Measurement studies of Internet performance are critical for
validating or refuting widely held beliefs about Web behavior, and for
shedding light on unknown behaviors. Results from these studies can
guide Internet architects in making decisions that affect Internet
Service Providers (ISPs), content providers and end-users. Examples of
decisions that can benefit from measurement information include
provisioning network capacity, placing Domain Name System (DNS) and
Web servers, and tuning parameters of transport layer protocols.
Internet protocols and services may exhibit different performance
characteristics when observed from different locations in the Internet
topology; to date, however, there has been little work investigating
the differences in these characteristics from multiple vantage
points. Typically, performance studies present results of measurements
taken in only one or two locations. Some of the reasons for the lack
of work in this area are the following. First, performance
measurement was not a high priority of Internet design and was not
built into the network architecture. Second, it is difficult to obtain
the necessary level of privilege at many different locations in the
Internet topology to perform measurement studies. Finally, high
expectations for real-time Internet performance is a relatively recent
phenomenon.
In this thesis we develop several methods for gathering
Internet performance data from multiple locations throughout the
world, and to analyze data gathered. Our focus is on the protocols
and services that support the World Wide Web.
In the first method we utilize a modified Web proxy. Our proxy
captures and logs fine-grained performance information on a per-user
basis. Our second method is to create and deploy a measurement package
for examining DNS performance. We modified the BIND DNS server and
packaged it with a script to drive the data collection. Our final
method is to create and distribute an application to be run at user
sites worldwide. One of the primary tasks of the application is to
provide performance data from each instance of the application
executing at locations throughout Internet topology. We can use the
information provided by this application to examine user-perceived
Internet performance throughout the globe.
Identifer | oai:union.ndltd.org:GATECH/oai:smartech.gatech.edu:1853/4803 |
Date | 27 August 2004 |
Creators | Liston, John Richard |
Publisher | Georgia Institute of Technology |
Source Sets | Georgia Tech Electronic Thesis and Dissertation Archive |
Language | en_US |
Detected Language | English |
Type | Dissertation |
Format | 484967 bytes, application/pdf |
Page generated in 0.0012 seconds