Mcdata Corporation

Hewlett-Packard Company

David Book

Robert H Grant

Bent Stoevhase

Stephen Trevitt

Robin Purohit

Gregory T Sullivan

A method and apparatus are presented for operating a time slicing shared memory switch. The apparatus includes a bus for receiving a plurality of data frames in a respective plurality of input channels to the switch. A slice crosspoint applies the plurality of data frames to a shared memory in a time sliced manner. The time slice is established for each section of a shared memory to be staggered so that on any clock cycle, one memory portion is being accessed for writing at least some of the data frames and on a next clock cycle the memory portion is accessed for reading at least a portion of the data.

Low latency shared memory switch architecture

The present invention is a fiber channel switch employing a distributed queuing algorithm for interconnecting a plurality of devices (workstations, supercomputer, peripherals) through their associated node ports (N_ports) and employs a fabric having a shared memory coupled to a plurality of fabric ports (F_ports) through a bi-directional bus over which memory addresses, frame data and communications commands are transmitted. Each F_port includes a port controller employing a distributed queuing algorithm associated with a control network for communicating commands between the ports related to when and where frame transfers should be made, wherein the bi-directional bus provides an independent data network for access to the shared memory such that frames can be transferred to and from the shared memory in response to port controller commands.

Fibre channel switch employing distributed queuing

The present invention is a fiber channel switch employing a distributed queuing algorithm for interconnecting a plurality of devices (workstations, supercomputer, peripherals) through their associated node ports (N.sub.-- ports) and employs a fabric having a shared memory coupled to a plurality of fabric ports (F.sub.-- ports) through a bi-directional bus over which memory addresses, frame data and communications commands are transmitted. Each F.sub.-- port includes a port controller employing a distributed queuing algorithm associated with a control network for communicating commands between the ports related to when and where frame transfers should be made, wherein the bi-directional bus provides an independent data network for access to the shared memory such that frames can be transferred to and from the shared memory in response to port controller commands.

Fiber channel switch employing distributed queuing

A fiber optic switch interconnects fiber optic channels so that a fiber optic network can be implemented. Channel modules provide ports (p1-pi) for connection of the fiber optic channels. Each channel module has a receive memory for temporarily storing incoming data frames from the fiber optic channels associated therewith. A switch module having a data distribution network interconnects each of the channel modules and permits ultimate connection of a source channel to a destination channel. A path allocation system, which controls the switch module, allocates the data paths between the channels. The path allocation system has a scheduler which maintains a destination queue (Q.sub.p1 -Q.sub.pi) for each of the channels, a sentry which determines when a new data frame is ready to be routed, and an arbitrator which arbitrates port availability and which grants transfer requests proposed by the scheduler.

High performance path allocation system and method for a fiber optic

A frame time monitoring system tracks the time in which data frames reside within a fiber optic switch for a fiber optic network. The network switch transfers data frames from source ports to destination ports. The frame time monitoring system comprises a digital signal processor (DSP), which is configured by a software program to implement a plurality of timers relative to frames to be routed through the switch from a source port to a destination port. The processor operates as an incrementer and is configured to output a series of sequential timer states corresponding to each particular frame. The timer states are generally indicative of the amount of time in which the frame has resided in the switch. A logic network of logic gates is connected to the processor to receive and interpret the timer states. The logic network has frame busy (FBSY) and delete mechanisms for determining elapse of respective FBSY and delete time periods based upon the timer states. The FBSY and delete mechanisms generate respective FBSY and delete signals after the elapse of the periods, which can vary for optimization reasons depending upon frame class and type. Thus, in the foregoing configuration, a processor is utilized as a timing incrementer and logical decisions are allocated to the logic network, resulting in an optimum balance between hardware and software so as to minimize cost, space requirements, and complexity, and maximize frame tracking resolution.

High performance frame time monitoring system and method for a fiber

A fiber optic switch interconnects fiber optic channels so that a fiber optic network can be implemented. Channel modules provide ports (p1-pi) for connection of the fiber optic channels. Each channel module has a receive memory for temporarily storing incoming data frames from the fiber optic channels associated therewith. A switch module having a data distribution network interconnects each of the channel modules and permits ultimate connection of a source channel to a destination channel. A path allocation system, which controls the switch module, allocates the data paths between the channels. The path allocation system has a scheduler which maintains a destination queue (Q.sub.p1 -Q.sub.pi) for each of the channels, a sentry which determines when a new data frame is ready to be routed, and an arbitrator which arbitrates port availability and which grants transfer requests proposed by the scheduler. Significantly, the arbitrator is equipped with fairness insurance mechanisms for insuring that transfer requests are always timely and efficiently serviced.

High performance path allocation system and method with fairness

A fiber optic switch interconnects ports (p1-pi) for connection with respective fiber optic channels so that a fiber optic network is realized. Channel modules provide the ports. Each channel module has a port intelligence mechanism for each port and a memory interface system for temporarily storing data passing to and from the ports. A switch module having a main distribution network, an intermix distribution network, and a control distribution network interconnects the memory interface systems and permits exchange of data among the ports and memory interface systems. A path allocation system controls the switch module and allocates the data paths therethrough. The path allocation system has a scheduler which maintains a destination queue (Q.sub.p1 -Q.sub.pi) for each of the ports. The destination queues are implemented with a double link list in a single memory configuration so that a separate queue structure in hardware is not necessary. Moreover, the scheduler is implemented with a digital signal processor (DSP) with on-chip memory so that the queues are implemented within the on-chip memory and can be accessed at high speed.

Growing community of inventors

Thornhill, Canada

David Book