(Page 2 of 2)
The key lies in identifying "hidden" bits of information that could help routers decide where to send a packet, Boguñá says. The people in Milgram's experiment used such information to figure out how to forward their letters. Instead of passing them on to a random friend, they identified criteria, such as a person's profession, that meant that they might be a step closer to the intended recipient. The work of Boguñá and his colleagues focuses on identifying and exploiting hidden information on other kinds of networks. In the case of Internet routing, the physical location of a router or the type of information it last handled could provide useful clues for forwarding information toward a final destination without knowing the complete structure of the network.
Kleinberg says that the work is "a very elegant approach to exploring the underlying structures that make navigability possible in real networks." He adds that Boguñá's group's "techniques have the potential to inform a new class of routing strategies in which global information is replaced by local strategies that follow hidden metrics."
However, Jon Crowcroft, a professor of communication systems at the University of Cambridge, U.K., warns that, while Boguñá's group has done good work in applying Kleinberg's theoretical models, it's too early to tell if the approach would actually work. "When you look at it in reality," he says, "there are other additional constraints," such as the requirements of particular applications. Nonetheless, Crowcroft believes that this direction is "absolutely worth exploring" and says that he would like to see the researchers try some real-world experiments.
Boguñá himself admits that his work is "very preliminary." The next step, he says, is to identify what "hidden metrics" could be used for Internet routing. But Boguñá expects that this could take several more years to figure out.
A proposed system would let cloud developers control the way their data travels across different machines.
Nick McKeown believes that remotely controlling network hardware with software can bring the Internet up to speed.
How an outdated algorithm put secure Internet transactions at risk.
Hybrid approach will be best, 2/3
Hybrid approach will be best, 2/3 look up 1/3 dynamic "pass along to a friend" approach...
As 1/3 finds more efficient low volume routes (or low congested routes if these routes handle less traffic) These can be swapped for over taxed main pathways to free up congestion on those routes only.
So lets say the 1/2 of the total or 3/6 would be static and when changed would need a system wide look up table update only thus rarely updated. the other 1/6 of the 4/6 of the static 2/3 would be suto-static and only be upated on a sub-set of total look up tables. The remaining 1/3 would be true dynamic with a constant sorting of free bandwitdth that can be swapped up to the top half of the 1/3 (1/6) with the suto-static 1/6 as needed.
Also the static and suto-static would be sorted per congestion to optimize at look up update time.
with most congested being sorted into suto-static prioritization,to offload traffic to dynamic maximum free bandwith sorted prioritzation candidates.
Your fractions may vary...
:)
Peace out
Wouldn't it be interesting if switched systems turned out to be a better solution than routing? And the Bell system's century-old methods were the most efficient?!
The flip side of the small world phenomenon is non-localisation. When events at the very edge of the network cause changes at the core of the network. Particle physics provides plenty of evidence.
The six degree paths are two-way channels and so design must proceed with caution to avoid becoming a victim of unintended consequences.
Manufacturing in the United States is in trouble. That's bad news not just for the country's economy but for the future of innovation.
Our list of the 50 most innovative companies, including the following:
On Twitter
Become a Fan on Facebook
Subscribe to the Feed
phoenix
172 Comments
information overload
All arguments of a particular function are defined as a set of parameters and as such fall into a definite set of categories each displaying a different set of characteristics. The only problem with using probability factors is that you have to distinguish between free or bound variables. Although the probability framework may apply, attention cannot shift from estimating the variables from observed data to testing hypothesis about them. While classical estimations consider parameters as fixed but unknown, Bayesian models constitute random variables with their own distribution. In other words trying to assign an arbitrary set of constants to find the hidden metrics of a particular function is like herding cats. While it is theoretically possible, the eventual outcome is pretty much a forgone conclusion.
Reply