1. QoS-Aware Resource Allocation for Slowly Time-Varying Channels InfoCom Department - University of Rome La Sapienza giancola@infocom.uniroma1.it lucadn@newyork.ing.uniroma1.it gaby@acts.ing.uniroma1.it Presented at the 58 th IEEE Vehicular Technology Conference, Orlando, U.S.A., October 2003

2. High level application Physical Channel Data (packets) MAC Medium Access Control Two basic problems in the design of a MAC protocol are: 1)the efficient management of the resource 2)the need for fulfilling QoS requirements despite the unpredictable behaviour of the channel We propose an analytical approach for resource allocation at the MAC level. The resulting algorithm maximizes transmis- sion efficiency by adapting error protection to both channel status and required QoS. Reference Scenario

3. Traffic sources are characterized by two sets of parameters: Tspecs Tspecs – collects parameters describing source traffic activity Qspecs Qspecs – defines QoS requirements Tspecs Qspecs p peak rate r mean rate M max packet size b token buffer Dmax maximum tolerable end-to-end delay F minimum tolerable percen- tage of packets delivered within Dmax MACPDU The MAC protocol works with fixed-size MAC Protocol Data Units (MACPDUs) header payload effective payload FEC size Model Assumptions

4. SOURCE BUFFER Resource Allocation (1/4) CD Two functions are introduced in order to express in analytical terms the trade-off which exists between reserved capacity C and the delay D. Capacity function Required capacity in bps Delay function system delay Maximum number of retransmissions Minimum capacity Round Trip Time

5. Resource Allocation (2/4) MACPDUs FEC Effective Payload In order to evaluate the effect of segmentation on required capacity, the MAC must evaluate the size of required overhead on each MACPDU. Effective Capacity F FEC size can be evaluated by taking into account the QoS parameter F.

6. Resource Allocation (3/4) Target packet loss probability on each MACPDU N R L EFF Given the required QoS, it depends on N R and on L EFF Corrective capability on each MACPDU p b It depends on channel status, i.e. on the BER value p b The value of k detemines the FEC size L FEC L EFF We obtain a new L FEC which can affect the size of the effective payload L EFF We propose an iterative algorithm based on successive approximations. This algorithm is computationally efficient and returns the FEC size which is necessary on each MACPDU.

7. Resource Allocation (4/4) Effective Capacity Required capacity in terms of the number of MACPDUs per frame which are necessary for the application. D F is the frame duration N ARQ is a corrective term due to the ARQ N R Transmission efficiency is maximized by selecting the N R value leading to the minimum number of MACPDUs per frame.

8. Performance in static channels

9. Performance in slowly time-varying channels Percentage of source packets delivered to destination as a function of the receiver speed for a real-time source (circles) and a non-real-time source (crosses). 00.10.20.30.40.50.60.7 90 100 speed of the receiver [m/s] percentage of packets delivered to destination Performance of the proposed algorithm was verified in the case of a slowly time-varying channel. The Jakes channel model was used for characterizing multipath propagation in a generic indoor environment. comparable Performance degradation is observed when the channel coherence time is comparable to the maximum end-to-end delay. In a scenario with high mobility, QoS cannot be guaranteed for real-time applications only.

10. Acknowledgements This work was supported by the European Union under Project No. IST-2000-25197 "Whyless.com - The Open Mobile Access Networks" …special thanks to John Silver for providing the PDF conversion of the poster. Special thanks to all the people in the ACTS lab their contribution in both technical and not-technical issues. And finally…