2nd International ICST Conference on Broadband Networks

Research Article

On the use of bandwidth aggregation over heterogeneous last miles

  • @INPROCEEDINGS{10.1109/ICBN.2005.1589789,
        author={Manoj Balakrishnan and Rajesh  Mishra and Ramesh R. Rao},
        title={On the use of bandwidth aggregation over heterogeneous last miles},
        proceedings={2nd International ICST Conference on Broadband Networks},
        publisher={IEEE},
        proceedings_a={BROADNETS},
        year={2006},
        month={2},
        keywords={},
        doi={10.1109/ICBN.2005.1589789}
    }
    
  • Manoj Balakrishnan
    Rajesh Mishra
    Ramesh R. Rao
    Year: 2006
    On the use of bandwidth aggregation over heterogeneous last miles
    BROADNETS
    IEEE
    DOI: 10.1109/ICBN.2005.1589789
Manoj Balakrishnan1,*, Rajesh Mishra1,*, Ramesh R. Rao1,*
  • 1: California Institute of Telecommunications and Information Technology, Department of Electrical and Computer Engineering, University of California at San Diego, CA 92093, USA
*Contact email: bsmanoj@ucsd.edu, ramishra@ucsd.edu, rrao@ucsd.edu

Abstract

The next generation wireless communication devices are expected to be capable of communicating with the best possible last mile networks as well as to utilize multiple networks simultaneously. Existing solutions such as the interoperability mechanisms and the always best connected (ABC) paradigm limit the access of wireless devices to only one, preferably the best possible network. Such schemes, though found to be better than the traditional single interface communication, are limited in their ability to utilize the services in the best possible way. Existing work focuses mainly on network layer or transport layer bandwidth aggregation mechanisms which either need to change the existing TCP protocol or require proxy nodes to perform the bandwidth scheduling process. We, in this paper, propose a new wireless access paradigm for multi-homed hosts based on a session layer bandwidth aggregation mechanism. The major advantages of our solution are the high end-to-end throughput, glitch free transition during both mobility and interface changes, dynamic selection of number of end-to-end paths, and above all our solution can work with existing transport and network layer protocols in today's Internet. In this paper, we provide the architectural and protocol solutions for the proposed scheme and results from extensive simulations and a Linux based implementation