The OLSR wikipedia page doesn't do a very good job of analyzing the strengths and weaknesses of OLSR.
The big problem with OLSR is that it's fairly new, immature, and not widely used or supported (mainly open source roll your own solutions and StarOS are the only ones that I know of off the top of my head that use it). Still, why it is attractive... (or could be if more common / standardized).... Yes, OLSR does push routing tables to all devices (as does OSPF and BGP)...I call that a feature, not a flaw. Link-state (ie OSPF and BGP) protocols are much better than distant vector (ie RIP) simply because routers will make much better decisions if they can "see" the entire network at once instead of just what the next node is reporting. Sure, that does take more memory and CPU, but the alternative is much worse... There are some "theoretical" other approaches, but nothing that, as far as I know, is more than a gleam in the eye of some grad student. The OLSR page failed to mention the main reason why OLSR is theoretically attractive over OSPF--link state quality (there has been some noise about adding this onto OSPF, but, it's largely just noise at this point and nothing that one could really implement). In other words, OLSR (technically via an extension) has the ability to choose routes based not just on link speed, load, link state (is it up or down), but also on how little packet loss is being experienced across the link. So, with OSPF, a 10Mb/s interface that is has no packet loss will "lose out" to a 100Mb/s interface that has some packet loss (as long as the packet loss doesn't "down" the interface or is the result of load, which can also be calculated). Which, is great for wired connections, where you're dealing with very low bit error rates and so forth. One wired Ethernet link is, pretty much 100% of the time, pretty much identical to the next. Wireless, of course, does have a wider variance. OLSR performs rudimentary packet loss calculations across the links and takes this information into account to give preference to good links over not so good links. http://www.olsr.org/docs/README-Link-Quality.html is a good writeup on this... OSPF is good for wireless if you are using very well engineered links (think nice point to point connections). So, if you are deploying mesh simply as a way of getting some redundancy in a network, then OSPF is definitely good. For some situations, though, the point of doing wireless mesh is that you make up for quality with quantity. Mesh takes the concept that, to some degree, multiple "less than perfect" links can, in aggregate, be as reliable as one very solid link...so, if you're going block by block in a city (for example), you may realize that some of your links will be problematic, at best. This is especially true among community wireless networks where your links are based on volunteers, not on design per-se. If that is the reason why you are using a mesh topology, then you would ideally need something that can differentiate based not just on speed and state of a link, but also on the quality of the connection of the link. Still, it is important to note that there are other problems associated with mesh that don't necessarily have anything to do with a routing protocol per-se; relying on multiple unreliable links to synthisize a reliable connection is problematic on other levels, since, if your network topology changes pretty frequently, you'll get packets coming in out of order and so forth... Clint Ricker -Kentnis Technologies -------------------------------------------------------------------------------- ** Join us at the WISPA Reception at 6:30 PM on October the 16th 2007 at ISPCON ** ** ISPCON Fall 2007 - October 16-18 - San Jose, CA www.ispcon.com ** ** THE INTERNET INDUSTRY EVENT ** ** FREE Exhibits and Events Pass available until August 31 ** ** Use Customer Code WSEMF7 when you register online at http://www.ispcon.com/register.php ** -------------------------------------------------------------------------------- WISPA Wants You! Join today! http://signup.wispa.org/ -------------------------------------------------------------------------------- WISPA Wireless List: wireless@wispa.org Subscribe/Unsubscribe: http://lists.wispa.org/mailman/listinfo/wireless Archives: http://lists.wispa.org/pipermail/wireless/
