This paper introduces King, a tool that measures latency between
arbitrary end hosts using DNS. King is an active measurement tool that
neither requires active cooperation of end hosts nor the deployment of
special infrastructure such as landmarks. Based on the premise that most
hosts are near their authoritative name server, King uses recursive DNS
resolution to get the the latency between the name servers of the end
hosts and use this value to estimate the latency between the end hosts
themselves.
King is an insightful tool that makes use of existing infrastructure
(DNS) in an interesting way to get an accurate estimation of latency.
The paper presents a simple description of the tool and answers 4 key
questions about it well. Subsequently, King's performance, accuracy and
limitations are discussed at length supported with a huge amount of data
collected from previous experiments which makes their claim strong.
However, the entire evaluation section is based on the premise that the
values reported by the traceroute servers are accurate, which I am not
necessarily completely convinced about. I also do not have a more
accurate suggestion to be used as reference. Maybe King is as inaccurate
as traceroute.
Received on Thu Nov 03 2005 - 09:13:45 EST
This archive was generated by hypermail 2.2.0 : Thu Nov 03 2005 - 09:21:30 EST