A New Handover Decision Management Algorithm in Wireless Overlay Networks

1. IEEE Global Telecommunications Conference, 2006. A New Handover Decision Management Algorithm in Wireless Overlay Networks Yildirim, M.; Alagoz, F. Computer Engineering Department Boğaziçi University, Istanbul, Turkey

2. Outline • Introduction • Related Work • A New Handover Decision Scheme for Wireless Overlay Networks • Performance Evaluation • Conclusion

3. Introduction • Wireless Overlay Networks (WONs) • combine different wireless technologies • to provide wide-area coverage, best possible bandwidth and lowest possible latency • have a hierarchical structure with different levels • Higher levels usually have a larger coverage but lower bandwidth • high-mobility users • lower levels contain high-bandwidth cells covering a small area. • low-mobility users

4. Introduction (cont.) • Challenges • High handover latency impairs both applications and protocols transferring information. • Power consumption of mobile devices • No comparable signal strength available to aid the decision of the “best”' network for vertical handovers • During a handover, upper-layer applications are interested in network conditions • such as available bandwidth, delay and user preferences

5. Introduction (cont.) • an efficient handover decision system in WONs must • consider many factors • such as network's load, cost, coverage and data rate, minimize MN's power consumption and • make accurate and timely decisions • This paper focuses on building a new handover decision manager (HDM) for MN • performs different tasks for different handover types • to minimize the number of unnecessary handovers and power consumption as well as maximize throughput. • decides handovers by being aware of network conditions and application requirements.

6. Horizontal Handovers • the MN switches between cells of the same network overlay • The MN does not need to change the network interface it is using. • mandatory since the MN cannot continue its communication without performing it.

7. Vertical Handovers • the MN moves from a cell in one overlay to a cell in a different overlay in WON • not strictly mandatory • a result of the tradeoff between coverage and bandwidth • the MN's network interface changes • upward vertical handover • moves to an WON layer with larger coverage area but lower bandwidth • This happens as the MN becomes highly mobile • and the lower layer becomes unreachable. • downward vertical handover • moves to a WON layer with greater bandwidth but smaller coverage area

8. Related Work • Zhang et al. uses different methods for switching to 802.11 WLAN and GPRS WWAN [9] • WLANWWAN: periodically monitors RSS • the MN switch to WWAN very quickly as the WLAN becomes unavailable • WWANWLAN: uses the network allocation vector (NAV) in the MAC layer to estimate network conditions (e.g. available bandwidth and access delay) • the NAV busy state can reflect the media's busy state • the MN decide handover by being aware of network conditions. • not suitable for different types of wireless networks not using NAV. • does not consider many network-related factors • such as the coverage, cost and data rate. • Some approaches use SNR together with conventional two-level thresholding scheme [10], [11], [12] [9] Q. Zhang, Z. G. C. Guo, and W. Zhu, “Efficient mobility management for vertical handoff between wwan and wlan,” IEEE Commun. Mag., 2003.[10] M. M. Buddhikot, G. P. Chandranmenon, S. J. Han, Y. W. Lee, S. Miller, and L. Salgarelli, “Integration of 802.11 and third-generation wireless data networks,” in INFOCOM, 2003.

9. Related Work (cont.) • Chen et al. presents an adaptive scheme which makes use of utility functions [13]. • wireless network j is composed of several normalized factors fi,j multiplied by their weight or importance wi. • calculated and compared with the utilities of networks for a stability period. • changes according to handover latency and the utility ratios • consumes more power relative to other methods • neglects the computational advantage of performing different tasks for upward and downward vertical handovers • Making a wrong estimation (of the handover latency) before even doing a handover may lead to erroneous results [13] W.-T. Chen, J.-C. Liu, and H.-K. Huang, “An adaptive scheme for vertical handoff in wireless overlay networks.,” in ICPADS, 2004.

10. A New Handover Decision Scheme for Wireless Overlay Networks • Downward vertical handovers must be carried out cautiously by being aware of the network conditions. • the objective of the handover is to improve QoS rather than maintaining connectivity. • A timely and accurate decision mechanism for upward vertical handovers. • a result of the necessity to maintain connectivity

11. the threshold timer value is determined by the application

12. Performance Evaluation • The simulation environment is a 7.5 km x 3 km area and consists of 61 WLAN AP with 300 m of coverage and 2 GSM BS with 5 km of coverage. • By C++ • The channel propagation model is taken from [15] and RSS at distance d is calculated according to • Pt : the transmitted power • Xσ : a zero-mean Gaussian random variable with standard deviation σ • PL(d): the path loss at distance d and • S: the path loss constant depending on the propagation environment and • n: the path loss exponent [15] A. Majlesi and B. H. Khalaj, “An adaptive fuzzy logic based handoff algorithm for interworking between wlans and mobile networks,” in The 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, Sep. 2002. [17] “Linksys wag54g v2 (wireless-g adsl gateway) datasheet,” 2005. http://www-uk.linksys.com. [18] “Contela bts specification.” Contela, Inc, 2006. http://www.contela.com/product/btsspec.pdf.

13. Performance Evaluation (cont.) • 500 MNs moving randomly for 3 hours within the map borders • move a random amount of distance in the chosen direction • northwest, northeast, southwest and southeast • Such a trajectory generation asserts the MNs to move larger distances and come across a greater number of candidate wireless networks • either do web browsing or video conferencing • With reference to the video traces analysed in [19], disruption continuing for 5 s will degrade QoS

14. does not consider the MN's speed

15. determined according to instruction execution times of various operations listed in [20].[20] “Mc68000 16-bit mp user manual.” Motorola Semicon. Prod. Inc., 1980. MNs move with 10 m/s

16. an MN performs handovers more freely without considering many factors (which takes times), the resulting throughput is relatively higher

17. Conclusion • The proposed scheme gives priority to lower layer candidate networks and performs different tasks for deciding different types of handovers. • The simulation results show that the proposed HDM decreases the number of unnecessary handovers as well as save power. • Besides, the average throughput of the MNs does not decrease significantly as they move faster. • planning to combine HDM with a WON architecture based on Subnet Agents controlling each overlay. • will be able to determine the signaling between network components and optimize the handover

18. comments • Simple and smart handover decision algorithm • simulation未考慮cross subnet/domain handover • 忽略某wireless access technology的tolerant moving speed • 考慮到vehicle上的MR, power consumption並非最重要的考量