| Publication Title: |
M&M: A Passive Toolkit for Measuring, Correlating, and Tracking Path Characteristics |
| Publication Author: |
Katti, Sachin |
| Additional Authors: |
Dina Katabi, Eddie Kohler, Jacob Strauss |
| LCS Document Number: |
MIT-LCS-TR-945 |
| Publication Date: |
4-14-2004 |
| LCS Group: |
Networks and Mobile Systems |
| Additional URL: |
|
| Abstract: |
| This paper presents M&M, a passive measurement toolkit
suitable for large-scale studies of Internet path characteristics.
The multiQ tool uses equally-spaced mode gaps in TCP flows’
packet interarrival time distributions to detect multiple bottleneck
capacities and their relative order. Unlike previous tools,
multiQ can discover up to three bottlenecks fromthe tcpdump
trace of a single flow, and can work with acknowledgment as
well as data interarrivals.We also describe the mystery tool, a
simple TCP loss event, packet loss, and RTT analyzer designed
to work in concert with multiQ. The M&M toolkit can measure
simple path properties; correlate different types of measurement
of the same path, producing new kinds of results; and
because M&M is passive, it can use publicly-available traces to
track the value of a measurement over multiple years.
We validate our tools in depth using the RON overlay network
[4], which provides more than 400 heterogeneous Internet
paths and detailed information about their characteristics.
We compare multiQ with Nettimer and Pathrate, two other
capacity measurement tools, in the first wide-area, real-world
validation of capacity measurement techniques. Each tool accurately
discovers minimum capacities (85% of measurements
are within 10%of the true value); multiQ additionally discovers
multiple bottlenecks and their orderings. We also use our
toolkit to perform several measurement studies using a reservoir
of 375 million traced packets spanning the last two years.
Among the results of these studies are that bottleneck capacity
on our traced links has gone up by around an order ofmagnitude
from 2002 to 2004, and that differences in levels of statistical
multiplexing on 10 Mb/s and 100 Mb/s bottleneck links result
in flows over those links having similar fair-share bandwidths. |
| To obtain this publication: |
|
|
|
To purchase a printed copy of this publication please contact
MIT
Document Services.
|