Handoffs in Cellular Wireless Networks: The Daedalus Implementation and Experience

  • Authors:

  • Srinivasan Seshan;Hari Balakrishnan;Randy H. Katz

  • Affiliations:

  • Computer Science Division, Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, CA 94720-1776, U.S.A.;Computer Science Division, Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, CA 94720-1776, U.S.A.;Computer Science Division, Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, CA 94720-1776, U.S.A.

  • Venue:
  • Wireless Personal Communications: An International Journal
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Network protocols in cellular wireless data networks must update routesas a mobile host moves between cells. These routing updates combinedwith some associated state changes are called handoffs. Most currenthandoff schemes in wireless networks result in data loss or largevariations in packet delivery times. Unfortunately, many applications,such as real-time multimedia applications and reliable transportprotocols, adapt to long term estimates of end-to-end delay and loss.Violations and rapid fluctuations of these estimates caused by handoffprocessing often result in degraded performance. For example, lossduring handoff adversely affects TCP performance [4], and highpacket loss and variable delays result in poor real-time multimediaperformance. In this paper, we describe a multicast-based protocol thateliminates data loss and incurs negligible delays during a handoff. Thebasic technique of the algorithm is to anticipate a handoff usingwireless network information in the form of received signal strengthsand to multicast data destined for the mobile host to nearby basestations in advance. This routing, combined with intelligent bufferingtechniques at the base stations, enables very rapid routing updates andeliminates data loss without the use of explicit data forwarding. Wehave implemented this protocol using IP Multicast and Mobile IP-likerouting. In our implementation, handoffs typically take between 8 and15 ms to complete and result in no data loss.