Abstract: A method of determining frame loss between two management points (C, D) in an Ethernet network, in which the management points each transmit frames to each other and each of the two management points transmits to the other, in regular intervals, measurement messages which contain current counts of frames transmitted and received by the respective transmitting management point. At least one of the two management points responds to a received management message to compute from counts of actual packets transmitted and/or received by a given one of the management points the frame loss at the given management point. At least one of the management points computes the frame loss only once in a measurement interval which consists of a multiplicity of the regular intervals and employs in the computation the counts indicated by the measurement message most recently received by the one of the management points.

Abstract: According to an example embodiment, a technique may be provided for managing a label data base of which each entry comprises a data field for insertion into a respective packet in a multicast transmission and a validity field. The technique may include, for example, (a) establishing a fixed size for a cluster of labels, (b) associating with each cluster a pointer to another cluster, (c) establishing a sentinel cluster of which all entries are indicated to be invalid and of which the cluster pointer points to this sentinel cluster, (d) establishing a root table which defines the number of replications of the packet in a multicast group and identifies a first label cluster associated with the multicast group and (e) establishing a chain of clusters of which each cluster which is full of valid entries points to the next cluster in the chain and the last cluster which contains valid entries points to the sentinel cluster.

Abstract: A method of determining frame loss between two management points (C, D) in an Ethernet network, in which the said management points each transmit frames to each other and each of the said two management points transmits in regular intervals to the other measurement messages which contain current counts of frames transmitted and received by the respective transmitting management point. At least one of the said two management points responds to a received management message to compute from counts of actual packets transmitted and/or received by a given one of the management points the frame loss at said given management point. At least one of the management points computes the said frame loss only once in a measurement interval which consists of a multiplicity of said regular intervals and employs in the computation the counts indicated by the measurement message most recently received by said one of the management points.

Abstract: Ethernet ‘operation, administration and management’ (OAM) circuits contain monitoring points that exchange ‘management” messages, particularly connectivity fault management messages and performance monitoring messages. This invention is a search process to identify a receiving monitoring point and also a further process to verify, particularly for continuity check messages, a management endpoint. The search processes use information carried in a message, in a hardware efficient method with predictable search times independent of search table size.

Abstract: Multicast packets that are received on a port of a network device and forwarded to multiple output ports are stored in memory and respective primary control entries which define them are converted to secondary control entries defining multiple unicast packets before a scheduling algorithm is applied. The packets are reconstituted after the application of the scheduling algorithm has been applied to the control entries. For VPLS packets that are received on a single port and replicated multiple times on one or more output ports, a replication database may be used in conjunction with a replication engine to convert the control entry for a received packet into multiple control entries defining unicast packets for each of the destination ports before the scheduling algorithm is applied. This method is applicable to the replication of packets onto a Virtual Private LAN.

Abstract: Multicast packets that are received on a port of a network device and forwarded to multiple output ports are stored in memory and respective primary control entries which define them are converted to secondary control entries defining multiple unicast packets before a scheduling algorithm is applied. The packets are reconstituted after the application of the scheduling algorithm has been applied to the control entries. For VPLS packets that are received on a single port and replicated multiple times on one or more output ports, a replication database may be used in conjunction with a replication engine to convert the control entry for a received packet into multiple control entries defining unicast packets for each of the destination ports before the scheduling algorithm is applied. This method is applicable to the replication of packets onto a Virtual Private LAN.

Abstract: A network switch (10) for a packet-based data communication network, comprises a plurality of ports for the reception and transmission of data and means for establishing a database for controlling the passage of data between the ports. The database comprises a data table (17) for holding data entries each comprising a media access control address and an identification of a port, and a pointer table (16) of which the entries each comprise a network address and an associated pointer to an entry in the said data table. The pointers are accessed by hashing network addresses in received packets.

Abstract: The present invention relates to transparent access to network attached storage devices, configured to any of several protocols, such as SCSI over IP, NAS or NASD. In particular, the present invention provides for using a switch to transparently aggregate storage devices. The switch appears as a virtual storage device. It responds to requests to initiate file sessions and selects one of a plurality of storage devices to participate in the file session. A file session can be handed off to a different storage device. Both the setup and handoff are transparent to the client and its TCP/IP client. The present invention may be practiced either as a method or device. It may provide a virtual storage device or it may aggregate storage devices already attached to a network.

Abstract: A method of operating a network device in a communication system for the transmission of data packets which include network addresses identifying sources and destinations of data, the network device being capable of both bridging and routing decisions and including a forwarding database by means of which a packet including network address data can be forwarded to at least one port and thereby to at least one network path identified by a network address, and packets can be forwarded to at least one port in response to a media access control address. The method comprises establishing a data table which contains entries comprising a network address of an end station to which a packet is destined, a respective media access control address and an identification of at least one port to which the packet will be directed within the device. The table is accessed in response to network addresses and media access control addresses, whereby the same table can be used for both routing and bridging decisions.

Abstract: The present invention relates to transparent access to network attached devices, such as network attached storage devices, Web servers or database servers. In particular, the present invention provides hand-off protocols that are transparent to a network client, with the advantage that physical devices can be added to, replaced on or removed from a network without reconfiguring network clients or applications running applications at levels above the network clients. One aspect the present invention is a method for handing off TCP sessions in a system including a client in communication with a switch and to or more devices. This method includes determining in the first device that a handoff should take place, identifying a second device to take over the session, sending handoff messages to and receiving an acknowledgment from the second device, and reporting the handoff to and receiving an acknowledgment from the switch.

Abstract: A last-in first-out data stack is able to store a plurality (n) of elements wherein the plurality is greater than two. For each of n storage locations in the stack a selector is adapted to receive n elements read out from the stack and a new element. The selector is arranged to make a controllable selection of its received elements. Combinational logic capable of performing logic manipulations on the selection of elements provides a resultant for pushing onto the stack.

Abstract: A network switch comprises a first device which includes a store for data packets, a second device which includes a plurality of ports from which the packets may be transmitted, a plurality of temporary stores each associated with one of the ports, a data bus connecting the first device to the second device, a unit for maintaining a count of units of storage space in respect of each of the ports, a unit for incrementing the respective count on transmission of data from one of the stores, and a unit for decrementing the count in response to dispatch of data from the temporary store in the first device to the second device.

Abstract: A method is described that converts a first flow of data words into a second flow of data words. The first flow of data words has a first data rate and the second flow of data words has a second data rate. The second data rate is greater than the first data rate such that the second flow of data words under-runs. The method also includes transmitting the second flow of data words over a plurality of communication links. A data alignment data structure is transmitted over each of the communication links for each under-run.

Abstract: A method of processing data contained in data packets comprises storing a set of microcode instructions of which some are test instructions prescribing a respective test between a data pattern in a packet and a test pattern; defining by means of bit masks a respective multiplicity of programs each consisting a group of instructions selected from the said set of instructions; providing a particular one of said bit masks; reading out from storage the selected group of instructions in the program defined by said particular one of said bit masks, said selected group including at least one of said test instructions; and executing the instructions in the selected group on data packets. Matches between data patterns in said packets and a test pattern defined by said at least one test instruction may be detected.

Abstract: A method of testing data packets uses a program of selected microcode instructions of which at least one prescribes a comparison between a series of data patterns and a series of test patterns. Data patterns and test patterns are stored in dynamic random access memory. A first group of one of said series of patterns is read out and stored in temporary storage locations. A first group of the other of said series of patterns is read out and stored in temporary storage locations. Each of the patterns in the first group of the other series of patterns is compared, as it is stored, with a respective one of the stored patterns of the first series. The invention renders the testing of data packets more efficient.