Performance Analysis of Multiplane, Nonblocking ATM Switches.pdfVIP

into a single curve This is in a way consistent with our easily derive the maximum throughput of the switch in observation regarding throughput which leads us to the question conclusion that n is sucient for implementing the The switches under consideration had common out typ e of multiplane switch we describ ed in this section put queues and common or separate input queues For the various values of m we see that m qualies schemes AB and C resp ectively In the case of separate as a go o d choice for m input queues we can avoid any sequencing problems by As far as the cell drop probability is concerned we can forcing cells that b elong to the same trac stream ie write from Equation A where now an input queue VC to b e queued in the same input thus avoiding having is mo deled as a Geom nGeom q queue that them routed through dierent planes In general sepa rate queues can b e used to implement trac streams with K 1 2 n m nm 2 n priorities or with individual trac requirements ie as 2(n )(m ) to guarantee the promised QoS or according to the cells P B 1 K 1 destination n m m 2 n Moreover we demonstrated the tradeo b etween re 2(n )(m ) dundancy and wiring cost schemes A and B whereas the since q 1 2(m ) Note that for fairness ie p erformance is not aected whatso ever In addition we S 2m maintain the same