Agere Systems Inc.

John T Musacchio

Jean C Walrand

Gaurav Agarwal

Jeonghoon Mo

Roy Thomas Myers, Jr

Shyam Parekh

Chung Kuang Chin

Roy T Myers

Roughly described, a packet switching fabric contains a separate queue scheduler for each combination of an input module and a fabric output port. The schedulers may also be specific to a single class of service. Each queue scheduler schedules its packets without regard to state of other input queues and without regard to packets destined for other output ports. In an aspect, the fabric manages per-flow bandwidth utilization of output port bandwidth capacity by monitoring the same and asserting backpressure toward the queue scheduler for any thread that is exceeding its bandwidth allocation. In another aspect, a switching fabric uses leaky buckets to apply backpressure in response to overutilization of downstream port capacity by particular subflows. In another aspect, a switching fabric includes a cascaded backpressure scheme.

Backpressure mechanism for switching fabric

A striping algorithm selects a route on which to transmit each next data segment, in dependence upon relative channel loading so far, taking account of multicast. Input modules can keep a channel loading history for each route it has, and can update its history for each route that a data segment follows through the fabric. In an embodiment, the input module transmits each data segment toward an i'th intermediate stage module, where i minimizesq(i,a(G),c)+q(i,b(G),c)+ . . . +q(i,k(G),c),where q(i, j, c) indicates the number of bytes of data sent, during a given prior time period, from the input module to each j'th one of the output modules via each i'th one of the intermediate stage modules, and a(G), b(G), . . . , and k(G) are the output module(s) in the multicast group G to which the data segment is destined.

Striping algorithm for switching fabric

In one embodiment, data segments routed through a switch fabric are identified as being either a first or a second type. First and second counters are maintained for a destination node to characterize the transmission of data segments of the corresponding type within the switch fabric. For each first-type time period, all transmitted data segments are identified as being the first type. For each second-type time period, all transmitted data segments are identified as being the second type. The first- and second-type time periods are interleaved. The first counter is reset after termination of each first-type time period and the second counter is reset after termination of each second-type time period to correct for a possible error in characterizing the transmission of the data segments within the switch fabric during the corresponding time period.

Characterizing transmission of data segments within a switch fabric using multiple counters for each destination node

Roughly described, a striping algorithm selects a route on which to transmit each next data segment, pseudorandomly from among a subset of eligible routes, the subset being chosen in dependence upon relative channel loading so far. Preferably each ingress node to a switching system chooses an outgoing route for each given next data segment, according to a pseudorandom algorithm, from among a respective given subset containing only those routes via which the amount of data sent from the ingress node during a respective prior time period is no greater than an average of the amount of data sent from the ingress node via any of its outgoing routes during the same prior time period.

Growing community of inventors

Berkeley, CA, United States of America

John T Musacchio