, and its staff have that service's mobile app and working cellular coverage. A phone on a cellular connection doesn't depend on the office circuit. Atlantech's page documents apps of this kind for its own hosted service, for Android and iOS. A branch that can't accept those assumptions needs its own outage plan. A second internet connection is one option, and on Direct Routing the distribution center's design is another.

Each piece of site equipment in that design has a named reason to be there. We compare the two ends of this path in how UCaaS compares to an on-premises PBX. When the carrier contract ends or the PBX is retired, you know which equipment can go, and the deployment moves closer to plain cloud calling.