Spyros Psychis, Polychronis Koutsakis and Michael Paterakis Electronics & Computer Engrg. Dept. &

1. “Multiplexing Live Video Streams & Voice with Data over a High Capacity Packet Switched Wireless Network” Spyros Psychis, Polychronis Koutsakis and Michael Paterakis Electronics & Computer Engrg. Dept. & Telecommunication Systems Institute Technical University of Crete Chania, Greece

2. Introduction • Wireless Networks: currently allow users to experience services that until now only the wire-line networks provided. • The main concern remains: how to extend the broadband frontier to the end user given the constraints of the wireless media.

3. Introduction (cont.) • Well designed MAC protocols are needed in order to: • Maximize system’s capacity • Integrate the different classes of traffic • Satisfy the diverse and sometimes contradictory QoS requirements of each traffic class

4. Concept • Within the cell, spatially dispersed MTs share a radio channel that connects them to a fixed base station or a wireless hotspot. • Base station allocates the channel resources, delivers scheduling and feedback information and serves as an interface to the MSC, which provides access to the fixed network infrastructure and the Internet.

5. Channel Structure Frame Information • Uplink channel time is divided into time frames of equal length. • Each frame consists of a request interval and an information interval. The request intervals consist of slots, which are subdivided into two mini-slots, and each mini-slot accommodates exactly one, fixed length, request packet.

6. Channel Structure (cont.) • The size of the request interval per channel frame is variable. The number of the request slots varies depending on the number of video terminals that “live” in the system. • Frame duration is selected such that an active Voice Terminal generates a new voice packet (of ATM size) at the beginning of each channel frame • Only Voice and Data Terms use the request intervals in order to transmit their request to the BS. • Voice Terms are given priority to request transmissions • When all Voice request have been transmitted the Data request transmission follows

7. Voice Traffic • We assume that the Voice Terminals (VTs) are equipped with Voice Activity Detectors (VAD). The output of the VAD is modeled by a two-state discrete time Markov chain. • VTs only require channel access during talkspurt. (when in talkspurt, they generate traffic @ 32 Kbps). • The upper delay limit that a voice packet can suffer is assumed equal to 40 ms. • Maximum Pdrop=0,01

8. Data Traffic • Data traffic model is based on statistics collected on email usage from a University and Research Network. • The pdf for the length of the data msgs was found to be well approximated by the Cauchy (0.8,1) distribution. • The msg inter-arrival time distribution is exponential. • An upper bound on the average data msg delay equal to 2 secs is assumed (tolerable delay for email msg transmission).

9. Video Terminals • Video Terminals are streaming actual MPEG-4 streams (steady cam) encoded @ 25 fps (one Video Frame every 40 ms). • The mean bit rate is 400 Kbps, the peak rate is 2 Mbps, and the standard deviation of the bit rate is equal to 434 Kbps. • Maximum transmission delay for Video Packets is assumed to be equal to 40 ms. • Maximum Pdrop = 0,0001

10. BS Scheduling and Terminal Actions (Video) • The video terminals envoy their slot requests to the BS by transmitting them within the header of the first packet of their current video frame. • The BS allocates channel resources at the end of the corresponding request interval. • Video terminals have the highest priority in acquiring the slots they demand. • If a full allocation is possible, the BS then proceeds to the allocation. • Otherwise, the BS grants to the video users as many of the slots they requested as possible (partial allocation).

11. BS scheduling and Terminal Actions (Voice) • Voice terminals, which have successfully transmitted their request packets to the BS, do not acquire all the available (after the servicing of video terminals) information slots in the frame. • BS allocates a slot to each requesting voice terminal with a probability p*. When there are no video terminals in the system, p* is set equal to 1.

12. BS scheduling and Terminal actions (Data) • Data Terminals follow the same allocation procedure with Voice Terminals after the Voice contention period is over. • They can receive only one information slot per channel frame and can keep it just as long as they have data packets to transmit. • We do not use preemption of data reservations but still voice users are given priority both in slots and in allocation policy..

13. Performance Evaluation - Simulations • Channel Rate = 20 Mbps • Initially the system was simulated under all possible video loads from 0 to 22 streams with no active Data Terminals in order to specify the boundaries where the number of R-slots must change. • Each run simulated one hour of actual network activity (300005 channel frames). • These simulation runs helped us choose the value of p* (which turns out that must be equal to 0.1 in order to get close to optimal results for all the examined cases of video load).

14. Estimation Runs

15. Simulation (2nd Set) • During the second set of experiments we simulated the system under Voice and Data traffic only (Rslots=30). • We tried to accommodate as many Voice Terminals as possible under the constraint that the data throughput should match the data load value (steady state stable operation). • Crude theoretical estimation of the voice capacity: [(Total number of information slots – estimated data throughput) / probability of talkspurt]

16. Simulation Results (2nd Set)

17. Simulation (3rd Set) During the third set of experiments we simulated the system under certain mixtures of Voice, Video and Data traffic.

18. Conclusions and Contributions of this Work • The goal was to design efficient MAC scheduling mechanisms in order to satisfy the diverse nature of the traffic types that the network must accommodate and the contradictory QoS requirements of each traffic type. • Simulation results show that the proposed mechanism achieves high aggregate channel throughput in all cases of traffic load, while preserving the Quality of Service (QoS) requirements of each traffic type.

19. Ideas for Future Work • The evaluation of the proposed mechanisms when used over a wireless error prone channel. • Investigating the case in which different media encoding techniques together with less strict QoS requirements for the time sensitive traffic are used. • Incorporating the proposed mechanisms in a wider system framework (MAC and BS scheduling schemes together with a content placement scheme throughout the wireless and wireline network).