CA2084516A1 - Repeaters for digital data networks - Google Patents

Abstract

A multiport repeater for a data network operating on a CSMA or CSMA/CD protocol, for example an Ethernet local area network, has signals transmitted from port to port via a crosspoint switch or other appropriate switching means. The destination address of each incoming data frame is read and compared with a stored table to determine which port or ports the frame needs to be transmitted to, and the switching means is then operated to cease transmitting the frame unnecessarily to other ports. The remaining ports are thus returned to an idle state, and a second frame overlapping with the first can be received and retransmitted provided it does not require any of the ports that are in use in transmitting the first frame. If the second frame does require one of the ports that are already in use, then a collision signal is returned and the source station will stop transmission in the normal way, after which the switching means is reset to respond appropriately when the source station re-tries (or another signal is received).

Description

2084~16 Repeaters for Diaital Data Networks This invention relates to multiport repeaters for digital data networks operating on "CSMA"or more especially "CSMA/CD" lcarrier sensing multiple access with or without collision detection) protocols, such as the widely-used Ethernet protocol for local area networks. In such networks, individual stations are linked ~normally via a "transceiver"or "media access unit" (MAU)) to network hardware that includes at least one "multiport repeater" (MPR) and may also include one or more than one "router" and/or one or more than one "bridge".
The links may be formed by electrical oonnections through wires or coaxial cables, by optical links through fibres or free space, or in any other appropriate way.
Data for transmission through such a network is assembled at the source location into digital packets of pre-arranged format. ~he format includes (among other things) a destination address segment that identifies the station intended to receive the packet and a source address segment identifying the station from which it originated, both of which are normally located near the beginning of the frame. Normally each station is allocated a unique sequence of digits (its "address") which is used to identify it whether in the destination address segment or the source address segment.
In the CSMA/CD protocol, any station that has data ready to transmit will attempt to do so provided it is not currently receiving data from the network, and if a second station begins to transmit in the short interval after the first has begun to transmit but before the second has become aware of it, a "collision"
signal is generated and both stations stop transmitting and wait, before trying again, for a length of time that is either randomised or pre-allocated so as to be SUBSriTUTE SHEET

W092/17963 ~ PCT/GB92/00569-different for each station, so that repeated collisions between signals from the same two sources are avoided or minimised.
It is in th~e~Lnterest of both the speed and the capacity of the,network to minimise the occurrence of data collisions, and in small basic networks data frames are distributed to all stations as rapidly as possible by at least one MPR which (or each of which) regenerates incoming signals and distributes them indiscriminately to all the stations and/or other MPR's that are directly connected to it with a time delay that is only a fraction of the length of a data-frame.
In larger networks, this approach gradually becomes unsatisfactory, as the number of stations seeking access increases and the aggregate time delay through a chain of MPR's also increases, and even in relatively small networks there may ~e unacceptable security implications in transmitting all the frames to every station. Some networks are therefore divided into sections linke~ not by MPR's but by "bridges"or "routers". The difference between a ~ridge and a router is irrelevant to understanding of this inventioni each of these stores an incoming frame while it reads at least the destination address and compares it with stored data to determine which part(s) of the network need to receive it. Once this is done, the frame is retransmitted only to that part (or parts) of the network. Bridges and routers are complex and expensive, and they necessarily introduce a delay which will often be substantially longer than a frame length. On The other hand, they increase the capacity of the system by allowing separate control of collisions in the pre-arranged sections of the network and by reducing the overall amount of traffic on some elements of the network.
The present invention provides an MPR tha.
increases the capacity of a CSMA/CD network for SUBSTITUTE SHE~

W~92/17963 2 0 8 4 5 ~ 6 PCT/GB92/00569 comparatively little extra cost and that operates with no more delay than an ordinary MPR and without requiring the division of the network into pre-arranged sections.
The multiport repeater in accordance with the invention comprises (a) a plurality of bidirectional communication port~
each for linking the repeater to a station or to another repeater (or other hardware, if required) and (b) regenerators for regenerating incoming signals received on any one of the said ports and outputting the regenerated signal to the other said ports and is characterised by (c) switching means enabling the said regenerated signal to be outputted to all or to less than all of the said other ports and (d) control means with facilities for reading the address segment of each incoming data frame and, by comparison with stored data, identifying some of said other ports as ports to which the address read from the incoming frame does not apply and for operating said swit~hing means for a chosen time interval to cease outputting the said regenerated signal to the ports so identified and to return those ports to a condition in which an incoming signal thereon can be received and regenerated for onward transmission to other of those ports.
Preferably the repeater also includes means for reading the destination segment of a frame received by a port in the condition last referred to, for comparing it with stored data to determine whether it corresponds to another port that is currently in that condition and for outputting a collision signal to the first of these ports if it does not.
Preferably the switching means is a crosspoint switch, or the equivalent, capable of switching signals received at any port independently to each and every one of the other ports.

SUBSTITUTE SHI~ET

W092~17963 ~ 6 PcT~Gn92/0056~-The bidirectional ports (or some of them) may be composite ports constituted ~y distinct input and output ports. While the existence of other possible strategies is not excluded, the following operating sequences form part of t~e invention.
First seauence In the absence of any data signals, all the ports are recorded as "idle"(O) and all the switches are closed so that a signal appearing as input on any port of the repeater will be regenerated and initially outputted on all the ports, (one port receiving (Rl) and all others transmitting (Tl)). As the address segment of the data frame is received, it is extracted and looked up in a stored table to identify the port (or ports in the special case of a network with duplicated pathways to provide resilience against breakdown) to which the addressee station is connected, and the switching means is immediately actuated to disconnect all the (Tl) ports except the identified one(s) from the tRl) port but to maintain them in connection with each other, restoring them to the (O) state.
If a data frame is now received on one of those ports that has been restored to the (O) state, putting it into a receiving state (R2), it is regenerated and initially retransmitted to all other ports that were in the (O) state, changing their state to a transmitting state (T2). Again the address segment is extracted and compared with the stored table, but the essential purpose of the reference is to determine whether the addressee station is on a port that is in the (T2) state or not: if it is not (because it is in the (Rl) or (Tl) state) the incoming frame cannot be accepted and a collision signal should be applied as rapidly as possible to the (R2) port to stop the originating station transmittins the incoming frame.
If the addressee station is on a (T2) port, outputting may simply continue to all the (T2) ports, or SUBStITUTE SHE~

~92/17963 2 0 8 4 316 PCT/GB92/00569 if desired it may continue only to the port(s) to which the addressee station is connected and the other port~
may be disconnected from those ports but connected to each other and returned to the (Oj state to allow one or more than one further iterationi however, it will be realised that the benefits diminish with each iteration.
When the reception and outputting of a data frame has been completed, the ports concerned may be promptly returned to the idle tO) state, and reconnec~ed by the switching means to any other ports that are currently in that state. Howeverj in many networks there is a high probability that a station that has just transmitted a frame to a particular other station will immediately attempt to send at least one more frame to the same station. Preferably, therefore, there is a time delay before the idle state is restored such that if another frame is received on the same receiving port (say Rl) within the normal cycle time of the network it will initially be outputted only to the same transmitting port(s) (Tl) as before; the address segment is read and compared and if it again correspondsito the same (Tl) port~s), no action is needed; otherwise a collision signal needs to be returned to the receiving (Rl) port and the receive and transmit ports returned immediately to the idle state tO) so that if possible the required connection will exist when the originating station re-tries.
Second Se~uence In networks in which the probability of a succession of frames being transmitted from one source station to one destination station is high, it may be a useful alternative to operate the switching means to cease outputting the regenerated signal only when outputting of one frame has been completed.
This has the advantage that the switching operation creates no "fragments" tuncomplete frames), SUBSTITUTE SHEET

W092/17963 ~6 PCT/GBg2/0056r--~ 6 -but the minor disadvantage that single frames and the first frame of eacn successlon will be transmitted every station (except any that was not connected to an "idle"
port immediately ~efore the frame begar. to be received).
In this case it will generail~ be appropriate to apply a time delay between the end of a frame and the switching operation to restore the idle state.
The multiport repeater may be pre-programmed with the required ta~le relating station numbers (destination addresses) with its port numbers, or it may be programmed to "self-learn" and maintain its own table by reading the source address of each incoming frame and to enter it, with the number of the port on which the signal was received, in its table if it is not already included. Particularly (but not only) in a self-learning repeater so programmed, a frame addressed to a station not listed in the stored table should be transmitted to all available ports, or ideally to all of the ports -the latter will re~uire the use of collision signals unless and until all ports are in the idle(O) condition.
The invention will be further described, by way of example, with reference to the accompanying drawings in which Fig. l is a topology diagram of a small local area network including at least multiport repeater in accordance with the inventioni Fig. 2 is a simplified functional diagram of a multiport repeater in accordance with the invention; and Fig. 3 is a diagramatic representation of a crosspoint switch forming part of the multiport repeater of Fig. 2.
Referring first to Fig. l, the network shown for the purpose of illustration comprises two multiport repeaters, MPR-A which is a repeater in accordance with the present invention and MPR-B which can be another multiport repeater in accordance with the invention or SUBSTITUTE SHEET

2 0 8 4 ~ ~ 6 PCT~GB92/00569 can be an entirely conventional multiport repeater. MPR-A has six ports l to 6, of which port 6 is connected to MPR-B; the network also includes terminal stations distinguished by the binary numbers OOOl to lOlO of which numbers OOOl to OlOl are connected to various ports of MPR-A as indicated and the remainder to ports of MPR-B.
The multiport repeater selected to illustrate the invention is a self-learning one, and on initial start up the operation of the network is entirely conventional with the first incoming dataframe ~eing relayed to all--stations, except only the one that originated it.
Referring now mainly to Fig. 2, which is simplified inter alia by omitting all repetitions of functional units, the conventional operation of the repeater is served by a signal received on one of the ports 7 being passed through a conventional regenerator circuit 8 and then via a cross-point switch 9 to the remainder of the ports 7 (if preferred, the regenerator could be located downstream of the crosspoint switch).
- In parallel to its being fed to the crosspoint switch 9, the regenerated signal is passed to a shift register lO which extracts from the signal frame its destination address and source address, which are supplied respectively to a comparator/reader ll and comparator/updater 12. Both the comparator reader ll and the comparator updater 12 have access to memory 13, and the comparator/reader ll looks for the destination address in the memory 13; at this stage, since there is as yet no data in the memory, it will not find it, and consequently sends no instruction to the switch controller 14 which therefore leaves the crosspoint switch setting such that the signal continues to be transmitted to all of the ports (except the input port).

SUBSTITUTE SHEEr 3 ~ PCT/GB92/0~6'-Meantime, the comparator/updater 12 also compares the source address of the incoming frame with the list in memory 13, and likewise finds that it is not present; it then interrogates a port state memory 15 to identify the port that~is in a recei~ing (Rl) state, and thus the port to which the new source address is connected, either directly or indirectly; the comparator/updater 12 writes this information to memory 13.
This process is repeated as other stations transmit respective first frames, and in a relatively short time, memory 13 will build up a complete table of stations as shown in table 1 Table 1 Station No. Port No.

OlO0 2 OlOl 5 lOO0 6 As soon as MPR-A receives a frame that is addressed to a station that has previously transmitted a frame, the comparator!reader 11 will find the destination address it reads in the memory 13 and will then send instructions to the switch controller 14 such that it will maintain the connection between the receiving (Rl) port and the port to which the addressee station is connected tTl), as determined from the table in memory 13, but disconnect those ports from all the others (returning the others to the (O) state).

SUBSTITUTE~ SHEEr W092/17963 PCT/GB92/~569 - 9 - 2084~16 Referring to Figure 3, the cross-point switch 9 is shown as a standard 7 X 7 crosspoint switch with both rows and columns numbered O to 6 and with each row connected to the correspondingly numbered column by permanently closing the constituent switches at cross-points [O, O], [1, 1] .... [6, 6] or by equivalent hard wiring (No other connections are made to the columns) Each of rows 1-6 is connected via a respective regenerator circuit (assumed ~idirectional) to the correspondingly numbered port, while row O is connected to the collision signal generator 16. Thus the constituent switch at cross-point [n, m] where n>m and m>O serves to connect ports nos. n and m to each other, when required; in the absence of action by the switch controller 14, all these fifteen switches are closed The constituent switch at each cross-point [n, O]
provides the facility to connect the collision signal generator to port no. ni in the absence of action by switch controller 14, these six switches are open.
Switches at crosspoints [n, m] where n~m are redundant (e.g. the switch at [1, 4] is put in parallel with that at [4, 1] by the fixed connection at 4, 4) but the use of a standard cross-point switch with simple logic and some redundancy is likely to be preferable to the avoidance of redundancy in the switching means by adoption of more complex logic in the switch controller.
Suppose now that station 0001 transmits a frame intended for station 0110, which has previously transmitted at least one frameand is therefore logged in memory 13 against the port number 3. When the frame begins to appear on port 1, all the switches interconnecting the ports are already closed as just explained, but as soon as the comparator/reader 11 determines from the table in memory 13 that the destination station of the frame is on port 3, it instructs the switch controller 14 to operate the SUBSrITUTE SHEET

WO92/17963 ~ PCT/GB~2/00569-crosspoint switch so as to open the constituent switches at [2,1], [4, l], [5, l], [6, l] and [3, 2] (that is those which connect ports l and 3 other than to each other), leaving all the others closed. The remainder of the frame is therefore transmitted only to port 3, and for the remainder or the time taken to receive an~
retransmit that frame, port l is in the state (Rl), por_ 3 in the state (Tl) and all the other ports in the state (O) .
For the moment, assume that the other stations remain inactive until the frame from station OOOl has - been completely transmitted. At the end of the frame, the state of the crosspoint switch is maintained from a short time, in the expectation that station OOOl will immediately send another frame also addressed to station OllO. If it does not transmit another frame within the normal cycle time, then the switch controller 14 automatically resets the crosspoint switch 9 to its initial state. If it does transmit a frame within the cycle time, then it is initially transmitted, in accordance with the existing connections, only to port 3. Meantime its destination address is read by the shift register lO (Fig. 2) and passed to the comparator/reader ll which refers to the memory 13 and so determines whether the destination address corresponds to the same port as it instructed the switch controller 14 to route the first frame to.
If so, it instructs the switch controller 14 (if necessary) to maintain the existing setting in respect of ports l and 3; if not, then the frame is not being transmi$ted to the station intended to receive it and the comparator/reader ll instructs the switch controller 14 to close the constituent switch [l, 0] and thereby send a collision signal from generator 16 to station OOOl for long enough to terminate its transmission and then to reset by opening the switch at [l,0] and closing all the port to port switch units.

SUBSTITUTE SHEEr WO92/17963 PCT/GB92/00~69 - ll - 208~

Next suppose that, while a frame is being transmitted from port 1 to port 3, station 0010 wishes to transmit a frame to station O101, and that that station has previously transmitted. Since port 4 is not receiving the frame being transmitted from port 1 to port 3, a collision will not be generated on receipt of a frame on this port, but the frame will initially be retransmitted to all the ports except ports 1 and ~. As before, the incoming signal is passed to a shift register 10 (which depending on the organisation of the multiport repeater maybe a different shift register from that which received the first frame, or may be the same one) and its destination address is passed to a comparator/reader 11 as before. This interroyates the memory 13 (which will be common to all the units) and determines that the addressee station is on port 5 to which the signal is being transmitted. At this point, the simpler MPR's in acc~rdance with the invention will take no action, and the frame received on port 4 will continue to be transmitted to ports numbers 2, 5 and 6.
If desired however, ports 2 and 6 may be switched out by the compara~tor/reader sending appropriate instructions to the switch controller, either for improved security or so that a third frame received on one of those ports could be transmitted to the other (or vice versa), even while station 0001 continues to transmit to station OllO
and station 0010 continues to transmit to station O101.
Next suppose that, while a frame is being transmitted from station 0001 to station OllO while ports 2, 4, 5 and 6 remain in the (0) state, station 0011 attempts to transmit a frame to station 0001.
Initially, the incoming frame will be retransmittea to ports 4, 5 and 6 only, which is futile since the addressee station is not connected to the any of those ports. However, the destination address of the frame will, as before, be extracted by a shift register 10 and compared by a comparator/reader 11 with the content of SUBSTITUTE SHEEI' WO92/17963 ~ PCT/GB92/00~6-memory 13, whereby the comparator/reader 11 will determine that the destination address corresponds to a port that is not in the (0) state and which cannot therefore be successfully transmitted. The comparator/reader 11 therefore instructs the switch controller 14 to close the element [2, 0] of the crosspoint switch to supply a collision signal to port 2 and thereby in the usual way stop station 0011 from attempting to transmit and then re-sets to restore ports 2, 4, 5 and 6 to the (0) state. Station 0011 will in due course re-try according to the system media access - protocol.
Finally suppose that, while all the ports on MPR-A are in the (0) state, station Olll transmits a signal through MPR-B (assumed to be a conventional multiport repeater) to station 1000, then this frame will also appear on port 6 of MPRA. Initially, it will be repeated to all of ports 1 to 5, but once the destination address has been read by shift register 10 and referred to the table in memory 13 by comparator/reader 11, it will be determined that the addressee station is connected (indirectly, as it happens) to port 6, and comparator/reader 11 will therefore instruct switch controller 14 to operate crosspoint switch 9 by opening the switch elements [6,1], ~6,2], [6,3], [6,4] and [6,~], so that MPR-A no longer transmits at all until port 6 has either returned to the idle ( n ) state and remained in it for a prearranged length of time or has received a frame addressed to a station on one of the other ports.

SUBSTITUTE SHEET

Claims (7)

1. A multiport repeater comprising (a) a plurality of bidirectional communication ports each for linking the repeater to a station or to another repeater and (b) regenerators for regenerating incoming signals received on any one of the said ports and outputting the regenerated signal to the other said ports and characterised by (c) switching means enabling the said regenerated signal to be outputted to all or to less than all of the said other ports and (d) control means with facilities for reading the address segment of each incoming data frame and, by comparison with stored data, identifying some of said other ports as ports to which the address read from the incoming frame does not apply and for operating said switching means for a chosen time interval to cease outputting the said regenerated signal to the ports so identified and to return those ports to a condition in which an incoming signal thereon can be received and regenerated for onward transmission to other of those ports.

2. A multiport repeater as claimed in claim 1 including means for reading the destination address segment of a frame received by a port in the condition last referred to in claim 1, comparing it with stored data to determine whether it corresponds to another port that is currently in that condition and for outputting a collision signal to the first of these ports if it does not.

3. A multiport repeater in accordance with claim 1 or claim 2 in which the switching means is a crosspoint switch or the equivalent, capable of switching signals received at any port independently to each and every one of the other ports.

4. A multiport repeater as claimed in any one of the preceding claims which is pre- with a table relating station numbers (destination addresses) with its port numbers.

5. A multiport repeater as claimed in any one of claims 1 to 3 which is programmed to self-learn and maintain its own table relating station numbers with its port numbers by reading the source address of each incoming frame and entering it, with the number of the port on which the signal was received, in its table if it is not already included.

6. A multiport repeater substantially as described with reference to figures 1 and 2 of the drawings.

7. A multiport repeater substantially as described with reference to figures 1 to 3 of the drawings.