How SBCs Degrade Network Performance — and What You can Do About It

The controversy over the use of Session Border Controllers (SBCs) in the VoIP industry is well known. One of the biggest selling features of SBCs is NAT traversal, and many providers hosting services for VoIP devices on the public Internet have relied on SBCs to solve the NAT traversal problem, often deploying SBCs primarily for this reason. However, SBCs come with a host of disadvantages that must be carefully considered.

First, the need for specialized hardware capable of relaying a large number of simultaneous media sessions makes scaling out the service a relatively expensive venture. Oversubscription and redundancy/failover requirements further exacerbate these costs. Second, SBCs violate the end-to-end principle of the Internet. The lack of end-to-end transparency severely restricts the innovation of new features and extensions to the core signaling protocol. If the media relay service is not media-agnostic, you may be stuck with those CODECs and/or media types supported by the SBC. Third, your datacenter bandwidth costs and requirements will increase as a result of the need to relay all media through the datacenter. Finally, the media relay service introduces an unavoidable delay (latency) to the media path. Depending on the architecture and call scenario, a media stream may traverse multiple SBCs. Minimizing the number of relay “hops” on the media path should be a primary goal of any service designer.

The opposing, more scalable model to SBCs is to distribute the burden of NAT traversal out to the edge devices. For devices sitting behind non-symmetrical NATs (the majority of NATs), STUN is an ideal solution. Unfortunately, devices behind symmetrical (non-SIP aware) NATs require some form of public media-agnostic media relay service such as a TURN server. If your service model has influence over the customer installation, you may consider deploying SIP-aware NATs/firewalls which also offer important QoS features. The other benefits of SBCs (security, DoS attack prevention, etc.) should not be ignored however, but at the same time, many of these can also be distributed into other parts of the network.

In conclusion: if you are planning to deploy SBCs in order to solve the NAT traversal problem, know that there are many alternative, cheaper, more scalable solutions available that help preserve end-to-end transparency. Many SIP clients these days support a variety of NAT traversal techniques such as STUN, TURN, ICE and uPnP. The challenge is to accurately determine the right solution for each client on the network and, where possible, avoid latency inducing public media relay services. IT

Garth Judge (News - Alert), is Vice President of Research & Development of 8x8 (www.8x8.com), responsible for the development of new features and expansion of all current 8x8 communications platforms, including the 8x8 Virtual Office service, the company’s flagship hosted IP PBX (News - Alert) service for small to medium-sized businesses.

» Internet Telephony Magazine Table of Contents