Ad-Hoc TCP

TRACK: TCP PeRformance in Ad-hoC networKs


Issues: In TRACK, we are interested in addressing the following two design issues of TCP over ad hoc wireless networks:

  • The impact of multihop, shared wireless medium on TCP performance
  • How to achieve robust detection of network congestion in ad hoc networks where packet loss may be induced by congestion, disconnection, channel interference, MAC contention.

Our Approach and Main Findings: Our study of TCP over static ad-hoc networks reveals some interesting results.

  • There exists certain value for TCP congestion window size, at which the TCP throughput tends to be maximized. However, TCP does not operate around this optimal point, but typically grows its window much larger; this leads to decreased throughput and increased packet loss.
  • To better understand this behavior, we further study the characteristics of TCP packet loss. Our results show that, network overload is mainly signified by wireless link congestion. As long as the buffer size at each node/router is reasonable (e.g., larger than 10 packets in typical settings), buffer overflow-induced packet loss is rare and packet drops due to link-layer contention dominate. Link-layer drops offer the first sign for network overload.
  • We further observe that multihop wireless links collectively demonstrate Random Early Detection (RED) like graceful drop behavior, i.e., the drop probability grows gradually as the network load increases, but the current TCP protocol does not adapt well to this built-in drop characteristic.

To ensure robust detection of variant connection behaviors over ad hoc networks, e.g., network congestion, disconnection, out-of-order delivery due to route changes induced by mobility, bursty channel interferences, we explore an approach of collaborative detection based on multiple metrics. The end hosts collect multiple measurement metrics such as delay, short-term throughput, packet loss ratio, packet out-of-order ratio, and these metrics will collectively infer the given connection behavior. We have shown through both analysis and simulations that this collaborative detection approach can greatly improve TCP performance but still maintain TCP friendliness. The reason is that, collaborative detection is able to significantly reduce the Type-II false detection probability (e.g., "congestion" is inferred when the network is not actually congested), but still keep the Type-I false detection probability negligible. Our extensive simlations and real hardware experiments have confirmed the results.

 

Project Members:

  • Students: Zhenghua Fu, Haiyun Luo, Xiaoqiao Meng
  • Faculty: Songwu Lu

Publications:

  • Zhenghua Fu, Xiaoqiao Meng, and Songwu Lu, "How Bad TCP can Perform in Mobile Ad-Hoc Networks," to appear in IEEE ISCC (IEEE Symposium on Computers and Computers) 2002, Italy, July 2002.
  • Zhenghua Fu, Ben, and Songwu Lu, "Design and Implementation of a TCP-Friendly Transport Protocol for Ad Hoc Wireless Networks," to appear in IEEE ICNP'02, 2002.
  • Zhenghua Fu, Petros Zerfos, Haiyun Luo, Songwu Lu, Lixia Zhang, Mario Gerla. The Impact of Multihop Wireless Channel on TCP Throughput and Loss. to appear in IEEE INFOCOM'03, 2002.

 
WiNG