doi:10.1016/S0140-3664(02)00134-2
Copyright © 2002 Elsevier Science B.V. All rights reserved.
Exploiting the adaptive properties of a probing device for TCP in heterogenous networks
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
V. Tsaoussidis
, 
and A. Lahanas
College of Computer Science, Northeastern University, Boston, MA 02115, USA
Received 26 November 2001;
revised 2 May 2002;
accepted 2 May 2002. ;
Available online 20 July 2002.
Abstract
End-to-end protocols lack the functionality to efficiently adjust their error control strategies to the distinct characteristics of network environments and to specific constraints of communicating devices. In the context of heterogeneous networks, error control needs to be responsive to the nature of the errors spanning a conservative-through-to-aggressive behavioral spectrum. In the context of mobile, battery-powered devices, the recovery strategy should yield good performance with a minimal transmission effort. We exploit the potential of a probing device to implement an adaptive error control strategy efficiently by shaping data transmission to be assorted with the distinctive characteristics of the underlying wired or wireless network components. We graft our mechanism onto standard TCP; we present encouraging experimental results with wired, wireless and mobile networks.
Author Keywords: Heterogeneous networks; TCP-probing; Adaptive properties; Wired/wireless networks; energy savings; Transmission Control Protocol
Fig. 1. TCP-probing recovery strategy.
Fig. 2. Hand-off layout with NS-2 simulator.
Fig. 3. Task completion time performance of TCP-probing and Reno (see Table A1).
Fig. 4. Task completion time performance with 50 ms propagation delay (see Table A2).
Fig. 5. Task completion time performance with 100 ms propagation delay (see Table A3).
Fig. 6. Overhead of protocols with 100 ms propagation delay (see Table A3).
Fig. 7. Overhead of Reno and TCP-probing (see Table A1).
Fig. 8. Overhead of protocols with 50 ms propagation delay (see Table A2).
Fig. 9. Task completion time performance with congestion (see Table A4).
Fig. 10. Task completion time performance with 50 ms propagation delay and congestion (see Table A5).
Fig. 11. Overhead of protocols with congestion (see Table A4).
Fig. 12. Overhead of protocols with 50 ms propagation delay and congestion (see Table A5).
Fig. 13. Task completion time distribution of the protocols. The average value is presented in Table A1).
Fig. 14. Task completion time with small file transfer (300KB, see Table A6).
Fig. 15. Task completion time with small file transfer (300KB) and 50 ms propagation delay (see Table A7).
Fig. 16. Goodput vs Throughput. Base-stations distance: 200 m.
Fig. 17. Goodput vs Throughput. Base-stations distance: 300 m.
Fig. 18. Goodput vs Throughput. Base-stations distance: 400 m.
Fig. 19. Goodput vs Propagation delay. Moving speed 16 m/s.
Fig. 20. Goodput vs Propagation delay. Moving speed 20 m/s.
Fig. 21. Overhead: Simulator. Base-stations distance 200 m.
Fig. 22. Overhead: Simulator. Base-stations distance 300 m.
Fig. 23. Overhead: Simulator. Base-stations distance 400 m.
Fig. 24. Impact of handoff on goodput and overhead. Base-stations distance 300 m. y-Axis presents the difference between the goodput and overhead of TCP-probing and TCP-Reno.
Fig. 25. Impact of handoff on goodput and overhead. Base-stations distance 400 m. y-Axis presents the difference between the goodput and overhead of TCP-probing and TCP-Reno.
Fig. 26. Fairness over a 10 Mbps link.
Table A1. Performance of Reno and TCP-probing
Table A2. Performance of protocols with propagation delay 50 ms
Table A3. Performance of protocols with propagation delay 100 ms
Table A4. Performance of protocols with random congestion
Table A5. Performance of protocols with random congestion. Propagation delay 50 ms
Table A6. Performance of protocols with small file transfer (300KB) and congestion
Table A7. Performance of protocols with small file transfer (300KB) and congestion. Propagation delay 50 ms
Table A8. Performance of protocols with handoffs (PER 15%)
Table A9. Performance of protocols with handoffs of different PER
Abstract
Purchase PDF (547 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (0)
Purchase PDF (375 K)
