Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Neil Hockert cs740 2/14/07

Summary: Neil Hockert
cs740 2/14/07
This paper sought to measure the effects of changes in the Internet using the BGP routing
algorithm. One main conclusion they found was that the Internet does not actually
support an effective method for dealing with implicit route replacement. The delay of
route convergence between ASes is due mostly to the differences between individual
vendor implementations of routers. The type of change to the network makes a
significant difference in convergence time. When a route is added to the network
convergence occurs much more quickly than when routes are removed. Using only
minor modifications to BGP implementations it would be possible to reduce minimum
convergence time from linear to constant complexity.
This paper used extensive actual network measurements from which they drew most of
their conclusions and they also had mathematical explanations for the theoretical bounds
on convergence complexity. A tremendous amount of data was collected over two years
for this paper so it seems like it should be possible to get a lot more information from it
than is presented in this paper. This paper was more focused so they may have used the
data for other papers as well and it was mentioned that related discussions were still
being researched.
This paper states that they confronted vendors about implementing the optimizations


Source: Akella, Aditya - Department of Computer Sciences, University of Wisconsin at Madison


Collections: Computer Technologies and Information Sciences