Presentation is loading. Please wait.

Presentation is loading. Please wait.

Packet implementation: discretization Rates Control: or.

Similar presentations


Presentation on theme: "Packet implementation: discretization Rates Control: or."— Presentation transcript:

1 Packet implementation: discretization Rates Control: or

2 Window dynamics: Packet implementation: discretization Equilibrium: ECN based price control: Link: Source: Shift-register to save the last N ECN bits

3 Packet implementation: Source Side

4 Packet implementation: Link Side

5 Ns Simulation: two-way, long-lived traffic 2Gbps, delay=20ms 32 ftps 64ftps 128 ftps 256 ftps 32 ftps 64ftps 128 ftps 256 ftps Duplex links td1 td2 td3 td4 td5 td2 td3 td4 td5 td_i=(i-1)*10ms 2^6, 2^6, 2^7,2^8,2^9 sources started at 0, 20, 40, 60, 80 seconds RTTs 40, 80,120,160, 200ms,link capacity 2Gbps (250pkts/ms) Scenario:

6 Ns Simulation: two-way, long-lived traffic Marking Prob Utilization Estimated Prob Queue Cwnds Rates (pkts) (pkts/sec) New Protocol

7 NewReno/REDNewReno/AdaptiveRED Utilization Queue Cwnds (pkts) Utilization Queue Cwnds (pkts) Ns Simulation: two-way, long-lived traffic Paremeters: Thresh_ 100, maxthresh_ 2500

8 NewReno/VQ Utilization Queue Cwnds (pkts) Utilization Queue Cwnds (pkts) NewReno/PI Ns Simulation: two-way, long-lived traffic Paremeters: qref_=100

9 Ns Simulation: small marking Prob Marking Prob Utilization Estimated Prob Queue Cwnds Rates (pkts) (pkts/sec) New Protocol

10 Power law Pareto law lognormal: log X normally distr., (q,s,m) Ns Simulation: “heavy-tailed” traffic Long/heavy-tailed distributions Crovella data set, 36208 files

11 Ns Simulation: “heavy-tailed” traffic flow size(pkts, log2(x)) Ex: Pareto(100,1.0) Count contribution of the flows packet contribution of the flows 34078 flows 3.33e+7 pkts Median size 200.5pkts Prob(X<x) Log2(Prob[X>x]) Pareto(scale,shape):

12 Marking Prob Utilization Queue Cwnds Scenario: 1024 sources started at [0,10] with RTT 100ms, link capacity 1Gbps (125pkts/ms) Percentage of the sessions Percentage of the packets Flow size (log2(x)) Cumulative Distribution of the flows Ns Simulation: “heavy-tailed” traffic New Protocol

13 Marking Prob Utilization Queue Cwnds (pkts) Percentage of the sessions Percentage of the packets Flow size (log2(x)) Cumulative Distribution of the flows Scenario: 2^6, 2^6, 2^7,2^8,2^9 sources started uniformly in [0,10] secs with RTTs 40, 80, 120,160, 200ms,link capacity 1Gbps (125pkts/ms) Ns Simulation: “heavy-tailed” traffic New Protocol

14 Utilization Queue Cwnds (pkts) Utilization Queue Cwnds (pkts) Ns Simulation: “heavy-tailed” traffic NewReno/REDNewReno/AdaptiveRED Paremeters: Thresh_ 100, maxthresh_ 2500

15 NewReno/VQ NewReno/PI Utilization Queue Cwnds (pkts) Utilization Queue Cwnds (pkts) Ns Simulation: “heavy-tailed” traffic

16 Packet implementation: tricks ? Window management Price estimation: Pacing output Capping the change of the cwnd Penalizing the real queue above a threshold Smoothing the average arrival rate of the queue

17 Conclusion: Equation-based implementation has the desirable performance. --scalable stable, fair, high utilization, small queue --especially for the high bandwidth links Price feedback and estimation is the main obstacle. Improvement may be made to be more efficient. To be done... -- packet-drop and timeout -- optimal parameter set -- new price estimation and transmission scheme -- more practical implementation

18 Packet implementation: Source Side

19 Ns Simulation: “heavy-tail” traffic


Download ppt "Packet implementation: discretization Rates Control: or."

Similar presentations


Ads by Google