Multiplex communications – Pathfinding or routing – Switching a message which includes an address header
Patent
1996-06-11
1998-05-12
Kizou, Hassan
Multiplex communications
Pathfinding or routing
Switching a message which includes an address header
370427, 370536, 395311, H04L 1200
Patent
active
057517100
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
This invention relates generally to network switches and, more particularly, to an arrangement for efficiently interconnecting cards of a network switch.
BACKGROUND OF THE INVENTION
A switched communication network comprises a collection of interconnected nodes in which data are routed through the network between communicating nodes. The nodes may be computers configured as end nodes and intermediate nodes, the latter connecting only to the end nodes or other intermediate nodes of the network. These intermediate nodes function primarily to provide a switching facility that transfers data among the other nodes and, thus, are generally referred to as network switches. As used herein, the term "network switch" includes bridges, routers and other similar intermediate network nodes.
The network switch typically comprises an enclosure having a plurality of connector slots for receiving network cards, such as input/output line cards. The network cards are, in turn, interconnected by a backplane of wires that is often controlled by a switching matrix to transfer data among the cards. Typically, the matrix comprises circuitry contained on a switch card that is centrally-located (with respect to the backplane) for receiving and transmitting data among source and destination cards internal to the switch.
The centralized switching matrix may be implemented as a bus, a cross-bar or other type of interconnection of wires, such as a cube. A bus is a multipoint data path that interconnects more than two cards. Because multiple cards share a single data path, only one may transmit at a time; this raises the problem of determining which card may transmit at any point in time. A centrally-located switch card matrix may also function as an arbiter to resolve such contention. Although an inexpensive solution, this approach provides relatively low performance because it cannot handle significant parallel activity.
In contrast, a cross-bar switching matrix provides a point-to-point connection between a source card transmitting data and the switch card, and a further point-to-point connection between the switch card and the destination card. Each connection thus requires the establishment of a physical, cross-point path through the switch card that is dedicated solely to the transfer of data between the source and destination cards. Here, the number of connections increases with the square of the number of cards in the switch; these connections may be large, even where a physical path consists of a single wire.
However, a single wire connection between cards may be insufficient to provide the bandwidth required for relatively high-data rate applications. In this case, each point-to-point connection typically comprises many wires which, for a conventional network switch, results in a voluminous number of wires in the enclosure. A very large connector assembly would be required to accomodate these wires at the switch card. Yet, a typical connector of such size may not fit in the enclosure, thereby necessitating use of an expensive, customized connector. In the alternative, the connector function must be apportioned onto two or three cards which consumes valuable card space in the enclosure.
A more significant problem with the cross-bar switch may concern its susceptibility as a single point of failure. That is, loss of a single cross-point path in the switch card destroys the connection between the two corresponding cards, a situation that effectively renders the network switch inoperable. A solution to this latter problem may be to provide two, redundant switch cards per enclosure, although this is an expensive and complicated option primarily because of redundancy and synchronization requirements.
An alternative to the cross-bar arrangement is the cube interconnect, e.g., the hybercube network. The hybercube network is typically employed for interconnecting many (N) processing elements in an n-dimensional cube arrangement, where N=2.sup.n. Here, each processing element is connected to n other elements w
REFERENCES:
patent: 4821176 (1989-04-01), Ward
patent: 4879551 (1989-11-01), Georgiou
patent: 4968977 (1990-11-01), Chinnaswamy et al.
patent: 5317562 (1994-05-01), Nardin et al.
patent: 5386567 (1995-01-01), Lien et al.
patent: 5491804 (1996-02-01), Heath et al.
patent: 5583867 (1996-12-01), Poole
patent: 5625780 (1997-04-01), Hsieh et al.
Brandt William P.
Crowther William R.
Cisco Technology Inc.
Kizou Hassan
LandOfFree
Technique for connecting cards of a distributed network switch does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Technique for connecting cards of a distributed network switch, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Technique for connecting cards of a distributed network switch will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-990278