Abstract: A scalable server architecture for use in implementing scaled media servers capable of simultaneous real-time data stream retrieval for large numbers of subscribers. A scalable server includes a plurality of stream pumping engines each accessing a particular storage device of a storage subsystem, and a server processor which receives retrieval requests from subscribers and directs the stream pumping engines to retrieve the requested data streams. Each of the stream pumping engines may include a storage controller coupled to its corresponding storage device for directing retrieval of the requested stream therefrom, a network controller for supplying the retrieved stream to a client network, and a processor for directing the operation of the storage and network controllers. Each of the stream pumping engines may also include a shared memory accessible by the corresponding stream pumping engine processor and the server processor.

Abstract: A system and method are provided for assigning routing tag bits for routing signals through a Benes network comprising an input stage and an output stage. The input and output stages each comprise a column of 2.times.2 .beta. elements which route an inputted signal to an upper output if the control sequence bit is 0 and to a lower output if the control sequence bit is 1. Each signal inputted to the Benes network is associated with control sequence. For a particular Benes, in a particular control stage of the Benes network, a 0 is assigned to a control sequence bit associated with a signal q.sub.0 received at an upper input of a topmost input stage .beta. element. A 1 is assigned to a control sequence bit associated with a signal q.sub.k received at the same output stage p element as the signal q.sub.0. A 1 is assigned to a control sequence bit associated with a signal q.sub.1 received at a lower input of the topmost input stage .beta. element.

Abstract: A method and apparatus for storage and retrieval of multiple data streams in a multimedia distribution system. A given data stream is separated into a plurality of portions, and the portions are stored in a multi-disk storage system with Y disks each having X zones such that the ith portion of the given stream is stored in zone (i mod X) of disk (i mod Y). The number X of zones per disk and the number Y of disks are selected as relatively prime numbers. The stored data are retrieved using Y independent retrieval schedulers which are circulated among the Y disks over a number of scheduling intervals. Each retrieval scheduler processes multiple requests separated into X groups, with the requests of each group accessing the same disk zone during a given scheduling interval. The retrieval schedulers are also configured such that the retrieval requests of a given retrieval scheduler access the same disk during a given scheduling interval.

Abstract: An inventive switch for transporting information cells without cell contention is described. The switch includes at least one parallel distribution network. Each distribution network includes an N.times.N first routing network for receiving cells at a plurality of input ports, where N equals the number the input ports. Illustratively, the routing network is self-routing and non-blocking, such as a Banyan Network. Connected to the N.times.N network are .rho..sup.k groups of shared buffers for storing the cells routed through the network for a period of time not greater than one cell cycle, where k is incremented from 1 to ?log.sub.2 N/log.sub.2 .rho.!.sup.-1 and .rho. equals a predetermined speed-up factor. In one aspect of this embodiment, the number of shared buffers is simply equal to N/.rho.. To prevent cell contention and cell loss, all of the contentious cells (cells destined for the same output during the same cycle) are stored in the same shared buffer. Connected to the shared buffers are .rho..sup.

Abstract: A method and system are disclosed for scheduling the assignment and writing of cells from cell sources into a outgoing bitstream transmitted from each device of an ATM communications network. The timeslots are organized into fixed length cycles which cycles each have a sequence of N timeslots, where N is an integer >1. Furthermore, each cycle is divided into at least one round comprising a variable length subsequence of the timeslots of the cycle. During each timeslot of a uniform timeslot clock, at least one subset of the sources is identified. Each subset corresponds to a round. One cell from each subset is assigned to, and written into, a respective timeslot of the corresponding round. During each timeslot of the timeslot clock, each of the sources is assigned a priority state depending on how many cells of that source have been previously assigned to timeslots during the current cycle and whether or not that source has a cell available for assignment to a round during that timeslot.