Distributed Control Algorithms for Service Differentiation in Wireless Packet Networks

1. Distributed Control Algorithms for Service Differentiation in Wireless Packet Networks INFOCOM 2001 Michael Barry, Andrew T. Campbell Andras Veres

2. Outline • Introduction • Modified MAC • Virtual MAC & Virtual Source algorithms • Virtual Delay Curves • Conclusions

3. Introduction • Wireless communications are becoming an essential feature of everybody’s life.

4. Introduction(cont’d) • There are two principal approaches to support better than best-effort services for Internet based services in a future wireless network: 1. Conventional circuit switched paradigm 2. Wireless LAN protocols such as IEEE 802.11 (PCF), etc.

5. Introduction(cont’d) • In this paper, we discuss the components of a differentiated services capable wireless network: 1. A distributed DiffServ capable wireless MAC 2. A distributed radio resource monitoring mechanism 3. Localized management of service level specifications and admission control 4. Fast mobility support in the access network

6. Introduction(cont’d) • The proposed modified MAC is based on the IEEE 802.11 radio MAC algorithm for mobile hosts and base stations. • Providing service differentiation solely at the radio interface is insufficient to enable predictable behavior for individual traffic types.

7. Modified MAC • The ideal MAC is adaptive and robust to both internal and external circumstances. • Use Backoff timer to achieve the service differentiation!!

8. Modified MAC(cont’d) • Backoff times are set to a random value in the range [ 0 , CW] * Tslot • By decreasing the CWmax for a service class, the maximum backoff time can be limited during congestion. • This limits the range of congestion control, thus we trade lower delay for increased collision probability, and eventually larger packet loss ratio.

9. Simulation Results of the Modified MAC • In the simulation, the traffic mix consisted of 5 voice traffics and 10 best effort traffics. • For high priority traffic: CWmin : [8, 32] ; CWmax : 64 • For best effort traffic: CWmin : [32, 128] ; CWmax : 1024

10. Simulation Results of the Modified MAC(cont’d)

11. Simulation Results of the Modified MAC(cont’d) • Channel load is increased by adding a new voice, video • and TCP session periodically every 5 seconds {16, 64}

12. Simulation Results of the Modified MAC(cont’d)

13. Virtual MAC & Virtual Source Algorithms • Although the modified MAC ensures effective service differentiation. However, it is not sufficient to ensure that high priority traffic gets better service than best effort traffic. • In most cases applications require absolute and not relative service quality. • The channel status is also important for different type of traffic!!

14. Virtual MAC & Virtual Source Algorithms(cont’d) • In what follows, we present the Virtual MAC and Virtual Source algorithms that are capable of passively observing the radio channel. • This passive monitoring capability allows a mobile host to evaluate the state of the channel, and estimate the level of service it would receive without actually loading the channel.

15. Virtual MAC & Virtual Source Algorithms(cont’d) • The channel can be in one of the three states: throughput limited, delay limited, or not congested. • These algorithms are operate in parallel to the real applications and MAC layer in the host and estimate the service level of the channel.

16. Virtual MAC & Virtual Source Algorithms(cont’d) VS Application Interface buffer TCP/IP VMAC MAC PHY

17. Virtual MAC & Virtual Source Algorithms(cont’d)

18. Virtual MAC & Virtual Source Algorithms(cont’d)

19. Virtual Delay Curves • The delay experienced by an application is made up of several components: packetization delay, delay accumulated in the buffers and the delay caused by the wireless MAC. • For the same data rate: psize* prate = const whereprateis the packet inter-arrival time psizeis the size of the application level packet

20. Virtual Delay Curves(cont’d) • Define d(prate) as the virtual delay curve of an application. • The mobile hosts and the base stations runs VS algorithms with several prate. • Similarly, it can define the virtual delay variance curve v(prate) which calculates the virtual delay variances.

21. Virtual Delay Curves(cont’d) Virtual Delay Curves Virtual Delay Variance Curve

22. Admission Control • Based on the VS and VMAC results, an admission control(AC) algorithm can determine whether the channel can support a new traffic stream or not. • Simulation environment: • In the simulations, new voice sessions is accepted if the estimated average delay less than 10 ms, with the CW value of {32, 64}. (BS): 10 (MH): 100 400m

23. Admission Control(cont’d)

24. Admission Control(cont’d)

25. Conclusions • This paper has shown how service differentiation can be provided in a mobile access network. • We have proposed two passive radio channel monitoring algorithms: VS and VMAC for estimating the achievable level of service without actually loading the channel

26. Discussions • In the case of MHs have packets to transmit simultaneously, and they don’t transmit before then, thus it would be a serious collision and a wrong virtual delay curve for each of the MH which wants to transmit. • The curves may be different with the real channel status because of the virtual detection of the radio. • The high priority traffic should be transmit than low priority traffic rather than the reverse way!!