TCP Testing

From Wiki

This site contains testing results of high-speed protocols. BIC and CUBIC official page

1 Testbed
2 Intra Protocol Testing
- 2.1 400Mbps
3 RTT Fairness Testing
- 3.1 400Mbps
4 TCP Friendliness Testing
- 4.1 400Mbps

Testbed

We use a dumbbell topology of dummynet routers where each end points consists of a set of Dell Linux servers dedicated to high-speed TCP variant flows and background traffic. Background traffic is generated by using a modification of a web-traffic generator, called Surge[1] and Iperf. We modified the traffic generator to generate a wider range of flow sizes in order to increase variability in cross traffic because medium size flows tend to fully execute the slow start and increase the variability in available bandwidth. The RTT of each background flow is set based on an exponential distribution found in [2]. The maximum bandwidth of the bottleneck router is set to 400 Mbps. The same amount of background traffic is pushed into forward and backward directions of the dumbbell. We use the drop-tail router at the bottleneck.

[1] Paul Barford and Mark Crovella, "Generating Representative Web Workloads for Network and Server Performance Evaluation", ACM SIGMETRICS 1998.

[2] Jay Aikat, Jasleen Kaur, F. Donelson Smith, and Kevin Jeffay, "Variability in TCP Roundtrip Times", ACM IMC 2003.

[edit]

Intra Protocol Testing

[edit]

400Mbps

Configuration
Protocols	CUBICv2.2, HTCP, and HSTCP
Kernel Version	Linux 2.6.25-rc3 (Senders & Receivers)
Bandwidth	400Mbps
Bottleneck Buffer	100% BDP
# Flows ( 2 flows )	2 flows of the same protocol with the same RTT (one for each machine)
Tested RTTs	8ms, 16ms, 46ms, 86ms, 166ms, and 246ms
Background Traffic	5% short-lived pareto traffic

Buffer Size	RTT / Protocols	CUBIC	HTCP	HSTCP
2Mbytes	8ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	16ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	46ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	86ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	166ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	246ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5

[edit]

RTT Fairness Testing

[edit]

400Mbps

Configuration
Protocols	CUBICv2.2, HTCP, and HSTCP
Kernel Version	Linux 2.6.25-rc3 (Senders & Receivers)
Bandwidth	400Mbps
Bottleneck Buffer	100% BDP
# Flows ( 2 flows )	one flow has a fixed RTT of 246ms and the other flow has a RTT ranges from 8ms to 246ms (one for each machine)
Tested RTTs	8ms, 16ms, 46ms, 86ms, 166ms, and 246ms
Background Traffic	5% short-lived pareto traffic

Buffer Size	RTT / Protocols	CUBIC	HTCP	HSTCP
2Mbytes	8ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	16ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	46ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	86ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	166ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	246ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5

[edit]

TCP Friendliness Testing

[edit]

400Mbps

Configuration
Protocols	CUBICv2.2, HTCP, and HSTCP
Kernel Version	Linux 2.6.25-rc3 (Senders & Receivers)
Bandwidth	400Mbps
Bottleneck Buffer	100% BDP
# Flows ( 2 flows )	one flow is fixed to TCP-SACK and the other flow is one of three high-speed protocols (CUBIC, HTCP, and HSTCP). Two flows have the same RTT. (one for each machine)
Tested RTTs	8ms, 16ms, 46ms, 86ms, 166ms, and 246ms
Background Traffic	5% short-lived pareto traffic

Buffer Size	RTT / Protocols	CUBIC	HTCP	HSTCP
2Mbytes	8ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	16ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	46ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	86ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	166ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5
	246ms	1 2 3 4 5	1 2 3 4 5	1 2 3 4 5