WO1996038958A2

WO1996038958A2 - Distributed cam integrity system

Info

Publication number: WO1996038958A2
Application number: PCT/EP1996/002334
Authority: WO
Inventors: Anne O'connell; Thomas C. Hogan; John Healy
Original assignee: 3Com Ireland
Priority date: 1995-05-31
Filing date: 1996-05-30
Publication date: 1996-12-05
Also published as: AU5902696A; GB9510930D0; WO1996038958A3; GB9805471D0; GB2322055A; GB2322055B

Abstract

A switch module for a computer network has a memory device for storing the relationship between the ports of the switch and the MAC address of the devices connected thereto. One port is designated a communications port and any packet destined for a device not listed in the memory device of the switch is sent to the communications port. A number of switch modules can be connected together with data flowing between modules via their communications ports. One of the modules will be provided with an output port for connection to the rest of the network.

Description

DISTRIBUTED CAM INTEGRITY SYSTEM

The present invention relates to computer networks and more particularly networks which are usually known as local area networks (LANs).

Local area networks are now well known and can vary in size considerably. With a network of anything more than a few ports, it is customary to separate the network into operational areas so that those devices which will normally wish to share information and/or send information to each other are grouped together but be able to communicate with other portions of the network via a device known variously as a bridge or switch. One of the features of a bridge or switch is that it contains an up-to-date list of all device addresses and the ports to which those terminals are connected.

For large networks, the network is separated by bridges and switches. Normally each bridge keeps a record in a memory of all ports in the network and the associated MAC addresses. This requires each bridge to have a very large memory. As part of the normal maintenance of the network, it is the practice to "age out" the address of any port and the MAC address of a device connected thereto which is not being used. In this way each CAM of each bridge or switch is kept up to date.

We propose to adopt a different approach. Rather than keep a record of all ports and all devices in the network we propose to keep a record within a switch of only those ports which are connected to the switch. All destination addresses which are not on the switch are sent to an output port from where they are sent to the rest of the network.

We further propose that a switch will be constructed using a plurahty of functionally identical modules each with a CAM, a plurality of ports for connection to devices and a communication port for transferring data between modules. One of the modules will be provided with an output port for connection to the rest of the network.

With this modular arrangement each module will know which ports and which devices are local to it. Packets for any non-local ports are sent to the communications port. In contrast with normal bridges which learn the ports and addresses of all devices in the network, each module will only learn the port numbers and addresses of all ports which are local to it. All addresses which are non-local will be allocated the port number of the communications port except for the module which is provided with the output port for the rest of the network.

Our modular arrangement works best as part of a large network having at least one conventional bridge or router which knows the location of all ports. This arrangement, however cannot use the conventional "age out" process and so a special ageing process has to be carried out in our switch. This is achieved by each module having marked as a "permanent entry" in its CAM, all non-local MAC addresses. These permanent entry addresses can only be removed upon instructions from other modules in the switch whose addresses these are local to.

In order that the present invention be more readily understood, an embodiment thereof will now be described by way of example with reference to the accompanying drawing, in which:- Fig. 1 shows diagrammatically a computer network incorporating a number of bridges or structures;

Fig. 2 shows one embodiment of a bridge or switch according to the present invention; and

Fig. 3 shows a part of the bridge or switch shown in Fig. 2 for explaining the operation thereof.

The preferred embodiment of the present invention assumes that a local area network will be separated up into a number of discrete areas using bridges or switches and that the ports incorporated within one or more areas will be provided as ports on a box which is also provided with its own microprocessor and CAM. This is shown in Fig. 1 where a network is assumed to comprise five multi-port bridges or switches 1,2,3 and 4 which are interconnected. Each switch supports a number of devices attached to ports of the switch in a conventional manner. It is common to supply multi-port bridges or switches with a predeteπnined number of ports but problems arise when a user wishes to connect more devices to a bridge than there are ports available. In the past this has required replacing the multi-port bridge with another having a greater number of ports. Users are not always happy with this approach as it tends to increase the cost to them. For parts of the network at the periphery e.g. switch 1 or 4, we propose that a multiport bridge or switch be assembled from one or more identical modules and for the sake of this example we will use four identical modules ABCD as shown in Fig. 2 which when connected together in a stack will function as a single bridge or switch as far as the network is concerned and is capable of operating as a switch. Each module is provided with a number of ports to which devices can be connected. In Fig. 2 we show that there are ten ports numbered 0-9 for each module for this purpose although it will be noted that port 0 of module A is in fact the port which connects to the rest of the network.

Each module is also provided with an additional port 10 which is used as a communications port with access to a local bus 11 so that data can be transferred from module to module within the stack. In this way additional ports can be provided for a bridge or switch simply by supplying one or more modules and connecting them to the data bus 11.

As is customary, each module has a CAM 12 for containing the ports and associated MAC address of all devices which are connected to the stack. It also knows which MAC addresses are its responsibility, being local to itself. In mis embodiment, any MAC address which the CAM and processor have identified as being non-local, is flagged as a "permanent entry". One way in which this may be achieved will now be described in relation to Fig. 2. The following explanation is given by way of example with reference to the CAM's of two modules, and for the sake of this example the modules will be identified as box A and box B. In Fig. 2 we assume that the MAC address of the device attached to port 4 of box A is in the form 0000010 and that this device wishes to utilize the network by sending a standard ethernet packet which includes a source address, a destination address and data. The processor of box A examines its CAM for the source address but fails to find it. The arrangement is such that in these circumstances the source address and port are "learned" by the CAM with an entry being made in the CAM for the box A. The processor then examines the CAM for the destination address for the packet which it finds. This indicates that the packet must be directed to the communications port 10.

For the sake of this example, we suppose that the destination of this packet is in fact a device connected to the box B. In this case the packet will be received by the communications port 10 on the box B via the local bus and the processor for the box B will communicate the packet to the appropriate port. The processor will insert the source address of the packet into its CAM 12 together with a flag, which will indicate that the source MAC address is not on a normal port of the box B but is somewhere else having come via the communications port 10. This flag will be discussed on more detail later. As more and more devices utilize the network, the CAMs of the modules will gradually fill with source addresses associated with ports or flags indicating that the source address is associated with another module and should not be handled in the same way as local ports. Periodically a message is generated requesting a so-called ageing operation in orderto delete old addresses or addresses which have not been in use for a specified period of time. This is part ofnormal network housekeeping. The ageing operation can be under the control of a "master" module if one of the modules is deemed to be a "master" with all other as slaves. Alternatively it could recur when any one CAM is full and needs updating. Each microprocessor then carries out an ageing out process in respect of its own CAM for all entries which are not flagged as permanent entries. However, instead of simply deleting old MAC addresses, the microprocessor records the MAC addresses of old entries prior to deletion from the CAM. It then forms a data packet with a list of MAC address entries which it has deleted locally and sends this list to each other microprocessor in the stack composed by modules interconnected by the local bus 11.

Each other microprocessor in the stack receives data packets from the other microprocessors during the ageing process and decodes and deletes those entries from its local CAM which it has been told by other microprocessors that it may delete. In this way, all addresses in all CAM's are kept up-to-date. All CAM's contain the same MAC addresses but the port numbers associated with the MAC addresses will differ depending on whether the address is on a port local to the particular CAM or not. The same will apply to the "permanent entry" flags.

Thus far it is being assumed that all communications are between ports on the same switch. However, it will be recalled that box A is provided with a port 0 via which the rest of the network can be accessed. If a packet is received via port 0 of box A, the processor and CAM of box A will process the packet by examining the destination address for the packet. Bearing in mind that box A will have a CAM which stores the MAC addresses of all devices connected to the stack, the processor and CAM of box A can determine whether the destination address is one of the MAC addresses on the stack. If it is, then the packet will be sent to the appropriate port of box A i.e. to the actual port to which the device having the destination address is connected or to communications port 10 if the address is elsewhere in the stack. The source address of this packet is not learned by any of the CAMs. If, however, the packet has a destination address which is not known to the CAM of box A, the packet is discarded or ignored in some convenient way. In our case, we prefer to send such a packet to a non existent port e.e. port 31.

In this way, the switch learns only source addresses associated with its own ports. It does not learn from destination addresses nor does it learn from the rest of the network.

Claims

CLAIMS :

1. A switching means for a network comprising a plurality of ports each arranged to be connected to external devices, means for storing addresses of devices attached to said plurality of ports, a further port arranged to be connected to the rest of the network, and means coupling the ports together in order to provide data flow between selected ports wherein control means is provided which comprises means for recognising the desire of a port to transmit data, means for determining if the destination of the data to be transmitted is one of said devices and means arranged to transmit said data to the appropriate one of said plurality of ports in the case that the determination is positive and to transmit said data to said further port in the case that the determination is negative.

2. A switching means according to claim 1 comprising a plurality of switching units each unit comprising at least one of said plurality of ports and means for storing the address or addresses of devices attached to said at least one of said plurality of ports, further comprising a bus means connected to each of said switching units for data transmission between said units.

3. A switching unit according to claims 1 or 2, wherein the means for storing MAC addresses is arranged to differentiate between devices local to said unit, and devices not local to said unit.

4. A switching unit according to any of claims 1 to 3 comprising means to keeping a record of all devices local to itself which have recently transmitted information, and means for informing any other switching units attached to it via said bus means of said devices. such that said other switching units can keep a record of said device.

5. A switching unit according to claim 4 comprising ageing means for removing devices from said record if they have not transmitted data for a specified period of time, and means for informing other units attached to said unit via said bus of said removal, such that said other units can also remove said devices from their device record.