Download presentation
Presentation is loading. Please wait.
1
Bufferless multiplexer
© Jörg Liebeherr, 2010
2
Internet Service Provider
A small internet service provider (ISP) sells internet access at rate R to its customers The ISP has leased a link with capacity C to get access to the Internet (P Mbps) Question: What is the maximum number of customers (Nmax) that can be supported on the link? © Jörg Liebeherr
3
Bufferless Multiplexer
Assumptions: Switch at ISP has no buffer. Traffic is dropped when rate from DSL customers exceeds We only look at upstream traffic from DSL customers to Internet If all users are transmitting all the time: But customers are not always sending traffic. © Jörg Liebeherr
4
ON-OFF Sources Each customer generates traffic at rate R and transmits packets of size L. The customer waits until a full packet is available and then transmits at the subscribed rate P. This results in an ON-OFF pattern of transmission ON Phase: Send at rate P OFF Phase: Wait until a full packet has been assembled ON ON ON © Jörg Liebeherr
5
ON-OFF Source ON ON ON Arrival time T1 T2 © Jörg Liebeherr
6
Multiple ON-OFF Sources
Arrival time © Jörg Liebeherr
7
Multiple ON-OFF Sources
Worst-case ON ON ON ON ON ON ON ON ON Arrival time © Jörg Liebeherr
8
CLT: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Computes P_loss at bufferless multiplexer % for Greedy ON-OFF Sources %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % CLT %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% N=[1:1:1000]; % number of flows C=100; % link capacity P=1; % ON rate ("peak") rho = 0.1; % average rate epsilon=0.01; subplot(2,1,1); Ploss = 1- normcdf(C, N*rho, sqrt(N*(P*rho - rho * rho ))); semilogy(Ploss,'b'); xlabel('Number of flows (N)'); ylabel('P_{loss}'); axis([ ^-16 1]); grid on title('Application of CLT'); © Jörg Liebeherr
9
Chernoff Bound: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Computes P_loss at bufferless multiplexer % for Greedy ON-OFF Sources %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Chernoff Bound %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% N=[1:10:1000]; a=C./N; l=a./P.*log(a/rho.*(P - rho)./(P-a)) - log((P-rho)./(P-a)); Ploss=exp(-N.*l); subplot(2,1,2); semilogy(N,Ploss,'b'); xlabel('Number of flows (N)'); ylabel('P_{loss}'); axis([ ^-16 1]); grid on; title('Application of Chernoff Bound'); © Jörg Liebeherr
10
Effective bandwidth Slotted system Single source:
In each slot, source transmits with Prob. p If source transmits, it sends at rate R Bk={0,R} transmission in slot k An are the transmissions in timer interval [0,n] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Effective bandwidth for Bernoulli On Off Source %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% p=0.4 % probability of transmit R=1 % peak rate C=0.75 % Capacity theta=[0:.1:100] % Theta semilogx(theta, log(1-p+p*exp(R*theta))./theta,theta,C, [ ], [0 1] ) © Jörg Liebeherr
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.