Peer-to-Peer SIP:
More and Less than You Think

“Peer-to-peer.” The term seems to evoke an emotional response of one sort or another from almost everyone in the technical community. Within the music and movie industries, it conjures images of piracy, copyright violations and teenage evildoers who think stealing music isn’t stealing at all. Among ISPs, it stirs up concerns over network congestion and overload due to the volume of traffic and induces fears about liability. Among the general populace, some see it as a tool for getting free stuff, and others see it as a vehicle for getting into trouble.

Given the excitement and controversy, peer-to-peer VoIP and specifically peer-to-peer SIP is, unsurprisingly, a contentious topic.

Certainly, the idea of using SIP in a peer-to-peer configuration is not new. Indeed, SIP was designed with many of the same concepts that P2P is built on — the placement of intelligence in the endpoints rather than centralization within the network, for example. However, P2P and VoIP were only really brought together when Skype (News - Alert) came into view and became a smashing success. Given how close SIP already was to being P2P, interest surged in developing standards and technologies around true P2P SIP. Numerous documents proposing mechanisms and requirements for P2P SIP have been submitted to the Internet Engineering Task Force (IETF), and a pre-working group is being held during the November IETF meeting in San Diego.

What does P2P SIP mean for the industry? Is it just a tool for building standards-based clones of the Skype experience? This common perception represents perhaps the least useful and least likely benefit of the technology. Indeed, PSP SIP applications are quite broad and far-reaching:

Small Enterprise Systems: The cost of PBX (News - Alert) systems, both in terms of capital and operational expenses, is a big impediment to moving them down market into small enterprises and even SOHO markets. One idea for reducing that cost is to eliminate the actual PBX itself and instead provide phones that self-organize into a low-overhead P2P network.

Emergency Response: In national emergencies, such as a hurricane or terrorist attack, the telecom infrastructure might be out of service. However, the need for communications among the fire, policy, and government officials that are onsite is more critical than at any other time. One way to solve this is for an ad hoc wireless network to be put in place, and then layer on top of that a P2P communications system. Because there is no centralized infrastructure at all, the network is robust against failure and easily set up in time of need.

Privacy: Achieving true privacy and anonymity on the Internet is actually very challenging. Anyone who knows your IP address can readily find information about where you are. Indeed, even knowing who your SIP service provider is can reveal information about you that may be as good as revealing your identity. Consider, for example, a small enterprise with 10 employees. A call from someone in that enterprise narrows the set of callers down to 10 — not much privacy at all. Consequently, a true system for privacy requires a large worldwide interconnected mesh of loosely organized servers through which calls can be routed. Such a mesh would make it virtually impossible to ascertain the identity of a caller. This is exactly the kind of topology that P2P SIP can provide.

Low Cost, Highly Reliable VoIP: There is some belief that P2P SIP would reduce the cost of providing consumer VoIP services, since it has no centralized infrastructure. Furthermore, the lack of such infrastructure makes it highly robust, so that ‘five 9s’ of reliability can readily be achieved. Debate rages about whether this is really true. I believe that it is not. In practice, these P2P VoIP systems end up needing centralized servers for bootstrapping purposes and for firewall and NAT traversal. They are also required for the storage of sensitive data (such as voicemail or buddy lists). Furthermore, it’s not clear that the costs of customer support are really any lower with P2P VoIP than any other kind of telephony system. Finally, a traditional SIP system, with traditional proxies, can be built with relatively little server infrastructure. Indeed, traditional SIP proxies are stateless and systems built on them are highly resilient as a consequence. Is a stateless SIP proxy more expensive to operate than a bootstrapping P2P server? It’s far from clear that it is.

Finally, there are those who believe that the secret of Skype’s success is its use of P2P technology. However, I believe this is not true at all. The secret of its success is that the product features a great user interface, works reliably, sounds great, and has a fabulous distribution channel. None of that has anything to do with P2P. PC-to-PC clients based on traditional SIP can be built with similar properties: the Gizmo Project is one such example.

All of that said, there is no doubt that P2P SIP has many important applications in the VoIP marketplace. If you take something as controversial as P2P and combine it with something as hot as VoIP, how can the results be anything but exciting?

Jonathan Rosenberg is co-author of the original SIP specification (RFC 3261). He is currently a Cisco Fellow and Director of VoIP Service Provider Architecture for the Broadband Subscriber Applications Business Unit in the Voice Technology Group at Cisco Systems (News - Alert) .