# CROSS LAYER DESIGN CMPT 820: Multimedia Systems Kaushik Choudhary.

## Presentation on theme: "CROSS LAYER DESIGN CMPT 820: Multimedia Systems Kaushik Choudhary."— Presentation transcript:

CROSS LAYER DESIGN CMPT 820: Multimedia Systems Kaushik Choudhary

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Problem Definition Define problem as an optimization to select a joint strategy across multiple OSI layers (PHY, MAC and APP) One-hop wireless network where network and transport layers play less important roles.

Problem Definition Autonomous Wireless Stations (WSTAs) = M, Available wireless resources = R ∈ R +, Channel condition experienced by WSTA i = SNR i, Video source characteristics = ξ i, Current state information (private information) of WSTA i = x i = (SNR i, ξ i ) Resource allocation = T( R ) = [t 1,……,t M ] ∈ R + M Time allocation by resource coordinator to WSTA i = t i (0 <= t i <= t SI ) (PCF or other contention)

Problem Definition Given the above constraints, the cross layer design problem may be formulated as an optimization with objectives like maximizing goodput or minimizing consumed power etc. If s i is the cross layer strategy available to WSTA i from a set S i then given x i and t i, s i will result in a utility u i (t i,s i,x i )

Problem Definition The optimal cross layer strategy can be found by maximizing the video quality in terms of perceived quality or PSNR Mathematically, s i opt = arg max u i (t i,s i,x i ) s i ∈ S i such that Delay(t i,s i,x i ) <= Delay i max

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Motivation for Optimal Strategy The above functions u i (t i,s i,x i ) and Delay(t i,s i,x i ) are non-deterministic, non-linear, complex and multi-variate optimization. The strategies to solve them must take into account different practical considerations like buffer sizes, modulation schemes etc. and should be procedurally grouped and ordered for cross layer optimization.

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Categorization of Cross Layer Solutions Possible solutions based on the order in which cross layer optimization is performed: – Top-down approach – Bottom-up approach – Application-centric approach – MAC-centric approach – Integrated approach

Top-down approach Higher layer protocols optimize their parameters and strategies at the next lower layer. Very widely used for example in systems where APP dictates MAC parameters and MAC selects optimal PHY parameters.

Bottom-up approach Lower layers try to insulate higher layers from losses and bandwidth variations. Not optimal due to incurred delays and throughput reductions.

Application-centric approach APP layer optimizes parameters of lower layers one at a time in either top-down or bottom-up manner. Not optimal since APP layer operates on slower timescales and coarser data granularities.

MAC-centric approach MAC layer decides which APP layer packets should be transmitted with which delay along with selecting PHY layer parameters. MAC layer is unable to perform adaptive source channel coding

Integrated approach Strategies are determined jointly across various protocols. Complex and introduces delays.

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

MAC Layer Retransmission Limit Adaptation To maximize video quality, minimize the MAC packet loss rate (PLR). MAC packet losses occur due to: – Link erasures – Buffer overflows Define a strategy to optimally select retransmission limit R that minimizes the overall MAC packet loss.

Fig 1: MAC PLR under fixed- and RTRO-based retransmission strategies.

MAC Layer Retransmission Limit Adaptation From Fig 1 we note that the optimal R is located at the point where p B (R) = p L (R) (intersection) Thus, optimal R = arg min R |p B (R) - p L (R)| (3)

MAC Layer Retransmission Limit Adaptation Li et. al. [2] performed empirical analysis using M/G/1 queuing model and proposed real-time retransmission limit optimization (RTRO) algorithm: 1.The network queue and the MAC layer monitor the overflow rate p B (R) and the packet error rate p L (R). 2.If p B p L, then R should be decreased

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Joint Application-MAC Cross Layer Optimization By associating different retransmission limits to different priority packets the MAC-layer RTRO optimization can be jointly optimized by the APP layer.

Joint Application-MAC Cross Layer Optimization Tolerable MAC packet loss rates of all video layers = P V = [P V1 P V2 … P VN ] Video quality = Q To maximize Q Unequal Error Protection (UEP) must be provided. To provide UEP multiple priority queues are maintained with a common absolute Priority- Queuing (PQ) discipline.

Joint Application-MAC Cross Layer Optimization

The above model can be further extended to include a multiqueue system based on which a systematic retry-limit configuration method for MAC can be determined to optimize video quality.

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Impact of Cross Layer Strategies Deployed StrategiesVisual Score No optimization at MAC and application1.4 MAC-layer optimization (RTRO)1.9 Application layer optimization3.8 Joint application-MAC cross-layer optimization 4.6 Table 1: Subjective video quality experiment. Very Annoying AnnoyingSlightly annoying Perceptible but not annoying Imperceptib le 12345 Table 2: Decoding the visual scores in Table 1.

Outline Problem Definition Motivation for Optimal Strategy Categorization of Cross Layer Solutions MAC Layer Retransmission Limit Adaptation Joint Application-MAC Cross Layer Optimization Impact of Cross Layer Strategies Conclusion

Cross layer design can be modeled as a multivariate optimization problem expressed in the form of various strategies to maximize perceived quality and improve user experience. Cross layer solutions can be categorized into various approaches based on a layer dictating strategies and parameters for other layers. MAC-layer optimization yields improved video quality. MAC-APP layer optimization performs even better.

References 1.P. A. Chou and M. van der Schaar. “Multimedia over IP and Wireless Networks”, Academic Press, ISBN 10: 0-12-088480-1, pp. 351-360, 2007. 2.Q. Li and M. van der Schaar. “Providing Adaptive QoS to Layered Video over Wireless Local Area Networks through Real-Time Retry Limit Adaptation,” IEEE Trans. on Multimedia, vol. 6, no. 2, pp. 278–290, April 2004.