Swing
Abstract:
The goal of this work is to generate high-speed packet
traces that reflect the characteristics of a target
link subject to particular application mix and
wide-area topology characteristics. Our methodology
accurately captures the statistical properties of
existing links based on available packet traces and
further enables extrapolation to future application
characteristics, bandwidth settings, and user
scenarios. The generated traces are valuable in a
variety of settings including design and evaluation of
queueing disciplines, traffic engineering policies
within or between networks, and studying higher-level
application sensitivity to variable and bursty cross
traffic.
This paper describes Swing, a closed-loop, network
responsive workload generator that accurately captures
the packet interactions of a wide variety of TCP and
UDP based applications using a simple set of
parameters. Starting from observed traffic at a single
point in the network, Swing automatically generates
individual settings for structural models of common
applications including HTTP, P2P, FTP, and SMTP along
with the wide-area characteristics like link delays,
capacities and loss rates. We verify the accuracy of
our models by extracting application characteristics
from available large-scale packet traces and by
validating our generated traces against the originals
using a variety of statistical techniques.
The primary contribution of this work is our ability to
capture and reproduce the burstiness in traffic across
a wide range of timescales using a handful of machines
running real TCP stacks over commodity Operating
Systems. We also demonstrate how Swing enables
fine-grained control over user characteristics,
application properties, and the underlying topology to
project traffic characteristics into the future.
Examples
University
of Auckland traces
Papers
Swing: Generating Realistic Background Packet Traces for Large-Scale Internet Topologies: Technical Report
Software
Undocumented Pre-release Coming Soon!
Mail Bugs to:kvishwanath[at]cs.ucsd.edu
People
Amin Vahdat
Kashi Vishwanath