Virtual-chassis switch network topology

Details Virtual-chassis switch network topology Virtual-chassis switch network topology

1998-04-30

2003-05-20

Hsu, Alpus H. (Department: 2665)

Multiplex communications

Pathfinding or routing

Switching a message which includes an address header

C370S392000

Reexamination Certificate

active

06567403

ABSTRACT:

BACKGROUND OF THE INVENTION
The present invention relates to computer networks and, more particularly, to computer networks incorporating network switches.
A major objective of the invention is to provide a high-performance network that can be initiated economically on a small scale and then expanded as needed.
Much of modern progress is marked by the proliferation of computers, and with their proliferation has come the need for computers to communicate with each other. Computer networks link computers and peripherals to facilitate their mutual communication. As the number of computers and peripherals increases, networks must expand. Simple peer-to-peer networks are superceded by networks that use network-specific devices (e.g., hubs, switches, routers) to forward data packets (frames) from one node device (e.g., computer, peripheral) to another.
The simplest network-specific device is a repeater, which regenerates a received signal to compensate for signal degradation over distances. A hub is a multi-port repeater, transmitting a received signal out all ports other than the one at which it was received. A hub can accommodate as many node devices (e.g., computers, printers) as it has ports. To accommodate greater numbers of devices, multiple hubs can be used and linked to each other. However, since each hub is broadcasting every received signal, the burden on network bandwidth can be significant.
Switches (including 2-port switches called “bridges”) selectively retransmit received data, thereby reducing the bandwidth burden relative to hubs. Each port of a switch can be coupled a network segment. A “network segment” is a portion of network in which all transmissions are receivable by all included node devices, as well as any interconnecting hubs.
The operation of a typical switch conforms to an IEEE 802 standard, according to which switches handle “layer 2” packets. Each layer 2 packet consists of a 48-bit destination “media access control” (MAC) address, a 48-bit source MAC address, control information, and data. The source and destination addresses identify network devices. Switches selectively forward packets by analyzing the destination address, and switches learn the location of network devices by examining the source address. The learning function is defined by IEEE Standard 802.1D.
A switch conforming to the IEEE 802.1D learning bridge standard checks its forwarding database to determine if the destination address of a packet has been mapped to a port of the switch. The forwarding database includes mappings between the source addresses of packets previously received and the ports at which those packets were respectively received. If the destination address has been mapped, the packet is forwarded out the port associated with the destination address. If the destination address has not been mapped to a port, the packet is flooded, i.e., transmitted out all ports other than the one at which the packet was received, so that it reaches all possible network destinations.
If a network expands to the point where the required number of ports exceeds those available, there are three possibilities. The first is to replace the switch with a switch with more ports. The second is to increase the number of ports in a switch. The third is to add additional switches to the network. The first solution, replacement, is obviously wasteful.
The second solution, adding ports, only applies to certain expandable switches that involve a relatively high expense per port in the minimal configurations that might be preferred at the early stages of a network's growth. For example, a network chassis might include a backplane on which slots are mounted for several switch cards. Each switch card might have four ports. Buying an initial four-port chassis-based switch can be expensive because the chassis and backplane are included. The additional port cards tend to be proprietary, and therefore more expensive than their manufacturing costs would imply.
The third solution, adding switches, is economically attractive in that the number of ports can be scaled almost linearly with the number of switches. For example, two twelve-port switches can provide up to twenty-two ports for network segments, with one port on each switch being used for communicating with the other switch. An additional advantage of adding switches to a network rather than adding ports to a switch is that the maximum distance between network segments increases. For example, if the maximum recommended cable length is 100 meters, then in a one-switch network, network segments can be at most 200 meters apart. In a two-switch network, this limit is raised to 300 meters.
Communication bandwidth is an issue when multiple switches are used. In a chassis-type switch, a common backplane provides for intercard communication. The bandwidth provided by the backplane can be readily designed to handle the maximum load provided by a full set of network cards. However, where a single network cable couples separate switches, that cable's bandwidth can be exceeded easily. In an extreme example using two 12-port switches, eleven pairs of network nodes could compete for bandwidth over a single cable connecting the two switches. “Trunking” is an approach to increasing inter-switch bandwidth that involves using more than one pair of ports for inter-switch communication. In related-art
FIG. 1
, three ports PA
1
, PA
2
, and PA
3
of one switch SWA can be respectively coupled to three ports PB
1
, PB
2
, PB
3
of another switch SWB to define a trunk TRK and triple the available inter-switch bandwidth. The remaining ports PA
4
, PA
5
, PA
6
, PB
4
, PB
5
, PB
6
can be coupled to other unique network segments.
For trunking to work, normal switch operation must be modified. Accordingly, proposed IEEE standard 802.3 defines a special “trunk mode”. In the foregoing example, “physical” ports PA
1
, PA
2
, and PA
3
collectively define a “logical” trunk TPA; likewise, physical ports PB
1
, PB
2
, and PB
3
collectively define a logical trunk TPB.
When a packet with an unknown destination is received by switch SWA, flooding it out all physical “trunked” ports PA
1
, PA
2
, PA
3
of its trunk TPA would cause switch SWB at the other end of the trunk to receive the packet on all its trunked ports PB
1
, PB
2
, PB
3
. If it responded in a like manner, switch SWB would flood multiple replicas of the packet back to switch SWA, which would flood packets again to switch SWB. Thus, a network loop would be formed, tying up network bandwidth unless and until some network fault handling procedure intervened. Furthermore, the redundant transmission would offset some of the additional bandwidth that trunking is intended to provide. Accordingly, trunk mode includes a “flood-reduction” procedure to ensure no packet is flooded out more than one physical port of a trunk.
In addition, the proposed IEEE standard 802.3 provides for a load-balancing procedure. The theoretical limit of bandwidth that increases linearly with the number of trunk lines can only be approached if the communications load is evenly distributed among the trunk lines. To help ensure even distribution, the normal learning procedure of mapping unknown source addresses to the trunked port at which a packet is received is replaced in trunk mode by a “trunk-load-balancing” procedure. For example, unknown source addresses can be mapped to trunked ports in a round-robin manner; thus, when a packet with an unknown source address is received at a trunk port, the source address may be mapped to the physical port at which the packet was received, but it may also be mapped to one of the other physical ports in the trunk. (A switch in trunk mode maps an unknown source address received at a non-trunk port to that port, just as it would in normal mode.) Alternatives to the round-robin procedure include random port selection and load-based port selection.
The main cost of trunking is the loss of ports for connecting segments. For example, two 12-port switches with a three-line trunk have 18 lines for coupling network s

Affiliated with

Also associated with

Rating

Virtual-chassis switch network topology does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Virtual-chassis switch network topology, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Virtual-chassis switch network topology will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3039391