Measuring BGP pass-through times
Even though BGP has been studied extensively within
the last few years, e.g., [1], we still do not have
a good understanding on how and where the delays occur.
Unexplainably large convergence times have been reported.
It is unclear whether they are attributable to protocol
interactions, implementation
idiosyncracies, hardware limitations or other factors.
While one may attribute convergence times greater than
30 minutes to route ap damping [2] and those that are
roughly multiples of 30 seconds and are less than 2
minutes to the MRAI timer [3] the wide variability of
in-between times is puzzling [4]. An unknown component
of the delay is the contribution of each router along
the path of the BGP update. A detailed exploration on
the delay of each router and how it relates to factors,
such as CPU load, number of BGP peers, etc., can be
very helpful for understanding such larger convergence
time. In this paper, a methodology for studying the
dependency of BGP pass-through times on
such factors is presented together with some initial
results.
|
Network Services 
