Dynamic Guard Bandwidth Scheme for Wireless Broadband Networks

1. Dynamic Guard Bandwidth Scheme for Wireless Broadband Networks IEEE INFOCOM 2001

2. Outline • Introduction • Framework • Boundary approximation • Estimation functionality based on GPS • Dynamic guard bandwidth adaptation • Simulation Results • Conclusion INFOCOM 2001

3. Introduction • Guard bandwidth / guard channel scheme • Fixed • Dynamic • Proposed Scheme: Dynamic Guard Bandwidth Adaptation. INFOCOM 2001

4. Irregular Cell Boundary INFOCOM 2001

5. Approximating the Cell Boundary • Boundary Approximation Points (BAP). • When a MT makes a handoff request, it is required to report its position and its target handoff cell to the BS. • The region around each BS is divided into M sectors, and the reported handoff request positions within each sector are reduced to a single BAP, associated with a most likely target handoff cell . • The BAPs and their corresponding are stored in a table located at each BS. • The table could be updated based on handoff-request positions collected over a certain period. INFOCOM 2001

6. Estimating the Remaining Time to Handoff / the Target Handoff Cell • The trajectory of each MT is predicted to estimate the BAP that is likely to be closest to the actual handoff request location (Closest BAP, CBAP). • Each MT obtains its own position information at a regular time interval from its GPS receiver, and keeps track of its previous positions over the last (Q-1) intervals. • The current position of the MT is labeled as , and the previous (Q-1) points are labeled as … . • The MT’s direction of travel is estimated using linear regression over these Q points. INFOCOM 2001

7. Estimating the Remaining Time to Handoff / the Target Handoff Cell • Method for the search of the CBAP: • The M BAPs are divided into four quadrants, where points (ag, bg), g = M(i-1)/4, M(i-1)/4 + 1, …, M(i-1)/4 – 1, lie in the ith quadrant. • Once the quadrant containing the CBAP is determined, a bisection search method is performed within that quadrant recursively until we obtain two BAPs that enclose the CBAP, and then we choose the closer BAP as CBAP. • O(log2 M) complexity. INFOCOM 2001

8. Estimating the Remaining Time to Handoff / the Target Handoff Cell INFOCOM 2001

9. Estimating the Remaining Time to Handoff / the Target Handoff Cell INFOCOM 2001

10. Important Points • The location information obtained from the GPS receivers are only used to estimate Tremain and Ctarget. • Actual handoff requests are still initiated based on received signal strength measurements, error rates, interference, and handoff protocols used. • The remaining time to handoff as well as the target handoff cell estimations are to be performed by individual MTs. • Distributed. • The BS would not be overloaded with these computations. INFOCOM 2001

11. Important Points • This scheme has reasonable tolerance for prediction errors. • While the MTs could change direction and speed as it approaches the cell boundary, the target handoff cell prediction is erroneous only if the trajectory of the MT changes so much that it enters a different cell from the one predicted. • Predictions are made periodically and alternative reservations would be attempted if the previous decision has become invalid. INFOCOM 2001

12. Dynamic Guard Bandwidth Adaptation • Adaptation of wireless guard bandwidth: • When a new call request arrives at a BS, the network shall classify it into one of the I traffic classes that is supported. • MTs make periodic predictions of Tremain and Ctarget every . If Tremain is shorter than a threshold time (the remaining time to handoff threshold, RTHT), the MT will inform the target BS about its prediction, and the BS will increase BG accordingly. INFOCOM 2001

13. Dynamic Guard Bandwidth Adaptation INFOCOM 2001

14. Dynamic Guard Bandwidth Adaptation • BG is only increased by a fraction of the actual bandwidth requirement. • Reservation ratio (RR). • For a MT that belongs to traffic class i, with bandwidth requirement Bi, and is anticipated to handoff to neighboring cell j within RTHT, the BG in cell j should be increased by RR(i,j) x Bi. • 0 < RR(i,j) < 1, and it is dynamically adjusted based on measured values of PF(i, j). INFOCOM 2001

15. Dynamic Guard Bandwidth Adaptation • When to decrease of BG: • The MT has entered the cell • The call ends before the handoff • The MT is no longer expected to handoff into the cell. • The BG for the cell is reduced if a handoff is not expected to occur within RTHT consistently over a number of prediction time intervals (the release Reservation Threshold Time, RRTT). • Typically the RRTT is set to a few . INFOCOM 2001

16. Dynamic Guard Bandwidth Adaptation • Adaptation of backbone guard bandwidth • While handoffs into a radio cell could only originate from its neighboring cells, handoff rerouting through a wired backbone link is dependent on network topology and the rerouting scheme used. • With the ability to predict the most likely handoff cell, the links whose BG need to be increased are limited to those between the COS and the target handoff cell. INFOCOM 2001

17. Dynamic Guard Bandwidth Adaptation • Admission control of new calls • A new call of traffic class i is admitted only if INFOCOM 2001

18. Dynamic Guard Bandwidth Adaptation • Admission control of handoff calls • A common BG pool that is shared by different traffic classes would favor those with smaller bandwidth requirements. • A handoff call of traffic class i is only admitted if it satisfies the following conditions: INFOCOM 2001

19. Simulation Results INFOCOM 2001

20. Simulation Results INFOCOM 2001

21. Simulation Results INFOCOM 2001

22. Simulation Results INFOCOM 2001

23. Conclusion • Measurement-based approach. • Suitable for heterogeneous bandwidth requirements. • High scalability due to the distributed operation. INFOCOM 2001