WO1986002511A1 - Load balancing for packet switching nodes - Google Patents

Load balancing for packet switching nodes Download PDF

Info

Publication number
WO1986002511A1
WO1986002511A1 PCT/US1985/001839 US8501839W WO8602511A1 WO 1986002511 A1 WO1986002511 A1 WO 1986002511A1 US 8501839 W US8501839 W US 8501839W WO 8602511 A1 WO8602511 A1 WO 8602511A1
Authority
WO
WIPO (PCT)
Prior art keywords
load balancing
data packets
output port
routing tag
tag signals
Prior art date
Application number
PCT/US1985/001839
Other languages
French (fr)
Inventor
Robert J. Mcmillen
Original Assignee
Hughes Aircraft Company
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hughes Aircraft Company filed Critical Hughes Aircraft Company
Priority to DE8585905022T priority Critical patent/DE3574659D1/en
Priority to KR1019860700370A priority patent/KR900006792B1/en
Publication of WO1986002511A1 publication Critical patent/WO1986002511A1/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/54Store-and-forward switching systems 
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic control in data switching networks
    • H04L47/10Flow control; Congestion control
    • H04L47/12Avoiding congestion; Recovering from congestion
    • H04L47/125Avoiding congestion; Recovering from congestion by balancing the load, e.g. traffic engineering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00Digital computers in general; Data processing equipment in general
    • G06F15/16Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
    • G06F15/163Interprocessor communication
    • G06F15/173Interprocessor communication using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
    • G06F15/17356Indirect interconnection networks
    • G06F15/17368Indirect interconnection networks non hierarchical topologies
    • G06F15/17393Indirect interconnection networks non hierarchical topologies having multistage networks, e.g. broadcasting scattering, gathering, hot spot contention, combining/decombining
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic control in data switching networks
    • H04L47/10Flow control; Congestion control
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic control in data switching networks
    • H04L47/10Flow control; Congestion control
    • H04L47/11Identifying congestion
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/25Routing or path finding in a switch fabric
    • H04L49/253Routing or path finding in a switch fabric using establishment or release of connections between ports
    • H04L49/254Centralised controller, i.e. arbitration or scheduling
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/20Support for services
    • H04L49/205Quality of Service based
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/30Peripheral units, e.g. input or output ports
    • H04L49/3009Header conversion, routing tables or routing tags
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/40Constructional details, e.g. power supply, mechanical construction or backplane

Definitions

  • the present invention relates generally to packet switching nodes employed in multi-processor and parallel computer systems, and the like, and more particularly to a load balancing circuit arrangement for use in such packet switching nodes which redistributes incoming data packets to its output ports for more efficient processing.
  • SIMD single instruction stream
  • MIMD multiple instruction stream
  • a SIMD computer typically comprises a control unit, N processors, N memory modules and an interconnection network.
  • the control unit broadcasts instructions to all of the processors, and all active processors execute the same instruction at the same time. Each active processor executes the instruction on data in its own associated memory module.
  • the interconnection network provides a communications facility for the processors and memory modules.
  • a MIMD computer typically comprises N processors and N memories, and each processor can execute an independent instruction stream. Each of the processors may communicate to any other processor. Similar interconnection networks may be employed in the MIMD computer.
  • interconnection networks may be employed to interconnect processors and memories employed in either type of computer system. These interconnection networks include delta networks, omega networks, indirect binary n-cube networks, flip networks, cube networks and banyan networks, for example.
  • Packet switching involves sending one or more words of data at time through the system.
  • a multiple queue packet switching node is described in a presently copending patent application entitled "Packet Switched Multiple Queue NxM Switch Node and Processing Method," invented by R. J. McMillen, and assigned to the assignee of the present invention.
  • This patent application discloses a packet switching node which processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are to be applied and transfers these packets to those output ports.
  • the packet switching node comprises a plurality of input ports and a plurality of output ports.
  • a plurality of queue selectors are individually coupled to corresponding ones of the plurality of input ports.
  • Each of the plurality of queue selectors are adapted to route data packets applied to each of the input ports in accordance with the output port destination of the data packets.
  • a plurality of queue sets are individually coupled to corresponding ones of the plurality of queue selectors.
  • Each of the queue sets comprise a plurality of queues for storing and forwarding data packets applied thereto as a function of output port destination.
  • a plurality of output arbitrators are individually coupled between corresponding ones of the plurality of output ports and the respective queue of each of the queue sets which store and forward data packets whose destinations are the corresponding output port.
  • the output arbitrators are adapted to transfer the data packets stored in the queues to the corresponding output port in accordance with a predetermined priority arbitration scheme.
  • This packet switching node processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are to be applied.
  • the packet switching node comprises a plurality of input ports and output ports with a multiport memory coupled therebetween.
  • the memory has a predetermined number of memory locations available for storage of data packets applied to each of the input ports.
  • Control logic is coupled to the input and output ports and the multiport memory which controls the storage of data packets in the memory. The control logic also controls the routing of the data packets to the output ports in accordance with the routing tag signals.
  • this invention comprises an N x M switch node that accepts data packets at any of N input ports and routes each to any of M output ports. The output selected is determined by the routing tag signal in the packet.
  • the control logic is designed so that the data packets are effectively sorted according to their desired output port destination. Arbitration logic randomly, in a statistical sense, chooses among any data packets that are directed to the same output port. The algorithm implemented by the arbitration logic is designed so that data packets will not wait indefinitely to be routed from the switch node.
  • processor 1 applies data packets to input port 1. Assume that all output ports of the node are connected to identical execution units. Processor 1 assigns routing tag signals to each of the data packets transmitted thereby. If processor 1 continually assigns routing tag signals corresponding to output port 1, for example, then the remaining output ports are not used, and the execution units connected thereto are not used. This clearly lessens the efficiency of the node.
  • the present invention provides for a load balancing circuit arrangement that improves the efficiency and performance of existing packet switching nodes.
  • the load balancing circuit arrangement of the present invention comprises a logic circuit for use with a packet switching node that processes applied data packets that contain routing tag signals indicative of the output port destinations to which the data packets are addressed.
  • the load balancing circuit arrangement comprises a load balancing circuit coupled to the packet switching node which monitors the output port addresses of the applied data packets. In addition, the load balancing circuit monitors the number of data packets addressed to each of the output ports. The load balancing circuit is adapted to generate new routing tag signals identifying output port addresses which redistribute the output port load.
  • the load balancing circuit arrangement includes a tag selection circuit coupled to the load balancing circuit and the packet switching node which selectively replaces the routing tag signals of the applied data packets with the new routing tag signals in order to redistribute and balance the output port load.
  • One embodiment of the present invention provides for a load balancing circuit arrangement for use with a multiple queue packet switching node that distributes data packets to the least full queues. This operation re-routes the data packets to output ports that are being infrequently used and hence provides for more efficient processing.
  • a second embodiment of the load balancing circuit is provided for use with a multiport memory packet switching node.
  • the switching node comprises queue selection logic coupled to a plurality of queues which route the applied data packets to the desired output port destinations.
  • the load balancing circuit arrangement for use with the multiple queue packet switching embodiment comprises the load balancing circuit coupled between a plurality of queues and their associated queue selection logic. The load balancing circuit monitors the number of data packets stored in each of the queues.
  • the load balancing circuit is coupled to the tag selection circuit which provides input signals to the queue selection logic that replace the routing tag signals of the data packet being processed with new routing tag signals.
  • the new routing tag signals are representative of the queue which has the fewest packets stored therein. This processing redistributes applied data packets to the least full queue and hence redistributes the output port loading of the circuit.
  • the switching node comprises a multiport memory and control logic which similarly route applied data packets to the desired output ports.
  • the load balancing circuit arrangement for use in the multiport memory packet switching embodiment comprises the load balancing circuit coupled to the control logic which monitors the number and destination of the data packets stored in the multiport memory.
  • the load balancing circuit arrangement provides input signals to the control logic which replace the routing tag signals with new routing tag signals which redistribute the applied data packets to the output ports to achieve balanced output port loading.
  • the load balancing circuit arrangement comprises a load balancing circuit that monitors the output port addresses of the applied data packets and the number of data packets addressed to each of the output ports. This load balancing circuit generates the new routing tag signals.
  • a tag selection circuit coupled to the load balancing circuit selectively applys the new routing tag signals to the switching node in response to enabling signals applied thereto.
  • the load balancing circuit comprises a minimum index circuit for generating the new routing tag signals and an adder circuit coupled to the mimimum index circuit for combining the new routing tag signals with offset signals which modify the new routing tag signals in order to implement a predetermined output port priority scheme.
  • a switching node employing the load balancing circuit arrangement of the present invention may be used to construct a multistage interconnection networks which have several capabilities. The first has the ability to distribute incoming data packets over a pool of shared resources. These resources are typically identical execution or memory units connected to the output ports of the node. The second employs an extra node stage at the input of the network. Enabling the load balancing circuit in the switch nodes of the newly added stage balances the load throughout the entire multistage network.
  • interconnection networks employing the present invention may be used to connect telephone systems that communicate both data and voice information by way of data packets.
  • Fig. 1 illustrates a block diagram of an embodiment of a load balancing circuit in accordance with the principles of the present invention incorporated into one channel of a multiple-queue packet switching node
  • Fig. 2 illustrates a more detailed block diagram of the load balancing circuit of Fig. 1;
  • Figs. 3a and 3b comprise a detailed schematic of the logic contained in the minimum index circuit of the load balancing circuit of Fig. 2;
  • Figs. 4a, 4b and 4c show detailed schematics illustrating the logic contained in the modulo four adder of the load balancing circuit of Fig. 2 for threepriority conditions;
  • Figs. 5a and b comprise a detailed block diagram of a multiport memory packet switching node without load balancing;
  • Fig. 6 illustrates the load balancing circuit of the present invention coupled to one channel of the multiport memory packet switching node of Fig. 5;
  • Fig. 7 illustrates a minimum index circuit for use in the load balancing circuit of Fig. 6.
  • FIG. 1 a block diagram of one embodiment of a load balancing circuit arrangement 20 in accordance with the principles of the present invention is shown.
  • the load balancing circuit arrangement 20 is shown incorporated into one channel of a multiple queue packet switching node 21.
  • This load balancing circuit arrangement 20 is made part of a single stage packet switching node 21.
  • the packet switching node 21 processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are addressed.
  • One channel of the packet switching node 21 comprises an input port 22 and a plurality of output ports 26.
  • a queue set 23 is coupled to the input port 22.
  • the queue set 23 comprises a plurality of queues which individually process and store data packets whose destinations are a particular one of the output ports 26.
  • a queue selector 24 is coupled between the input port 22 and each of the plurality of queues in the queue set 23. The queue selector 24 is adapted to route applied data packets arriving at the input port 22 to corresponding ones of the queues in accordance with the addresses contained in the routing tag signals.
  • An output arbitrator 25 is coupled between each output port and respective queues of the queue set. The output arbitrator 25 processes data packets whose destinations are the output port coupled thereto and applies the data packets stored in each of the respective queues to that output port in accordance with a predetermined priority arbitration scheme.
  • the load balancing circuit arrangement 20 is coupled between each of the queues of the queue set 23 and the queue selector 24.
  • the load balancing circuit arrangement 20 monitors the number of data packets stored in each of the queues.
  • the load balancing circuit arrangement 20 provides input signals to the queue selector 24 which replace the routing tag signals of the data packets currently being processed. These input signals are representative of the queue in the queue set 23 which has the fewest packets stored therein.
  • the load balancing circuit arrangement 20 comprises a load balancing circuit 28 coupled to the queues 23 of the packet switching node 20.
  • the load balancing circuit 28 monitors the number of data packets addressed to each of the output ports 26 by examining the number of packets in the queue 23 associated with each output port 26.
  • the load balancing circuit 28 is adapted to generate new routing tag signals identifying output pert addresses which redistribute the output port load.
  • a tag selection circuit 27 is coupled between the load balancing circuit 28 and the queue selector 24 which is adapted to selectively replace the routing tag signals of the applied data packets with the new routing tag signals in order to redistribute and balance the output port load.
  • the load balancing circuit 28 may comprise a minimum index circuit coupled to an adder circuit. The minimum index circuit is adapted to generate the new routing tag signals and the adder circuit is adapted to combine the new routing tag signals with offset signals which modify the new routing tag signals in order to implement a predetermined output port priority scheme.
  • the tag selection circuit 27 may comprise a two to one multiplexer, or the like, which is adapted to selectively choose between two input signals for output to the queue selector 24.
  • the multiple queue packet switching node 21 employing the load balancing circuit arrangement 20 functions in the following manner.
  • a process controller or the like, is coupled to the input port 22.
  • This process controller generates data and instructions which are sent to functionally identical execution units individually coupled to the output ports 26.
  • the data and instructions are sent in data packets which are processed by the packet switching node 21.
  • the switching node 21 provides a communications link between the process controller and the execution units.
  • Each data packet includes a routing tag signal identifying the particular output port 26, and hence the particular execution unit, to which the packet is to be sent.
  • the data packet also includes data that is to be operated upon in that execution unit.
  • a data packet enters the switching node 21 and the queue selector 22 reads the routing tag signal. It then routes the data packet to the appropriate queue of the queue set 23 which processes signals whose destinations are the particular output port 26 corresponding to the routing tag signal.
  • each of the output ports are connected to functionally identical execution units. Therefore, it is not particularly important where the instructions contained in the data packets get executed, but that they are executed in a timely manner. Since all of the execution units are functionally identical, routing tag signals may be changed in order to more efficiently utilize the execution units.
  • the load balancing circuit arrangement 20 produces the desired re-routing of the data packets to accomplish more efficient operation of the switching node 21 and the computer system in general.
  • the load balancing circuit 20 processes count signals generated by each of the queues 23 in the queue set 24.
  • the count signals indicate the number of data packets stored in a particular queue.
  • the load balancing circuit arrangement 20 then analyzes the count signals to determine which queue has the least number of data packets stored therein. Once this has been determined, the routing tag signal in the data packet being processed by the queue selection logic 23 is changed to identify the least full queue.
  • Fig. 2 shows a more detailed block diagram of the load balancing circuit 28.
  • the load balancing circuit 28 is comprised of the minimum index circuit 30 and en adder circuit 31.
  • the load balancing circuit 28 is shown adapted for use with a node which comprises four output ports, hence it employs four sets of inputs to the minimum index circuit 30, identfied as L[0] to L[3].
  • Each set of inputs to the minimum index circuit 30 is individually coupled to count circuitry in the queues which counts the number of data packets currently being stored and processed by a particular queue.
  • the output of the minimum index circuit 30 is coupled to a modulo M adder 31 (modulo 4 in Fig. 2) which adds an offset to the index produced.
  • the load balancing circuit 28 implements an algorithm which determines the least full queue in the queue set. A typical algorithm is presented below.
  • INPUT SCALAR OFFSET 2.
  • INPUT ARRAY L[M]
  • ADDR: (minimum + OFFSET) MODULO M
  • This algorithm scans the inputs L[M] from low order to high, looking for the smallest value. In the case of a tie, the value is retained, thus implementing a particular priority scheme.
  • This priority scheme is as follows. Since there can be a tie between two or more queues for the lowest number, a default priority is assigned. Let L[0] be the first input to the minimum index circuit 30 and let L[M-1] be the last input. In the case of a tie, L[0] has the highest priority and L[M-1] has the lowest.
  • the queue i data packet couut is connected to L[0]
  • the queue with the next highest priority is connected to L[1] and so on.
  • the binary representation of i is connected to the OFFSET input to the modulo M adder 31.
  • N the load balancing circuits are required. If all circuits gave top priority to the lower numbered outputs, there would be a bias in the overall operation of of the switching node favoring these outputs. The distribution of the load over the outputs would not be uniform. To illustrate how the load balancing circuit
  • the load balancing circuit 28 comprises two parts, the minimum index circuit 30 and the modulo four adder 31 which combines the output of the minimum index circuit 30 with the OFFSET value.
  • a modified algorithm is required to implement the minimum index circuit only, and is presented below.
  • This algorithm has been programmed into a computer to generate a truth table with eight boolean input variables and two boolean output variables. This truth table was then made the input to a program which generated the following "sum-of-product" form logic equations. These equations correspond directly to a two level nand gate implementation of the minimum index circuit 30.
  • FIG. 3 A detailed schematic of the logic diagram representing a specific embodiment of the minimum index circuit 30 is shown in Fig. 3.
  • Fig. 3 is comprised of Figs. 3a and 3b.
  • Figs. 3a and 3b should be placed abutting one another and joined along the line of separation in order to obtain the complete drawing.
  • Figs. 4a, 4b and 4c show the circuits to be used when the OFFSET is 1, 2 and 3, respectively. A detailed discussion of the components and wiring of the circuit of Figs. 3 and 4 will be dispensed with.
  • the load balancing circuit arrangement of the present invention may also be incorporated into a multiport memory packet switching node.
  • Fig. 5 illustrates a portion of a four input, four output multiport memory packet switching node without load balancing. This circuit is identical to the circuit of Fig. 3 of the presently copending multiport memory patent application cited hereinabove.
  • Fig. 5 shows control logic 40 of the multiport memory packet switching node.
  • the control logic comprises four input port logic sections 41a-d coupled to four output port logic sections 42a-d.
  • handshaking signal lines 45 and status update control logic 46 are included in Fig. 5.
  • FIG. 6 there is shown a portion of one channel of the multiport memory packet switching node of Fig. 5 incorporating a load balancing circuit arrangement 20' of the present invention.
  • the discussion of this embodiment is undertaken with reference to the presently copending patent application entitled "Packet Switched Multiport Memory NxM Switch Node and Processing Method" cited hereinabove with particular reference to Fig. 3 thereof.
  • Fig. 6 shows a portion of the control logic 40 which includes an input port logic section 41a coupled to four output port logic sections 42a-d.
  • the signal lines illustrated in Fig. 6 of this application correspond to signal lines described in Fig. 3 of the copending patent application.
  • the load balancing circuit arrangement 20' has inputs coupled to request output lines REQ0, REQ1, REQ2 and REQ3 of the input port logic section 41. Outputs of the load balancing circuit arrangement 20' are coupled to a latch 45 which is used to ensure correct signal timing and then to the enabling circuit 27.
  • a load balancing circuit 28' comprises a minimum index circuit 30' and the adder 31.
  • the minimum index circuit 30' for use in this embodiment is shown in more detail in Fig. 7.
  • the adder circuit 31 is substantially the same as hereinabove described with reference to the load balancing circuit arrangement for the multiple queue node.
  • the load balancing circuit may be used to automatically redistribute data packets to external execution units coupled to a multiple queue or multiport memory switching node. This redistribution results in more efficient operation of the node and computer system employing networks incorporating the load balancing circuit.

Abstract

A load balancing circuit arrangement for use with a packet switching node. The packet switching node processes applied data packets containing routing tag signals indicative of the output port destinations to which the data packets are addressed, and routes these packets to the identified output ports. The present invention a load balancing circuit coupled to the packet switching node which monitors the output port addresses of the applied data packets and monitors the number of data packets addressed to each of the output ports. The load balancing circuit is adapted to generate new routing tag signals identifying output port addresses which redistribute the output port load. The load balancing circuit arrangement includes a tag selection circuit coupled to the load balancing circuit and the packet switching node which selectively replaces the routing tag signals of the applied data packets with the new routing tag signals in order to redistribute and balance the output port load. The load balancing circuit comprises a minimum index circuit for generating the new routing tag signals and an adder circuit coupled thereto. The minimum index circuit combines the new routing tag signals with offset signals that modify the new routing tag signals in order to implement a predetermined output port priority scheme. The load balancing circuit arrangement may be employed in both multiple queue and multiport memory packet switching nodes employed in computer or telephone communications applications.

Description

LOAD BALANCING
FOR PACKET SWITCHING NODES
BACKGROUND OF THE DISCLOSURE
The present invention relates generally to packet switching nodes employed in multi-processor and parallel computer systems, and the like, and more particularly to a load balancing circuit arrangement for use in such packet switching nodes which redistributes incoming data packets to its output ports for more efficient processing.
One developing area of computer technology involves the design and development of large-scale, multi-processor-based distributed and parallel computer systems. Typical of these classes of computer systems and architectural approaches are the single instruction stream, multiple data stream (SIMD) computer architecture and the multiple instruction stream, multiple data stream (MIMD) computer architecture.
A SIMD computer typically comprises a control unit, N processors, N memory modules and an interconnection network. The control unit broadcasts instructions to all of the processors, and all active processors execute the same instruction at the same time. Each active processor executes the instruction on data in its own associated memory module. The interconnection network provides a communications facility for the processors and memory modules. A MIMD computer typically comprises N processors and N memories, and each processor can execute an independent instruction stream. Each of the processors may communicate to any other processor. Similar interconnection networks may be employed in the MIMD computer.
Various interconnection networks may be employed to interconnect processors and memories employed in either type of computer system. These interconnection networks include delta networks, omega networks, indirect binary n-cube networks, flip networks, cube networks and banyan networks, for example.
The above-cited networks are discused in some detail in the following publications: "LSI implementation of modular interconnection networks for MIMD machines," 1980 Int'l. Conf. Parallel Processing, Aug.
1980, pp. 161-162; "Analysis and simulation of buffered delta networks," IEEE Trans. Computers, Vol. C-30, pp. 273-282, April 1981; "Processor-memory interconnections for multiprocessors," 6th Annual Int'l. Symp. Computer Architecture, April 1979, pp. 168-177; "Design and implementation of the banyan interconnection network in TRAC," AFIPS 1980 Nat'l. Computer Conf., June 1980, pp. 643-653; "The multistage cube: a versatile interconnection network," Computer. Vol. 14, pp. 65-76, Dec. 1981; "The hybrid cube network," Distributed Data Acquisition. Computing and Control Synp., Dec. 1980, pp. 11-22; and "Performance and implementation of 4x4 switching nodes in an interconnection network for PASM," 1981 Int'l Conf. on Parallel Processing, Aug.
1981, pp. 229-233.
Several types of data switching techniques may be employed to transfer data in SIMD and MIMD computers, and the like, including packet switching, message switching, time-division circuit switching or space-division circuit switching. Packet switching involves sending one or more words of data at time through the system.
A multiple queue packet switching node is described in a presently copending patent application entitled "Packet Switched Multiple Queue NxM Switch Node and Processing Method," invented by R. J. McMillen, and assigned to the assignee of the present invention. This patent application discloses a packet switching node which processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are to be applied and transfers these packets to those output ports.
The packet switching node comprises a plurality of input ports and a plurality of output ports. A plurality of queue selectors are individually coupled to corresponding ones of the plurality of input ports. Each of the plurality of queue selectors are adapted to route data packets applied to each of the input ports in accordance with the output port destination of the data packets.
A plurality of queue sets are individually coupled to corresponding ones of the plurality of queue selectors. Each of the queue sets comprise a plurality of queues for storing and forwarding data packets applied thereto as a function of output port destination. A plurality of output arbitrators are individually coupled between corresponding ones of the plurality of output ports and the respective queue of each of the queue sets which store and forward data packets whose destinations are the corresponding output port. The output arbitrators are adapted to transfer the data packets stored in the queues to the corresponding output port in accordance with a predetermined priority arbitration scheme. Another related presently co-pending patent application is entitled "Packet Switched Multiport Memory NxM Switch Node and Processing Method," invented by R. J. McMillen and A. Rosman, and assigned to the assignee of the present invention. This packet switching node processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are to be applied. The packet switching node comprises a plurality of input ports and output ports with a multiport memory coupled therebetween. The memory has a predetermined number of memory locations available for storage of data packets applied to each of the input ports. Control logic is coupled to the input and output ports and the multiport memory which controls the storage of data packets in the memory. The control logic also controls the routing of the data packets to the output ports in accordance with the routing tag signals.
In general, this invention comprises an N x M switch node that accepts data packets at any of N input ports and routes each to any of M output ports. The output selected is determined by the routing tag signal in the packet. The control logic is designed so that the data packets are effectively sorted according to their desired output port destination. Arbitration logic randomly, in a statistical sense, chooses among any data packets that are directed to the same output port. The algorithm implemented by the arbitration logic is designed so that data packets will not wait indefinitely to be routed from the switch node.
However, although both of these nodes improve upon the performance and throughput of conventional packet switching designs, their performance can be improved. An example will illustrate a typical problem. Assume that processor 1 applies data packets to input port 1. Assume that all output ports of the node are connected to identical execution units. Processor 1 assigns routing tag signals to each of the data packets transmitted thereby. If processor 1 continually assigns routing tag signals corresponding to output port 1, for example, then the remaining output ports are not used, and the execution units connected thereto are not used. This clearly lessens the efficiency of the node.
SUMMARY OF THE INVENTION
In order to overcome the limitations of prior art switching node designs, the present invention provides for a load balancing circuit arrangement that improves the efficiency and performance of existing packet switching nodes. In general, the load balancing circuit arrangement of the present invention comprises a logic circuit for use with a packet switching node that processes applied data packets that contain routing tag signals indicative of the output port destinations to which the data packets are addressed.
The load balancing circuit arrangement comprises a load balancing circuit coupled to the packet switching node which monitors the output port addresses of the applied data packets. In addition, the load balancing circuit monitors the number of data packets addressed to each of the output ports. The load balancing circuit is adapted to generate new routing tag signals identifying output port addresses which redistribute the output port load. The load balancing circuit arrangement includes a tag selection circuit coupled to the load balancing circuit and the packet switching node which selectively replaces the routing tag signals of the applied data packets with the new routing tag signals in order to redistribute and balance the output port load.
One embodiment of the present invention provides for a load balancing circuit arrangement for use with a multiple queue packet switching node that distributes data packets to the least full queues. This operation re-routes the data packets to output ports that are being infrequently used and hence provides for more efficient processing. A second embodiment of the load balancing circuit is provided for use with a multiport memory packet switching node. In the first embodiment, the switching node comprises queue selection logic coupled to a plurality of queues which route the applied data packets to the desired output port destinations. The load balancing circuit arrangement for use with the multiple queue packet switching embodiment comprises the load balancing circuit coupled between a plurality of queues and their associated queue selection logic. The load balancing circuit monitors the number of data packets stored in each of the queues. The load balancing circuit is coupled to the tag selection circuit which provides input signals to the queue selection logic that replace the routing tag signals of the data packet being processed with new routing tag signals. The new routing tag signals are representative of the queue which has the fewest packets stored therein. This processing redistributes applied data packets to the least full queue and hence redistributes the output port loading of the circuit.
In the second embodiment, the switching node comprises a multiport memory and control logic which similarly route applied data packets to the desired output ports. The load balancing circuit arrangement for use in the multiport memory packet switching embodiment comprises the load balancing circuit coupled to the control logic which monitors the number and destination of the data packets stored in the multiport memory. The load balancing circuit arrangement provides input signals to the control logic which replace the routing tag signals with new routing tag signals which redistribute the applied data packets to the output ports to achieve balanced output port loading. The load balancing circuit arrangement comprises a load balancing circuit that monitors the output port addresses of the applied data packets and the number of data packets addressed to each of the output ports. This load balancing circuit generates the new routing tag signals. A tag selection circuit coupled to the load balancing circuit selectively applys the new routing tag signals to the switching node in response to enabling signals applied thereto.
The load balancing circuit comprises a minimum index circuit for generating the new routing tag signals and an adder circuit coupled to the mimimum index circuit for combining the new routing tag signals with offset signals which modify the new routing tag signals in order to implement a predetermined output port priority scheme. A switching node employing the load balancing circuit arrangement of the present invention may be used to construct a multistage interconnection networks which have several capabilities. The first has the ability to distribute incoming data packets over a pool of shared resources. These resources are typically identical execution or memory units connected to the output ports of the node. The second employs an extra node stage at the input of the network. Enabling the load balancing circuit in the switch nodes of the newly added stage balances the load throughout the entire multistage network.
Although the present invention is discussed with reference to its use with computer systems and architectures, it is not limited to this application. The present invention may also be used in applications involving the communications field. In particular, interconnection networks employing the present invention may be used to connect telephone systems that communicate both data and voice information by way of data packets.
BRIEF DESCRIPTION OF THE DRAWINGS
The various objects and features of the present invention may be more readily understood with reference to the following detailed description taken in conjunction with the accompanying drawings, wherein like reference numerals designate like structural elements, and in which: Fig. 1 illustrates a block diagram of an embodiment of a load balancing circuit in accordance with the principles of the present invention incorporated into one channel of a multiple-queue packet switching node; Fig. 2 illustrates a more detailed block diagram of the load balancing circuit of Fig. 1;
Figs. 3a and 3b comprise a detailed schematic of the logic contained in the minimum index circuit of the load balancing circuit of Fig. 2; Figs. 4a, 4b and 4c show detailed schematics illustrating the logic contained in the modulo four adder of the load balancing circuit of Fig. 2 for threepriority conditions; Figs. 5a and b comprise a detailed block diagram of a multiport memory packet switching node without load balancing;
Fig. 6 illustrates the load balancing circuit of the present invention coupled to one channel of the multiport memory packet switching node of Fig. 5; and
Fig. 7 illustrates a minimum index circuit for use in the load balancing circuit of Fig. 6.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to Fig. 1, a block diagram of one embodiment of a load balancing circuit arrangement 20 in accordance with the principles of the present invention is shown. The load balancing circuit arrangement 20 is shown incorporated into one channel of a multiple queue packet switching node 21. This load balancing circuit arrangement 20 is made part of a single stage packet switching node 21. The packet switching node 21 processes applied data packets containing routing tag signals indicative of the output port destination to which the data packets are addressed.
One channel of the packet switching node 21 comprises an input port 22 and a plurality of output ports 26. A queue set 23 is coupled to the input port 22. The queue set 23 comprises a plurality of queues which individually process and store data packets whose destinations are a particular one of the output ports 26. A queue selector 24 is coupled between the input port 22 and each of the plurality of queues in the queue set 23. The queue selector 24 is adapted to route applied data packets arriving at the input port 22 to corresponding ones of the queues in accordance with the addresses contained in the routing tag signals. An output arbitrator 25 is coupled between each output port and respective queues of the queue set. The output arbitrator 25 processes data packets whose destinations are the output port coupled thereto and applies the data packets stored in each of the respective queues to that output port in accordance with a predetermined priority arbitration scheme.
A more detailed description of the basic multiple queue packet switching node 21 is provided in the presently copending application entitled "Packet Switched Multiple Queue NxM Switch Node and Processing Method," cited hereinabove. This patent application is incorporated herein by this reference.
The load balancing circuit arrangement 20 is coupled between each of the queues of the queue set 23 and the queue selector 24. The load balancing circuit arrangement 20 monitors the number of data packets stored in each of the queues. The load balancing circuit arrangement 20 provides input signals to the queue selector 24 which replace the routing tag signals of the data packets currently being processed. These input signals are representative of the queue in the queue set 23 which has the fewest packets stored therein. The load balancing circuit arrangement 20 comprises a load balancing circuit 28 coupled to the queues 23 of the packet switching node 20. The load balancing circuit 28 monitors the number of data packets addressed to each of the output ports 26 by examining the number of packets in the queue 23 associated with each output port 26. The load balancing circuit 28 is adapted to generate new routing tag signals identifying output pert addresses which redistribute the output port load. A tag selection circuit 27 is coupled between the load balancing circuit 28 and the queue selector 24 which is adapted to selectively replace the routing tag signals of the applied data packets with the new routing tag signals in order to redistribute and balance the output port load. The load balancing circuit 28 may comprise a minimum index circuit coupled to an adder circuit. The minimum index circuit is adapted to generate the new routing tag signals and the adder circuit is adapted to combine the new routing tag signals with offset signals which modify the new routing tag signals in order to implement a predetermined output port priority scheme. These two circuits will be described in more detail with reference to Fig. 2. The tag selection circuit 27 may comprise a two to one multiplexer, or the like, which is adapted to selectively choose between two input signals for output to the queue selector 24. In operation, the multiple queue packet switching node 21 employing the load balancing circuit arrangement 20 functions in the following manner. In a typical multi-processor computer application requiring load balancing, or the like, a process controller, or the like, is coupled to the input port 22. This process controller generates data and instructions which are sent to functionally identical execution units individually coupled to the output ports 26. The data and instructions are sent in data packets which are processed by the packet switching node 21. The switching node 21 provides a communications link between the process controller and the execution units. Each data packet includes a routing tag signal identifying the particular output port 26, and hence the particular execution unit, to which the packet is to be sent. In addition, the data packet also includes data that is to be operated upon in that execution unit. A data packet enters the switching node 21 and the queue selector 22 reads the routing tag signal. It then routes the data packet to the appropriate queue of the queue set 23 which processes signals whose destinations are the particular output port 26 corresponding to the routing tag signal. In the multi-processor computer of this example, each of the output ports are connected to functionally identical execution units. Therefore, it is not particularly important where the instructions contained in the data packets get executed, but that they are executed in a timely manner. Since all of the execution units are functionally identical, routing tag signals may be changed in order to more efficiently utilize the execution units. The load balancing circuit arrangement 20 produces the desired re-routing of the data packets to accomplish more efficient operation of the switching node 21 and the computer system in general. The load balancing circuit 20 processes count signals generated by each of the queues 23 in the queue set 24. The count signals indicate the number of data packets stored in a particular queue. The load balancing circuit arrangement 20 then analyzes the count signals to determine which queue has the least number of data packets stored therein. Once this has been determined, the routing tag signal in the data packet being processed by the queue selection logic 23 is changed to identify the least full queue.
To better understand the operation of the load balancing circuit 20, reference is made to Fig. 2. Fig. 2 shows a more detailed block diagram of the load balancing circuit 28. The load balancing circuit 28 is comprised of the minimum index circuit 30 and en adder circuit 31. In Fig. 2, the load balancing circuit 28 is shown adapted for use with a node which comprises four output ports, hence it employs four sets of inputs to the minimum index circuit 30, identfied as L[0] to L[3].
Each set of inputs to the minimum index circuit 30 is individually coupled to count circuitry in the queues which counts the number of data packets currently being stored and processed by a particular queue. The output of the minimum index circuit 30 is coupled to a modulo M adder 31 (modulo 4 in Fig. 2) which adds an offset to the index produced. The load balancing circuit 28 implements an algorithm which determines the least full queue in the queue set. A typical algorithm is presented below.
I. INPUT SCALAR: OFFSET 2. INPUT ARRAY: L[M]
3. OUTPUT SCALAR: ADDR
4. BEGIN
5. minimum:=0;
6. FOR (i=1; WEILE i<M; i:=i+1) 7. IF (L[i] < L[minimum])
8. minimum: =i ;
9 . ENDFOR
10. ADDR: = (minimum + OFFSET) MODULO M;
II . END This algorithm scans the inputs L[M] from low order to high, looking for the smallest value. In the case of a tie, the value is retained, thus implementing a particular priority scheme. This priority scheme is as follows. Since there can be a tie between two or more queues for the lowest number, a default priority is assigned. Let L[0] be the first input to the minimum index circuit 30 and let L[M-1] be the last input. In the case of a tie, L[0] has the highest priority and L[M-1] has the lowest. Alternatively, if the first queue is not to have the highest priority, rather queue i, 0<i<M, then the queue i data packet couut is connected to L[0], the queue with the next highest priority is connected to L[1], and so on. In addition, in order to obtain the correct re-routing tag signal, the binary representation of i is connected to the OFFSET input to the modulo M adder 31. To convert this algorithm into a logic circuit requires the specification of the maximum queue size and the number of output ports (M), in order to determine the number of input and output bits in the load balancing circuit 28.
The ability to specify priority is important. In an NxM packet switching node, for example, N of the load balancing circuits are required. If all circuits gave top priority to the lower numbered outputs, there would be a bias in the overall operation of of the switching node favoring these outputs. The distribution of the load over the outputs would not be uniform. To illustrate how the load balancing circuit
28 is reduced to logic, assume a four by four packet switching node with queues that can hold up to three data packets. Therefore, the number of input ports N = 4 and the number of output ports M = 4, and the number of data packets P = 3. The equation [Log2(P+1)] =
[Log2(3+1)] = 2 represents the number of bits required to indicate the data packet count in any queue. Since there are four queues, a total of eight inputs are needed for the minimum index circuit 30. Also, [log2M] = 2 bits are required to indicate the which queue is to be selected. Therefore, each input to the minimum index circuit 30, represented by L[i], is represented by two bits, Lil and LiO, and ADDR and OFFSET are represented by two bits each, ADDRl, ADDRO, and OFFSETl, OFFSETO, respectively. The load balancing circuit 28 comprises two parts, the minimum index circuit 30 and the modulo four adder 31 which combines the output of the minimum index circuit 30 with the OFFSET value. A modified algorithm is required to implement the minimum index circuit only, and is presented below.
1. INPUT ABBAY: L[M]
2. OUTPUT SCALAR: OUT
3. BEGIN 4. minimum:=0;
5. FOR (i=1; WHILE i<M; i:=i+1)
6. IF (L[i] < L [minimum])
7. minimum:=i;
8. ENDFOR 9. OUT:=minimum; 10. END
This algorithm has been programmed into a computer to generate a truth table with eight boolean input variables and two boolean output variables. This truth table was then made the input to a program which generated the following "sum-of-product" form logic equations. These equations correspond directly to a two level nand gate implementation of the minimum index circuit 30.
Figure imgf000017_0001
A detailed schematic of the logic diagram representing a specific embodiment of the minimum index circuit 30 is shown in Fig. 3. Fig. 3 is comprised of Figs. 3a and 3b. Figs. 3a and 3b should be placed abutting one another and joined along the line of separation in order to obtain the complete drawing.
The "sum-of-product" form of the equations for the modulo four adder 31 are as follows. Cl = +
CO =
Figure imgf000018_0001
Since the B input is the OFFSET constant, once the constant is selected, it is substituted into the above equations which results in simpler circuits. If the OFFSET = 0, no circuit is required. Figs. 4a, 4b and 4c show the circuits to be used when the OFFSET is 1, 2 and 3, respectively. A detailed discussion of the components and wiring of the circuit of Figs. 3 and 4 will be dispensed with. The load balancing circuit arrangement of the present invention may also be incorporated into a multiport memory packet switching node. Fig. 5 illustrates a portion of a four input, four output multiport memory packet switching node without load balancing. This circuit is identical to the circuit of Fig. 3 of the presently copending multiport memory patent application cited hereinabove. Fig. 5 shows control logic 40 of the multiport memory packet switching node. The control logic comprises four input port logic sections 41a-d coupled to four output port logic sections 42a-d. In addition, handshaking signal lines 45 and status update control logic 46 are included in Fig. 5.
Referring to Fig. 6, there is shown a portion of one channel of the multiport memory packet switching node of Fig. 5 incorporating a load balancing circuit arrangement 20' of the present invention. The discussion of this embodiment is undertaken with reference to the presently copending patent application entitled "Packet Switched Multiport Memory NxM Switch Node and Processing Method" cited hereinabove with particular reference to Fig. 3 thereof.
Fig. 6 shows a portion of the control logic 40 which includes an input port logic section 41a coupled to four output port logic sections 42a-d. A more detailed understanding of the complete circuit will be had with reference to Fig. 3 in the previously cited patent application. The signal lines illustrated in Fig. 6 of this application correspond to signal lines described in Fig. 3 of the copending patent application.
The load balancing circuit arrangement 20' has inputs coupled to request output lines REQ0, REQ1, REQ2 and REQ3 of the input port logic section 41. Outputs of the load balancing circuit arrangement 20' are coupled to a latch 45 which is used to ensure correct signal timing and then to the enabling circuit 27. A load balancing circuit 28' comprises a minimum index circuit 30' and the adder 31. The minimum index circuit 30' for use in this embodiment is shown in more detail in Fig. 7. The adder circuit 31 is substantially the same as hereinabove described with reference to the load balancing circuit arrangement for the multiple queue node. Thus, there has been described a load balancing circuit for use with a packet switching node which may be employed in multi-processor and parallel computer systems, and the like. The load balancing circuit may be used to automatically redistribute data packets to external execution units coupled to a multiple queue or multiport memory switching node. This redistribution results in more efficient operation of the node and computer system employing networks incorporating the load balancing circuit. it is to be understood that the abovedescribed embodiments are merely illustrative of some of the many specific embodiments which represent applications of the principles of the present invention. Clearly, numerous and varied other arrangements may be readily devised by those skilled in the art without departing from the spirit and scope of the invention. In particular, the present invention may also be used in applications involving the communications field. Interconnection networks employing the principles of the present invention may be used to connect telephone systems that communicate both data and voice information by way of data packets.

Claims

CLAIMS What is claimed is:
1. A load balancing circuit arrangement for use with a packet switching node that processes applied data packets that contain routing tag signals indicative of the output port destinations to which said data packets are addressed, wherein said load balancing arrangement comprises: load balancing means coupled to said packet switching node for monitoring the output port addresses of said applied data packets and the number of data packets addressed to each of said output ports, and for generating new routing tag signals identifying output port addresses which are adapted to redistribute the output port load; and tag selection means coupled to said load balancing means and said packet switching node for selectively replacing the routing tag signals of said applied data packets with said new routing tag signals in order to redistribute and balance the output port load.
2. The load balancing circuit arrangement of Claim 1 wherein: said load balancing means comprises a logic circuit which monitors the output port addresses of said applied data packets and the number of data packets addressed to each of said output ports, and generates said new routing tag signals; and said tag selection means comprises a logic circuit for selectively applying said new routing tag signals: to said switching node in response to enabling signals applied thereto.
3. The load balancing circuit arrangement of Claim 2 wherein said load balancing means comprises: a minimum index circuit for generating said new routing tag signals; and an adder circuit coupled to said minimum index circuit for combining said new routing tag signals with offset signals which modify said new routing tag signals in order to implement a predetermined output port priority scheme.
4. A load balancing circuit arrangement for use with a switching node that processes applied data packets containing routing tag signals indicative of the output port destinations to which said data packets are addressed, and wherein said switching node comprises a multiport memory and control logic coupled thereto for routing applied data packets in accordance with the output port destination thereof, wherein said load balancing circuit arrangement comprises: load balancing means coupled to said control logic for monitoring the output port addresses of said applied data packets and the number of data packets addressed to each of said output ports, and for generating new routing tag signals identifying output port addresses which are adapted to redistribute the output port load; and tag selection means coupled to said load balancing means and said control logic for selectively replacing the routing tag signals of said applied data packets with said new routing tag signals in order to redistribute and balance the output port load.
5. The load balancing circuit arrangement of Claim 4 wherein: said load balancing means comprises a logic circuit which monitors the output port addresses of said applied data packets and the number of data packets addressed to each of said output ports, and generates said new routing tag signals; and said tag selection means comprises a logic circuit for selectively applying said new routing tag signals to said switching node in response to enabling signals applied thereto.
6. The load balancing arrangement of Claim
5 wherein said load balancing means comprises: a minimum index circuit for generating said new routing tag signals; and an adder circuit coupled to said mimimum index circuit for combining said new routing tag signals with offset signals which modify said new routing tag signals in order to implement a predetermined output port priority scheme.
7. A load balancing circuit arrangement for use with a packet switching node that processes applied data packets containing routing tag signals indicative of the output port destinations to which said data packets are addressed, and wherein said switching node comprises queue selection logic coupled to a plurality of queues for routing applied data packets in accordance with the output port destination thereof, wherein said load balancing circuit arrangement comprises: load balancing means coupled to said plurality of queues for monitoring the number of data packets addressed to each of said output ports, and for generating new routing tag signals identifying output port addresses which are adapted to redistribute the output port load; and tag selection means coupled to said load balancing means and said queue selection logic for selectively replacing the routing tag signals of said applied data packets with said new routing tag signals in order to redistribute and balance the output port load.
8. The load balancing circuit arrangement of Claim 7 wherein: said load balancing means comprises a logic circuit which monitors the output port addresses of said applied data packets and the number of data packets addressed to each of said output ports, and generates said new routing tag signals; and said tag selection means comprises a logic circuit for selectively applying said new routing tag signals to said switching node in response to enabling signals applied thereto.
9. The load balancing circuit arrangement of Claim 8 wherein said load balancing means comprises: a minimum index circuit for generating said new routing tag signals; and an adder circuit coupled to said minimum index circuit for combining said new routing tag signals with offset signals which modify said new routing tag signals in order to implement a predetermined output port priority scheme.
PCT/US1985/001839 1984-10-18 1985-09-26 Load balancing for packet switching nodes WO1986002511A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
DE8585905022T DE3574659D1 (en) 1984-10-18 1985-09-26 LOAD COMPENSATION FOR PACKAGE SWITCHING NODES.
KR1019860700370A KR900006792B1 (en) 1984-10-18 1985-09-26 Load balancing for packet switching nodes

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US06/661,995 US4621359A (en) 1984-10-18 1984-10-18 Load balancing for packet switching nodes
US661,995 1991-02-27

Publications (1)

Publication Number Publication Date
WO1986002511A1 true WO1986002511A1 (en) 1986-04-24

Family

ID=24655964

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US1985/001839 WO1986002511A1 (en) 1984-10-18 1985-09-26 Load balancing for packet switching nodes

Country Status (7)

Country Link
US (1) US4621359A (en)
EP (1) EP0197103B1 (en)
JP (1) JPS62500627A (en)
KR (1) KR900006792B1 (en)
DE (1) DE3574659D1 (en)
IL (1) IL76533A (en)
WO (1) WO1986002511A1 (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2638260A1 (en) * 1988-10-26 1990-04-27 Onera (Off Nat Aerospatiale) SWITCHING DEVICES AND DATA COMMUNICATION NETWORKS FOR MULTIPROCESSOR SYSTEMS
WO1991004624A1 (en) * 1989-09-15 1991-04-04 The Plessey Company Limited An asynchronous time division multiplex switching system
EP0426990A2 (en) * 1989-11-06 1991-05-15 Masaru Kitsuregawa Method of and apparatus for data distribution
EP0497097A2 (en) * 1991-01-08 1992-08-05 Nec Corporation Switching system with time-stamped packet distribution input stage and packet sequencing output stage
EP0512141A1 (en) * 1991-05-07 1992-11-11 Siemens Aktiengesellschaft Procedure for switching high bit rate ATM data cell streams through a switching device with a lower bit rate
US5475680A (en) * 1989-09-15 1995-12-12 Gpt Limited Asynchronous time division multiplex switching system
US5603028A (en) * 1992-03-02 1997-02-11 Mitsubishi Denki Kabushiki Kaisha Method and apparatus for data distribution
GB2346302A (en) * 1999-01-29 2000-08-02 Ibm Load balancing in a computer network
EP1145505A1 (en) * 1999-01-12 2001-10-17 McData Corporation Method and apparatus for measuring traffic within a switch
WO2004084501A2 (en) * 2003-03-21 2004-09-30 Marconi Uk Intellectual Property Limited Paths in telecommunications networks
US7382736B2 (en) 1999-01-12 2008-06-03 Mcdata Corporation Method for scoring queued frames for selective transmission through a switch

Families Citing this family (72)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4792941A (en) * 1985-02-25 1988-12-20 Itt Corporation Data subsystem traffic control apparatus and method
US5047917A (en) * 1985-07-12 1991-09-10 The California Institute Of Technology Apparatus for intrasystem communications within a binary n-cube including buffer lock bit
US4734907A (en) * 1985-09-06 1988-03-29 Washington University Broadcast packet switching network
EP0219049B1 (en) * 1985-10-07 1992-09-02 Nec Corporation High-speed packet-switched communications system with end-to-end flow control and retransmission
EP0243563B1 (en) * 1986-04-30 1990-09-12 International Business Machines Corporation Non coded information and companion data switching mechanism
CA1262274A (en) 1986-06-20 1989-10-10 Randall D. Kun Isdn d channel handler
US4821265A (en) * 1987-04-06 1989-04-11 Racal Data Communications Inc. Node architecture for communication networks
US4949338A (en) * 1987-04-06 1990-08-14 Racal Data Communications Inc. Arbitration in multiprocessor communication node
CA1292054C (en) * 1987-05-15 1991-11-12 Hitachi, Ltd. Packet switching equipment and a packet switching method for controlling packet switched networks
FR2618280B1 (en) * 1987-07-16 1989-10-20 Quinquis Jean Paul PRIORITY INFORMATION SWITCHING SYSTEM.
US4797880A (en) * 1987-10-07 1989-01-10 Bell Communications Research, Inc. Non-blocking, self-routing packet switch
US5367518A (en) * 1987-10-15 1994-11-22 Network Equipment Technologies, Inc. Self-routing switching element and fast packet switch
US5222085A (en) * 1987-10-15 1993-06-22 Peter Newman Self-routing switching element and fast packet switch
US5245603A (en) * 1987-10-15 1993-09-14 Network Equipment Technologies, Inc. High-speed determining unit for prioritizing and arbitrating among competing input signals
GB8724208D0 (en) * 1987-10-15 1987-11-18 Newman P Self-routing switching element
JPH01126760A (en) * 1987-11-11 1989-05-18 Toshiba Corp Parallel computer system
JP2810043B2 (en) * 1987-11-13 1998-10-15 株式会社日立製作所 Data processing device
JPH01177239A (en) * 1988-01-06 1989-07-13 Nec Corp Packet concentrator and packet switching device
US5105424A (en) * 1988-06-02 1992-04-14 California Institute Of Technology Inter-computer message routing system with each computer having separate routinng automata for each dimension of the network
US5396491A (en) * 1988-10-14 1995-03-07 Network Equipment Technologies, Inc. Self-routing switching element and fast packet switch
US4899335A (en) * 1988-12-21 1990-02-06 American Telephone And Telegraph Company, At&T Bell Laboratories Self routing packet switching network architecture
CA2011934A1 (en) * 1989-07-19 1991-01-19 Theodore Heske, Iii Method and apparatus for source routing bridging
CA2015514C (en) * 1989-08-22 1996-08-06 Mitsuru Tsuboi Packet switching system having bus matrix switch
JP2531275B2 (en) * 1989-09-29 1996-09-04 日本電気株式会社 ATM cell transfer method
US5172371A (en) * 1990-08-09 1992-12-15 At&T Bell Laboratories Growable switch
US5274782A (en) * 1990-08-27 1993-12-28 International Business Machines Corporation Method and apparatus for dynamic detection and routing of non-uniform traffic in parallel buffered multistage interconnection networks
US5321813A (en) 1991-05-01 1994-06-14 Teradata Corporation Reconfigurable, fault tolerant, multistage interconnect network and protocol
US5235592A (en) * 1991-08-13 1993-08-10 International Business Machines Corporation Dynamic switch protocols on a shared medium network
SE515148C2 (en) * 1993-06-23 2001-06-18 Ericsson Telefon Ab L M Control of cell selector
US5515502A (en) * 1993-09-30 1996-05-07 Sybase, Inc. Data backup system with methods for stripe affinity backup to multiple archive devices
US5631897A (en) * 1993-10-01 1997-05-20 Nec America, Inc. Apparatus and method for incorporating a large number of destinations over circuit-switched wide area network connections
US5555378A (en) * 1994-03-02 1996-09-10 Bell Communications Research, Inc. Scheduling transmission multimedia information in broadband networks using a token passing scheme
US6704765B1 (en) 1994-12-14 2004-03-09 International Business Machines Corporation System for allocating resources among agent processes
US5706516A (en) * 1995-01-23 1998-01-06 International Business Machines Corporation System for communicating messages among agent processes
JP3095163B2 (en) * 1995-07-19 2000-10-03 株式会社日立製作所 How to control the automatic logon function
US5978844A (en) * 1995-09-08 1999-11-02 Hitachi, Ltd. Internetworking apparatus for load balancing plural networks
US6345041B1 (en) * 1996-10-24 2002-02-05 Hewlett-Packard Company Method and apparatus for automatic load-balancing on multisegment devices
US6263368B1 (en) * 1997-06-19 2001-07-17 Sun Microsystems, Inc. Network load balancing for multi-computer server by counting message packets to/from multi-computer server
US6363077B1 (en) * 1998-02-13 2002-03-26 Broadcom Corporation Load balancing in link aggregation and trunking
US6658565B1 (en) 1998-06-01 2003-12-02 Sun Microsystems, Inc. Distributed filtering and monitoring system for a computer internetwork
US6567377B1 (en) * 1999-03-18 2003-05-20 3Com Corporation High performance load balancing of outbound internet protocol traffic over multiple network interface cards
US6775280B1 (en) * 1999-04-29 2004-08-10 Cisco Technology, Inc. Methods and apparatus for routing packets using policy and network efficiency information
US6788692B1 (en) * 1999-05-03 2004-09-07 Nortel Networks Limited Network switch load balancing
US6412002B1 (en) 1999-11-15 2002-06-25 Ncr Corporation Method and apparatus for selecting nodes in configuring massively parallel systems
US6418526B1 (en) 1999-11-15 2002-07-09 Ncr Corporation Method and apparatus for synchronizing nodes in massively parallel systems
US6745240B1 (en) 1999-11-15 2004-06-01 Ncr Corporation Method and apparatus for configuring massively parallel systems
US6519697B1 (en) 1999-11-15 2003-02-11 Ncr Corporation Method and apparatus for coordinating the configuration of massively parallel systems
US6980533B1 (en) * 2000-04-19 2005-12-27 Lucent Technologies Inc. Load balancing technique for a wireless internet access system
JP3770058B2 (en) * 2000-06-16 2006-04-26 株式会社日立製作所 Frame distribution method and information processing apparatus having the function
CN1146261C (en) * 2000-10-27 2004-04-14 清华大学 Method for retransmitting lost packets in fading channel
US6982986B2 (en) * 2001-11-01 2006-01-03 International Business Machines Corporation QoS scheduler and method for implementing quality of service anticipating the end of a chain of flows
US7046676B2 (en) * 2001-11-01 2006-05-16 International Business Machines Corporation QoS scheduler and method for implementing quality of service with cached status array
US7280474B2 (en) * 2001-11-01 2007-10-09 International Business Machines Corporation Weighted fair queue having adjustable scaling factor
US7103051B2 (en) * 2001-11-01 2006-09-05 International Business Machines Corporation QoS scheduler and method for implementing quality of service with aging time stamps
US7310345B2 (en) 2001-11-01 2007-12-18 International Business Machines Corporation Empty indicators for weighted fair queues
US7187684B2 (en) * 2001-11-01 2007-03-06 International Business Machines Corporation Weighted fair queue having extended effective range
US7317683B2 (en) * 2001-11-01 2008-01-08 International Business Machines Corporation Weighted fair queue serving plural output ports
US7257124B2 (en) * 2002-03-20 2007-08-14 International Business Machines Corporation Method and apparatus for improving the fairness of new attaches to a weighted fair queue in a quality of service (QoS) scheduler
US7680043B2 (en) 2002-03-20 2010-03-16 International Business Machines Corporation Network processor having fast flow queue disable process
DE10212374B4 (en) * 2002-03-20 2005-12-08 Siemens Ag Load distribution to peripheral facilities of a packet-based switching system
US7554914B1 (en) * 2002-07-10 2009-06-30 Cisco Technology, Inc. System and method for adaptively balancing network traffic over router output ports
TW577237B (en) * 2002-11-18 2004-02-21 Inst Information Industry Load balance system and method of wireless LAN
US7397796B1 (en) * 2003-08-21 2008-07-08 Smiljanic Aleksandra Load balancing algorithms in non-blocking multistage packet switches
US20050060423A1 (en) * 2003-09-15 2005-03-17 Sachin Garg Congestion management in telecommunications networks
US7451209B1 (en) 2003-10-22 2008-11-11 Cisco Technology, Inc. Improving reliability and availability of a load balanced server
US7421695B2 (en) * 2003-11-12 2008-09-02 Cisco Tech Inc System and methodology for adaptive load balancing with behavior modification hints
US7760637B1 (en) * 2004-06-10 2010-07-20 Alcatel Lucent System and method for implementing flow control with dynamic load balancing for single channelized input to multiple outputs
ATE547872T1 (en) * 2007-03-30 2012-03-15 British Telecomm AD-HOC COMMUNICATION SYSTEM
US7768907B2 (en) 2007-04-23 2010-08-03 International Business Machines Corporation System and method for improved Ethernet load balancing
KR101754998B1 (en) * 2011-01-27 2017-07-06 삼성전자주식회사 Multi Core System and Method for Processing Data
CN104113486A (en) * 2013-04-18 2014-10-22 中兴通讯股份有限公司 Cell data processing method and device, and exchange network element
US10740155B1 (en) * 2018-09-28 2020-08-11 Cavium, Llc Methods and systems for load balancing in network devices

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1984001077A1 (en) * 1982-09-02 1984-03-15 Burroughs Corp Five port module as a node in an asynchronous speed independent network of concurrent processors

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2432804A1 (en) * 1978-08-03 1980-02-29 Trt Telecom Radio Electr HIGH SPEED DATA SWITCH
US4484326A (en) * 1982-11-04 1984-11-20 At&T Bell Laboratories Packet load monitoring by trunk controllers
JPS59135953A (en) * 1983-01-26 1984-08-04 Fujitsu Ltd System for deciding data transmitting path
US4550397A (en) * 1983-12-16 1985-10-29 At&T Bell Laboratories Alternate paths in a self-routing packet switching network

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1984001077A1 (en) * 1982-09-02 1984-03-15 Burroughs Corp Five port module as a node in an asynchronous speed independent network of concurrent processors

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Conference Proceedings: the 9th Annual Symposium on Computer Architecture, 26-29 April 1982 (Austin, US) PARKER et al.: "The Gamma Network: a Multiprocessor Interconnection Network with Redundant Paths", pages 73-80, see page 74, right-hand column, lines 11-47; figure 1 *
PATENT ABSTRACTS OF JAPAN, Vol. 8, No. 264 (E-282) (1701), 4 December 1984 & JP, A, 59135953 (Fujitsu) 4 August 1984 see the whole document *
Proceedings of the International Teletraffic Congress, Stockholm, 13-20 June 1973, Proc. 7, Part 2, Swedisch Communications 1973 (Stockholm, SW) HERZOG: "Message-Switching Networks with Alternate Routing", pages 415/1-415/8, see paragraph 2.1 "Structure and Operating Made of Data Switching Centers" and figure 2 of page 415/2 *

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2638260A1 (en) * 1988-10-26 1990-04-27 Onera (Off Nat Aerospatiale) SWITCHING DEVICES AND DATA COMMUNICATION NETWORKS FOR MULTIPROCESSOR SYSTEMS
EP0366520A1 (en) * 1988-10-26 1990-05-02 Office National d'Etudes et de Recherches Aérospatiales (O.N.E.R.A.) Data communication network for multiprocessor systems
WO1991004624A1 (en) * 1989-09-15 1991-04-04 The Plessey Company Limited An asynchronous time division multiplex switching system
US5475680A (en) * 1989-09-15 1995-12-12 Gpt Limited Asynchronous time division multiplex switching system
US5237683A (en) * 1989-11-06 1993-08-17 Masaru Kitsuregawa Method and apparatus for data distribution
EP0426990A3 (en) * 1989-11-06 1991-09-04 Masaru Kitsuregawa Method of and apparatus for data distribution
EP0426990A2 (en) * 1989-11-06 1991-05-15 Masaru Kitsuregawa Method of and apparatus for data distribution
EP0497097A2 (en) * 1991-01-08 1992-08-05 Nec Corporation Switching system with time-stamped packet distribution input stage and packet sequencing output stage
EP0497097A3 (en) * 1991-01-08 1992-09-09 Nec Corporation Switching system with time-stamped packet distribution input stage and packet sequencing output stage
EP0512141A1 (en) * 1991-05-07 1992-11-11 Siemens Aktiengesellschaft Procedure for switching high bit rate ATM data cell streams through a switching device with a lower bit rate
US5317561A (en) * 1991-05-07 1994-05-31 Siemens Aktiengesellschaft Method for switching ATM message cell streams having a high bit rate via a switching equipment having a lower bit rate
US5603028A (en) * 1992-03-02 1997-02-11 Mitsubishi Denki Kabushiki Kaisha Method and apparatus for data distribution
EP1145505A1 (en) * 1999-01-12 2001-10-17 McData Corporation Method and apparatus for measuring traffic within a switch
EP1145505A4 (en) * 1999-01-12 2007-08-01 Mcdata Corp Method and apparatus for measuring traffic within a switch
US7382736B2 (en) 1999-01-12 2008-06-03 Mcdata Corporation Method for scoring queued frames for selective transmission through a switch
GB2346302A (en) * 1999-01-29 2000-08-02 Ibm Load balancing in a computer network
GB2346302B (en) * 1999-01-29 2003-06-18 Ibm Pre-emptive network load balancing by predictive configuration
WO2004084501A2 (en) * 2003-03-21 2004-09-30 Marconi Uk Intellectual Property Limited Paths in telecommunications networks
WO2004084501A3 (en) * 2003-03-21 2005-01-20 Marconi Uk Intellectual Prop Paths in telecommunications networks

Also Published As

Publication number Publication date
KR880700567A (en) 1988-03-15
KR900006792B1 (en) 1990-09-21
US4621359A (en) 1986-11-04
EP0197103B1 (en) 1989-12-06
IL76533A0 (en) 1986-01-31
DE3574659D1 (en) 1990-01-11
JPS62500627A (en) 1987-03-12
IL76533A (en) 1990-08-31
EP0197103A1 (en) 1986-10-15

Similar Documents

Publication Publication Date Title
EP0197103B1 (en) Load balancing for packet switching nodes
EP0198010B1 (en) Packet switched multiport memory nxm switch node and processing method
US4623996A (en) Packet switched multiple queue NXM switch node and processing method
EP0334954B1 (en) Layered network
US5170482A (en) Improved hypercube topology for multiprocessor computer systems
US6314487B1 (en) Adaptive routing controller of a crossbar core module used in a crossbar routing switch
US5175733A (en) Adaptive message routing for multi-dimensional networks
JP2533223B2 (en) Multi-stage communication network
JP4128447B2 (en) Scalable interconnect structure for parallel computing and parallel memory access
JPS6367047A (en) Packet switch
Wah A comparative study of distributed resource sharing on multiprocessors
JPS63257052A (en) Multiprocessor system
Ramakrishnan et al. Analysis of network on chip topologies
WO1989001665A1 (en) Hypercube topology for multiprocessor systems with added communication paths between nodes or substituted corner topologies
JP3079068B2 (en) ATM switch
Ciminiera et al. Interconnection networks for MIMD machines
JPH01105642A (en) Data transfer network for parallel computer
Prakash et al. VLSI implementation of a wormhole router using virtual channels
Sharif et al. Design and simulations of a serial-link interconnection network for a massively parallel computer system
JPH11275613A (en) Cross-bar switch, sub-crossbar switch unit, parallel computer, telephone exchange, and frame data exchange
Kiamilev et al. Modular architecture for smart pixel switching networks
Lee et al. Performance analysis of STC104 interconnection networks
Shepherd et al. Next Generation Transputers and Beyond-1: Transputers and Routers: Components for Concurrent Machines
JPH09179840A (en) Inter-processor coupling network for parallel computer
JPH0830571A (en) Data transfer network

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): JP KR

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): DE FR GB IT NL

WWE Wipo information: entry into national phase

Ref document number: 1985905022

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1985905022

Country of ref document: EP

WWG Wipo information: grant in national office

Ref document number: 1985905022

Country of ref document: EP