Call flow basics

Important

Azure Communication Services is currently in public preview. This preview version is provided without a service-level agreement, and it's not recommended for production workloads. Certain features might not be supported or might have constrained capabilities. For more information, see Supplemental Terms of Use for Microsoft Azure Previews.

The section below gives an overview of the call flows in Azure Communication Services. Signaling and media flows depend on the types of calls your users are making. Examples of call types include one-to-one VoIP, one-to-one PSTN, and group calls containing a combination of VoIP and PSTN-connected participants. Review Call types.

About signaling and media protocols

When you establish a peer-to-peer or group call, two protocols are used behind the scenes - HTTP (REST) for signaling and SRTP for media.

Signaling between the client libraries or between client libraries and Communication Services Signaling Controllers is handled with HTTP REST (TLS). For Real-Time Media Traffic (RTP), the User Datagram Protocol (UDP) is preferred. If the use of UDP is prevented by your firewall, the client library will use the Transmission Control Protocol (TCP) for media.

Let's review the signaling and media protocols in various scenarios.

Call flow cases

Case 1: VoIP where a direct connection between two devices is possible

In one-to-one VoIP or video calls, traffic prefers the most direct path. "Direct path" means that if two client libraries can reach each other directly, they'll establish a direct connection. This is usually possible when two client libraries are in the same subnet (for example, in a subnet 192.168.1.0/24) or two when the devices each live in subnets that can see each other (client libraries in subnet 10.10.0.0/16 and 192.168.1.0/24 can reach out each other).

Diagram showing a Direct VOIP call between users and Communication Services.

Case 2: VoIP where a direct connection between devices is not possible, but where connection between NAT devices is possible

If two devices are located in subnets that can't reach each other (for example, Alice works from a coffee shop and Bob works from his home office) but the connection between the NAT devices is possible, the client side client libraries will establish connectivity via NAT devices.

For Alice it will be the NAT of the coffee shop and for Bob it will be the NAT of the home office. Alice's device will send the external address of her NAT and Bob's will do the same. The client libraries learn the external addresses from a STUN (Session Traversal Utilities for NAT) service that Azure Communication Services provides free of charge. The logic that handles the handshake between Alice and Bob is embedded within the Azure Communication Services provided client libraries. (You don't need any additional configuration)

Diagram showing a VOIP call which utilizes a STUN connection.

Case 3: VoIP where neither a direct nor NAT connection is possible

If one or both client devices are behind a symmetric NAT, a separate cloud service to relay the media between the two client libraries is required. This service is called TURN (Traversal Using Relays around NAT) and is also provided by the Communication Services. The Communication Services calling client library automatically uses TURN services based on detected network conditions. Use of Microsoft's TURN service is charged separately.

Diagram showing a VOIP call which utilizes a TURN connection.

Case 4: Group calls with PSTN

Both signaling and media for PSTN Calls use the Azure Communication Services telephony resource. This resource is interconnected with other carriers.

PSTN media traffic flows through a component called Media Processor.

Diagram showing a PSTN Group Call with Communication Services.

Note

For those familiar with media processing, our Media Processor is also a Back to Back User Agent, as defined in RFC 3261 SIP: Session Initiation Protocol, meaning it can translate codecs when handling calls between Microsoft and Carrier networks. The Azure Communication Services Signaling Controller is Microsoft's implementation of an SIP Proxy per the same RFC.

For group calls, media and signaling always flow via the Azure Communication Services backend. The audio and/or video from all participants is mixed in the Media Processor component. All members of a group call send their audio and/or video streams to the media processor, which returns mixed media streams.

The default real-time protocol (RTP) for group calls is User Datagram Protocol (UDP).

Note

The Media Processor can act as a Multipoint Control Unit (MCU) or Selective Forwarding Unit (SFU)

Diagram showing UDP media process flow within Communication Services.

If the client library can't use UDP for media due to firewall restrictions, an attempt will be made to use the Transmission Control Protocol (TCP). Note that the Media Processor component requires UDP, so when this happens, the Communication Services TURN service will be added to the group call to translate TCP to UDP. TURN charges will be incurred in this case unless TURN capabilities are manually disabled.

Diagram showing TCP media process flow within Communication Services.

Case 5: Communication Services client library and Microsoft Teams in a scheduled Teams meeting

Signaling flows through the signaling controller. Media flows through the Media Processor. The signaling controller and Media Processor are shared between Communication Services and Microsoft Teams.

Diagram showing Communication Services client library and Teams Client in a scheduled Teams meeting.

Next steps

The following documents may be interesting to you: