To implement SIP trunking, you must route the connection through a Mediation Server, which proxies communications sessions between Lync Server 2010 clients and the service provider. The Mediation Server decodes the media traffic from clients and servers and re-encodes it before sending it to the service provider. The re-encoding is needed because SIP trunks do not support some codecs used like Real Time Audio (RTA) or Interactive Connectivity Establishment (ICE) protocol negotiation for firewall traversal.

Each Mediation Server can have two network adapters, which provide an internal and an external network interface. The external interface is commonly called the gateway interface because traditionally it has been used to connect to a PSTN gateway or an IP-PBX. To implement a SIP trunk, you connect the external interface to a Session Border Controller (SBC) at a service provider.

Centralized SIP trunking routes all VoIP traffic, including branch site traffic, through your data center. The centralized deployment model is simple, cost-effective, and generally the preferred approach for implementing SIP trunks with Lync Server 2010.

Depending on usage patterns within your enterprise, you may not want to route all users through the centralized SIP trunk. To analyze your needs, answer the following questions:

• How big is each site? How many users?
  
  
• Which Direct Inward Dialing (DID) numbers at each site get the most phone calls?
  
  

Distributed SIP trunking is a deployment model in which you implement a local SIP trunk at one or more branch sites. VoIP traffic is then routed from the branch site directly to their service provider, without going through your data center.

Distributed SIP trunking is required only in the following cases:

• The branch site requires survivable phone connectivity (for example, if the WAN goes down). If the branch does need redundancy and failover, the service provider will charge more and the configuration will take longer. This should be analyzed for each branch site. Some of your branches may require redundancy and failover, while others do not.
  
  
• The branch site and data center are in different countries/regions. For compatibility and legal reasons, you need at least one SIP trunk per country/region.
  
  

The decision about whether to deploy centralized or distributed SIP trunking requires a cost-benefit analysis. In some cases, it may be advantageous to opt for the distributed deployment model even if it is not required. In a completely centralized deployment, all branch site traffic is routed over WAN links. Instead of paying for the bandwidth required for WAN linking, you may want to use distributed SIP trunking.

Note:
For details about why and how you might use distributed SIP trunking, see Branch Site SIP Trunking in the Planning documentation.

Lync Server 2010 supports the following connection types for SIP trunking:

• Multiprotocol Label Switching (MPLS) is a private network that directs and carries data from one network node to the next. The bandwidth in an MPLS network is shared with other subscribers, and each data packet is assigned a label to distinguish one subscriber’s data from another’s. This connection type does not require VPN. A potential drawback is that excessive IP traffic can interfere with VoIP operation unless VoIP traffic is given priority.
  
  
• A private connection with no other traffic is typically the most reliable and secure connection type (for example, a leased fiber-optic connection or T1 line). This connection type provides the highest call-carrying capacity, but is typically the most expensive. VPN is not required. Private connections are appropriate for organizations with high call volumes or stringent security and availability requirements.
  
  
• The public Internet is the least expensive connection type, but also the least reliable with the lowest call-carrying capacity.
  
  

  Note:
  Internet connection is the only Lync Server 2010 SIP trunking connection type that requires VPN.

Selecting a Connection Type

The service provider proxy must support the following codecs:

• G.711 a-law (used primarily outside North America)
  
  
• G.711 µ-law (used in North America)