In October 2008 the Communications Research Centre (CRC) hosted the 4th Optimized Link State Routing Protocol Workshop (OLSR), drawing participants from Europe, North America and Asia. They came with a single purpose in mind: to discuss and test the next generation of software that will lead to a truly wireless world. The software, being developed in part by CRC, will allow computers to communicate - to form networks - in the complete absence of infrastructure.

These mobile ad hoc networks (MANETs), explains Maoyu Wang, a research engineer with CRC's Mobile Ad hoc and Sensor Network Systems group, have tremendous potential for use in situations where the infrastructure has been destroyed by natural disasters or war. But they also have a distinctly Canadian context, she points out. MANETs could be used to link people in rural communities and, in some cases, even link those communities to the Internet by "stepping stone" nodes. While several prototype MANETs are up and running in Europe, the current version of routing software (OLSRv1) places limitations on their usefulness in real-life situations, something Wang and her CRC colleagues have set out to change. At CRC's October workshop, Wang showcased the lab's OLSRv2, a new version of the routing protocol that allows for fast and accurate movement of data within a mobile ad hoc network, but without the problems and limitations inherent in the earlier version.

"When you form a network via the Internet," she explains, "you require infrastructure. You have routers and communication beacons, fixed points that direct the flow of data packets to the various addresses. In a mobile ad hoc network you have no fixed points, so each computer must act as a router as well as an end host."

But, says Wang, getting each computer to act as a router is a complex problem. To function as a network router each computer must "know" how to send, receive and forward data packets such that they arrive at the right computer by the most efficient route, even if the end node is several "hops" away - meaning the packets must be routed through several other network computers to arrive at the end host. Each computer in the network, therefore, must be constantly aware of the network topography, and able to calculate the best possible route for a packet to travel, but the calculations must be made without using excessive memory or processing power since this is background activity invisible to the user. To complicate things further, the topography of the network is fluid. In the aftermath of a natural disaster, for example, new rescue teams may arrive by the hour and join the mobile ad hoc network to coordinate their efforts. In a peacekeeping or battle situation, a group of troops may secure an area then rapidly move on to the next village several kilometers away, all the while remaining an active node that must be coordinated with other troops.

While the first version of OLSR allows for the successful transfer of data packets within a mobile ad hoc network, says Wang, the system lacks the flexibility needed to accommodate real life applications. Compatibility is one of the issues.

"With the first version," Wang says, "if my computer sends a packet to you, your computer can only read the packet if it understands its strict definition. If there is even a little bit of the packet the computer doesn't understand, it rejects the whole packet and we can't communicate."

In addition, she says, the first version software is inflexible. To change one part of the software you have to change it all, which makes it difficult to add the features that researchers now know are necessary, such as network hierarchies and security.

"We needed a new type of protocol capable of supporting a new type of network that would be more flexible and could be easily extended."

Participants at the October 2008 OLSR Workshop. Back row, left to right: Pedro Villanueva-Pena, Canada; Justin Dean, U.S.; Tomas Heide Clausen, France; Henning Rogge, Germany; Joakim Flathagen, Norway. Front row, left to right: Maoyu Wang, CRC, Canada; Aaron Kaplan, Austria; Jiazi Yi, France; Ulrich Herber, France; Ronald in'telt, Norway; Yannick Lacharité, CRC, Canada; Toru Matsuda, Japan; Dang Nguyen, CRC, Canada.

To develop the new OLSRv2 Wang and her colleagues made some conscious decisions early on. First, she says, they chose to design the new software using object-oriented programming. This would allow them to build it in modules or components. The result would be a much more flexible implementation where one component could be changed while leaving all others intact.

"This way, the software can be ported from one operating system to another. All that's required is a change in the component that communicates with the operating system."

Wang also decided to program the new OLSR in C++. While this made the actual coding of the protocol more difficult, it gave the end product much better performance than could be achieved using other languages, and increased the flexibility.

"If you receive a packet using this second version of the routing protocol, your computer just throws away the part it doesn't understand but uses the part is does understand. That's the first level of flexibility," says Wang, who adds that features can be included and the networks can be scaled up as needed.

Because the new OLSR can be easily extended, she says, it allow for the creation of hierarchies - in other words, the most powerful machines will be designated as hub or beacon nodes thus improving the efficiency and data transfer rates. The flexibility to add components will also allow designers to incorporate security tools into a mobile ad hoc network, tools that allow the network to recognize an intruder or alert network users to suspicious activity. Adequate network security is a critical feature in almost all real-world applications.

The October seminar was a milestone for Wang and her colleagues. Along with several other groups developing second generation OLSRs, Wang unveiled CRC's work-in-progress. After only four months of development time they arrived with the only fully functioning software package ready for the test bed. Using the new software, workshop participants set up a 10-node mobile ad hoc network and took the new OLSRv2 out on its first trial run.

"We wanted feedback," says Wang with pride. "We were told that our implementation was the best - with the best structure, best scalability, and best extendibility - of any of the implementations at the workshop."

With the trial run under their belts and valuable feedback from workshop colleagues, Wang expects the new OLSRv2 to be ready for practical use sometime next year, a feat made possible by the unyielding support of her lab head, Louise Lamont, Research Manager for Mobile Ad hoc and Sensor Network Systems. And there is, says Wang, already considerable interest from military and civilian quarters.

"We're very proud of the new OLSRv2," says Lamont. "Here at CRC, we see a bright future for the use of the OLSRv2 protocol in tactical mobile ad hoc networks because of its flexibility and versatility."

For more information contact Maoyu Wang, Research Engineer, Mobile Ad hoc and Sensor Network Systems, at 613-991-1671 or maoyu.wang@crc.gc.ca.