Overview

Due to the rapid development of networking technologies in both access and core computer networks, as well as the sudden increase of the Internet population, new and varied types of service-oriented networks have emerged and currently co-exist in the Internet (e.g., CDNs, P2P network, Grid network, and IP-VPN). Under these circumstances, we believe that information concering bandwidth availability in a network path is important in adaptive control of the network.

ImTCP (Inline measurement TCP)

So we introduce a novel mechanism for actively measuring available bandwidth along a network path. In previous active available bandwidth measurement tools sending extra traffic into the network is the common weakness. Against the problem, our new mechanism exploits data packets transmitted in a TCP connection, so that the mechanism does not add probing packets into the network (we call the concept inline netowrk measurement technique). We call a modified version of TCP that incorporates the proposed mechanism ImTCP (Inline measurement TCP). ImTCP is based on TCP Reno, with a modification to the sender program only. The ImTCP sender adjusts the transmission intervals of data packets, then estimates available bandwidth of the network path between sender and receiver utilizing the arrival intervals of ACK packets. Moreover, we propose some applications that utilize the measurement results of ImTCP [7,10,11].

For more information about how ImTCP works, you can read [1,2]. When you want to get the implementation code of ImTCP, jump to ImTCP for Linux or ImTCP for FreeBSD.

ICIM (Interrupt Coalescence-aware Inline Measurement)

In high-speed networks, such as 1-Gbps or higher, bandwidth measurement mechanisms based on packet transmission/arrival intervals, such as packet trains and packet pairs, have some problems.

• Network measurement for large bandwidth requires short packet transmission intervals (e.g, 12 us for a 1-Gbps link), which causes a heavy load on the CPU.
• Network interface cards for high-speed networks usually employ Interrupt Coalescence (IC), which rearranges the arrival intervals of packets and causing bursty transmission, so that the mechanisms utilizing the packet arrival intervals do not work properly.

In order to overcome above problems we introduce ICIM (Interrupt Coalescence-aware Inline Measurement), a new bandwidth measurement approach. In ICIM the TCP sender adjusts the number of packets involved in a burst, rather than adjusting the packet transmission intervals, and estimates the available bandwidth by observing the number of corresponding ACK packets in the burst.

For more information about how ICIM works, you can read [4,5]. When you want to get the implementation code of ICIM, jump to ICIM for FreeBSD.

Papers

About ImTCP

1. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, A New Available Bandwidth Measurement Technique for Service Overlay Networks, in Proceedings of 6th IFIP/IEEE MMNS 2003 E2EMON Workshop, Sept. 2003. [PDF]
2. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, Available Bandwidth Measurement via TCP Connection, in Proceedings of 2nd IEEE/IFIP Workshop on End-to-End Monitoring Techniques and Services (E2EMON 2004), Oct. 2004. [PDF]
3. Tomoaki Tsugawa, Go Hasegawa, and Masayuki Murata, Implementation and evaluation of an inline network measurement algorithm and its application to TCP-based service, in Proceedings of 4th IEEE/IFIP Workshop on End-to-End Monitoring Techniques and Services (E2EMON 2006), Apr. 2006. [PDF]

About ICIM

4. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, ImTCP HighSpeed: Inline Network Measurement for High-Speed Networks, in Proceedings of Passive and Active Measurement Workshop (PAM 2006), Mar. 2006. [PDF]
5. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, ICIM: An Inline Network Measurement Mechanism for Highspeed Networks, in Proceedings of 4th IEEE/IFIP Workshop on End-to-End Monitoring Techniques and Services (E2EMON 2006), Apr. 2006. [PDF]
6. Tomoaki Tsugawa, Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, Inline network measurements: Implementation difficulties and their solutions, in Proceedings of 5th IEEE/IFIP Workshop on End-to-End Monitoring Techniques and Services (E2EMON 2007), May. 2007. [PDF]

Related Papers

7. Tomohito Iguchi, Go Hasegawa, and Masayuki Murata, A New Congestion Control Mechanism of TCP with Inline Network Measurement, in Proceedings of ICOIN 2005, Jan. 2005. [PDF]
8. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, A Merged Inline Measurement Method for Capacity and Available Bandwidth, in Proceedings of Passive and Active Measurement Workshop (PAM 2005), Mar. 2005. [PDF]
9. Cao Le Thanh Man, Go Hasegawa, and Masayuki Murata, An Inline Measurement Method for Capacity of End-to-end Network Path, in Proceedings of 3rd IEEE/IFIP Workshop on End-to-End Monitoring Techniques and Services (E2EMON 2005), May. 2005. [PDF]
10. Tomoaki Tsugawa, Go Hasegawa, and Masayuki Murata, Background TCP data transfer with inline network measurement, in Proceedings of Asia-Pacific Conference on Communications (APCC 2005), Oct. 2005. [PDF]
11. Kana Yamanegi, Go Hasegawa, and Masayuki Murata, Congestion control mechanism of TCP for achieving predictable throughput, in Proceedings of Australian Telecommunication Networks and Applications Conference (ATNAC 2006), Dec. 2006. [PDF]

Implementation Code

We implemented ImTCP, ImTCP-bg (one of applications using the measurement results of ImTCP), and ICIM by modifying the FreeBSD version 4.10 kernel system. We also implemented ImTCP in Linux 2.6.16.21 kernel system. If you try to use ImTCP, ImTCP-bg, and ICIM, download the modules in the following pages and rebuild your kernel system.

• ImTCP, TCP symbiosis, and TCP guarantee for Linux 2.6.16.21 ... Latest version 0.1.1 - 2007/02/21
• ImTCP and ImTCP-bg for FreeBSD 4.10 ... Latest version 1.0.1 - 2005/12/17
• ICIM for FreeBSD 4.10 ... Latest version 1.0.1 - 2006/05/11

If you have some troubles and questions about these mechanisms and implementations, please send me E-mail.