TCP Experiment Automation Controlled Using Python (TEACUP)
A Tool for Automated TCP Testbed Experiments
OVERVIEW
The
Transmission Control Protocol (TCP) underpins much of today's Internet.
Mechanisms for negotiating TCP connections, retransmitting lost packets
and
rearranging out of order arrivals have remained relatively stable for
years. On-the-wire packet formats are likewise well defined.
The challenge that continues to vex academia and industry is ensuring
individual TCP connections can transfer data as fast as possible
without
overloading either the destination host (flow control) or the underlying
network path (congestion control). We built TEACUP (TCP
Experiment Automation Controlled Using Python) to automate many aspects
of running TCP performance experiments in our specially-constructed
physical testbed. TEACUP
enables repeatable testing of different TCP algorithms over a
range of emulated network path conditions, bottleneck rate limits and
bottleneck queuing disciplines.We hope TEACUP proves useful to other researchers who already have (or are interested in setting up) similar network testbeds.
Testbed environment
Using a dedicated testbed sits somewhere between simulation-based evaluations
(e.g. ns-2 or ns-3) and wide-area experiments over real networks (e.g. the Internet). Our
typical use-case involves TEACUP controlling a testbed containing
multiple hosts either side of a Linux- or FreeBSD-based
bottleneck router in the classic dumbbell topology. Figure 1
illustrates the testbed we used in-house while developing TEACUP, built
according to the details in CAIA-TR-150210C.
Figure 1 Example TEACUP testbed topology
TEACUP runs on the 'data and control' server, orchestrating the configuration of end hosts and
bottleneck router as required for particular experiments or ranges of
experiments. It automates the process of configuring the
bottleneck router for emulating different path delays, intrinsic packet
loss rates,
bottleneck rate limits, Active Queue Management (AQM) selection, ECN
support, and so forth. Automated end-host OS selection is supported when the end hosts
have Windows, Linux and/or
FreeBSD on different local hard drive partitions, or are dedicated to
running Mac OSX. If
the testbed includes suitable remote power control units (IP-PDUs),
TEACUP can automatically power cycle machines that get stuck during
reboot.
In addition to launching full experiments, the user can query, configure or reboot testbed hosts (individually or in groups) using TEACUP from the command line of the control server.
Experiments
TEACUP utilises a text-based configuration file
to define experiments
as combinations of parameters specifying desired network path and end host
conditions. When multiple values are provided (e.g. for the bottleneck
rate limit, or TCP congestion control algorithm), an experiment is made
up of multiple tests
-- consecutively run instances of the experiment run for each parameter
combination. For each experiment and test, TEACUP collects a
range of data, such as tcpdump files of traffic seen at all network interfaces,
FreeBSD SIFTR
logs and Linux Web10G
logs. TEACUP also collects a variety of metadata from the end hosts and
bottleneck router (such as the actual OS/kernel version(s) in use,
network interface configuration, NTP state and so forth).
Analysis
TEACUP also provides some simple tools for analysing the
results of experiments. These include plotting
against time TCP's round trip time (RTT) estimate, congestion window size (CWND), SPP's
RTT estimate (from tcpdump files), and estimated throughput. TEACUP can
also plot any other TCP statistic gathered by SIFTR or Web10g over
time.
Rudimentary animation of captured TCP traffic is provided by TEAPLOT.
Further reading
The rest of this website contains an introduction to TEACUP and its functionality,
some examples on how to use TEACUP (including the configuration files), and the
source code of TEACUP itself and some TCP statistics logging tools we developed
for Linux, Windows and Mac OS X.
PROJECT MEMBERS
Up to v1.0, while developed at CAIA:
- Sebastian Zander (primary developer)
- Grenville Armitage (project lead)
From v1.0 onwards, see our Sourceforge site.
Projects using TEACUP
The following projects are (or have been) using TEACUP:
- 2016 -- "Implementing Active Queue Management at the home to reduce NBN speed demands" (final year engineering student project)
- 2015-present -- PhD research into content streaming and AQM
- 2015-present -- PhD research into delay-sensitive congestion control algorithms
- 2015-2016 -- "Implementing AQM in FreeBSD" (funded R&D)
- 2015-2016 -- "An integrated social and technical evaluation of household broadband service requirements for educational innovation and the Internet of Things" (funded R&D)
- 2015 -- "Queue management in Home Gateways: Potential impact on IoT traffic flows" (summer internship student project)
- 2015 -- TEAPLOT (animated visualisation of selected TEACUP results)
- 2015 -- "High-availability Internet Gateway for Uptime-Sensitive Medical Telemetry Applications" (final year engineering student project)
- 2014 -- "Alternative Backoff: Achieving Low Latency and High Throughput with ECN and AQM" (collaborative research)
- 2014 -- "Study in TCP Congestion Control Performance In A Data Centre" (funded R&D)
Acknowledgements
TEACUP versions prior to 1.0 were developed as part of a project
funded by Cisco Systems and titled "Study in TCP Congestion Control
Performance In A Data Centre". This was a collaborative effort
between Swinburne University of Technology's Center for Advanced
Internet Archectures and Fred Baker of Cisco Systems.
Starting with TEACUP v1.0, development will be community supported
and publicly hosted at our Sourceforge site.
|