Just when it seemed peer-to-peer architectures might fade into the sunset, another horizon opens up.
Spurred by the spread of 802.11 and related protocols, new mobile and wireless applications are creating situations in which the lack of access to central servers offers opportunities uniquely suited to P2P.
According to Peter Stanforth, CTO of Florida-based MeshNetworks, P2P wireless applications fall into several categories, with the common characteristic being a lack of reliable connectivity to host networks.
For example, in many situations a network must go up in a hurry or go up temporarily for emergency services, a sporting event, or military combat. (The US Department of Defence's top IT professional, John Stenbit, has made "Power to the Edge" a mantra for military research efforts).
"The whole concept of moving the intelligence to the edge is the real interesting part," says Stanforth, whose company is building on technology licensed from the US Defense Advanced Research Projects Agency (DARPA) that leverages a peer architecture to extend the range and to improve the efficiency of wireless networks.
Mesh Networks is starting with basic applications, such as leveraging "multihopping" to allow a fixed area such as a corporate campus to be covered by fewer access points. Several access-point vendors are also working on peer approaches to managing and controlling large wireless networks, for example using the open-source Jabber IM XML protocol to more efficiently pass configuration and troubleshooting information across nodes.
Mesh Networks has begun trials with Orlando, Florida, city buses, using off-the-shelf 802.11 radios and a variant radio it has developed that utilizes real-time equalisation and a multitap rake receiver -- which can handle multipath and fading -- to enable multihopping networks at vehicle speeds as fast as 70 mph. Combined with relay devices posted on light poles (two per square mile), the system can provide a QoS (quality of service) sufficient to stream video and to support VoIP calls, Stanforth claims.
But mobile peer applications are not without their challenges. On a network level, routing algorithms must be more subtle and complex because wireless signal quality can vary randomly, with lots of fading and jitter. But the biggest challenges involve security, authentication, discovery, and scalability.
"It makes the whole issue of controlling access and security a bigger thing" if you don't have access to central DNS servers or directory servers to help determine identity and authorisation, Stanforth says. "The [P2P] network can find out all the nodes that are there, but do you want to talk to them, are they spying?" he says.
Stanforth adds that security must be tackled at both the link layer via encryption and at the application layer via VPN-like functionality, but should remain transparent to the user. "The challenge is to make applications that are really useful, but that don't need you to be an expert in wireless networking [to use]," he says.
"You put crypto providers at the edge of the network ... really pump up the level of encryption," agrees Michael Helfrich, vice president of applied technology for Groove Networks, which is also developing P2P technology to support mobile and wireless applications. "We decided early on that the best decentralised security architecture leverages smart clients, allowing them to ride on dumb networks."
With this approach, says Helfrich, the client negotiates "at a moment's notice" how it will connect securely to the other clients. Groove is also tackling the challenges of P2P synchronisation in an intermittent environment -- how to synchronise seamlessly without constantly having to hunt for the other devices.
According to Helfrich, this requires "a reliable message switch based on industry standards like XML or SOAP," as well as logic that won't crush the network and can ride atop readily accessible protocols such TCP/IP without special drivers.
Indeed, management overhead is so critical to mobile p-top applications that scalability becomes a key challenge. A key design tradeoff in P2P networks that affects performance and scalability is between proactive and reactive routing, Stanforth explains. In proactive mode, each node is always pinging the others, knows where each node is, and can communicate without latency. But this comes at a cost of time and bandwidth.
"I want the speed of proactiveness but don't want the overhead of having to keep track of everybody," Stanforth says. To this end, Mesh Networks continues to develop both proactive and reactive algorithms, along with some predictive logic, to optimise performance.
Such performance improvements may be key to some of the more futuristic potential applications of mobile P2P. In fact, Delphi, a leading automotive electronics vendor, is testing Mesh Networks' P2P technology for a host of in-vehicle telematics applications, including streaming entertainment, web access, safety, and location-based services.