News

• 1/3/2002: Fix minor memory leaks. Reintegrate testing code with autoconf build system. Fix bug in use of input handler in buffered output. Fix boolean enumeration issue for C++. Release libkl-1.3.8, libdpcap-1.3.8, nettimer-2.3.8.
• 7/25/2001: Update man page. Move documentation to correct location. Release libkl-1.3.4, nettimer-2.3.6.
• 7/24/2001: Release active probing code. Fixed many bugs. Release libkl-1.3.2, libdpcap-1.3.4, nettimer-2.3.4.
• 3/25/2001: Document log mode. Make log and update file mode consistent. Remove extraneous flows. Return an error when an uncaught exception occurs. Add quit command. Fix an error with SBPP. Release libkl-1.2.4, libdpcap-1.3.2, nettimer-2.3.2.
• 2/17/2001: Make a statically linked version of nettimer for non-RedHat 7.0 Linux people. Update docs on how to install statically linked version.
• 2/13/2001: Release libkl-1.2.2, libdpcap-1.2.2, and nettimer-2.2.2. Fixes core dump when -h option used.
• 2/13/2001: Release libkl-1.2.0, libdpcap-1.2.0, nettimer-2.2.0, and updated the manual. The main changes are
  • Use autoconf/automake so non-Redhat people have some hope of installation.
  • New default ncurses-based interface so users can easily watch the status of flows.
  • Added ability to control whether unqualified hostnames, fully domain-qualified hostnames or IP addresses should be used in output.
  • Fixed many bugs.
• 11/15/2000: Release the first version of the nettimer manual.
• 11/15/2000: Release nettimer-2.0.1. This version splits off some of the code into libraries and fixes many bugs.

About Nettimer

Nettimer is a project to do end-to-end network performance measurement. It can listen passively to existing network traffic or actively probe the network. End-to-end means that we don't depend on any special information from the network and we don't depend on a particular transport protocol. The metric that we've currently implemented is bottleneck link bandwidth.

However, doing fast, accurate bandwidth measurement is difficult because we can't use transport protocol throughput. TCP throughput is only an accurate measure of bandwidth when it transfers large amounts of data, which be very slow. End-to-end bandwidth measurement is difficult because congestion can make an otherwise high bandwidth network appear to have low banndwidth. Passive measurements are difficult because we have to filter out the behavior of any particular transport protocol.

Possible applications of Nettimer:

• Benchmarking: A user can determine if some network equipment/service is really delivering the claimed bandwidth.
• Web cache selection: Given several web proxy caches, a client could use bandwidth measurements in selecting the cache that would give it the best performance.
• Multicast tree construction: Multicast routers could use bandwidth measurements in selecting a path to join a multicast tree.
• Interface selection: A multi-homed host could use bandwidth measurements to select which of several network interfaces to use for a packet.
• Protocol tuning: Someone tuning a transport protocol could use the measured bandwidth as a guideline for the maximum possible throughput he or she could achieve with the transport protocol.
• Application-level network adaptation: An application could transform its data based on the current network conditions. Examples of this are real time video applications that reduce their frame rates when bandwidth drops and web servers that reduce the quality of their content (e.g. jpegs) when bandwidth drops.

Download

• Nettimer Statically Linked Release 2.3.8 [tar.gz]
• Nettimer Release 2.3.8 [i386 RPM] [SRPM] [tar.gz]
• libdpcap Release 1.3.8 [i386 RPM] [SRPM] [tar.gz]
• libkl Release 1.3.8 [i386 RPM] [SRPM] [tar.gz]

Documentation

Manual

The nettimer manual describes how to install, configure, run, and interpret the results of the nettimer application. [HTML] [PDF] [PS]

Frequently Asked Questions

[HTML]

Publications

• Kevin Lai and Mary Baker, "Nettimer: A Tool for Measuring Bottleneck Link Bandwidth", Proceedings of the USENIX Symposium on Internet Technologies and Systems, March 2001. [PDF] [PS] [HTML]
• Kevin Lai and Mary Baker, "Measuring Link Bandwidths Using a Deterministic Model of Packet Delay", Proceedings of ACM SIGCOMM 2000, August 2000. [PDF] [PS] [HTML]
• Kevin Lai and Mary Baker, "Measuring Bandwidth", Proceedings of IEEE INFOCOM '99, March 1999. [HTML] [ps.gz]

Talks

"Measuring Link Bandwidth using a Deterministic Model of Packet Delay" at SIGCOMM 2000 in Stockholm, Sweden.

[html] [ppt]

"Measuring Network Bandwidth" at INFOCOM 1999

This is a short technical talk discussing the difficulty of measuring bandwidth, limitations of current techniques, why Packet Pair is a good technique, and improvements to the Packet Pair algorithm. [html] [ps.gz]

"Measuring Internet Bandwidth" at Stanford Center for Telecommunications

This is a high level talk discussing the importance of measuring bandwidth, limitations of current techniques, why Packet Pair is a good technique, and improvements to the Packet Pair algorithm. [html] [ps.gz]