Synchronization in a communication network of interconnected rings

Present invention is concerned with synchronization in a communication network consisting of a plurality of transmission rings interconnected by bridges. In particular, it is concerned with the system-wide coordination of the release of synchronous information frames at regular intervals on token-controlled rings which per se are not frame or slot synchronized. Local area networks are an application field for which this technique is of particular interest.

The use of transmission rings for communication networks is well known. They may be either fully slot or frame synchronized so that all fields for information appear in a fixed time raster at regular intervals, or they may use a special access technique such as a token mechanism which allows frames of variable lengths to occur asynchronously, i.e. at irregular intervals.

Systems with rings of the first kind are known e.g. from US Patent 3'586'782 and from US Patent 3'731'002. Systems with rings of the other kind, i.e. operating asynchronously with tokens, are known from an article by W. Bux et al. "A local area communication network based on a reliable token-ring system" published in "Local Computer Networks", North-Holland Publishing Co. 1982, pp. 69 ... 82, and from European Patent Application 0 054 077.

For larger networks, it may be desirable to interconnect several individual closed rings via connecting units. Such combined networks are also shown in above mentioned two US patents and the article by Bux et al.

The asynchronous token rings, though very effective with respect to bandwidth utilization and due to their ability to transport variable-length or irregularly occurring messages, were per se not suited for transporting synchronous information such as digitized voice samples. However, in above-mentioned European Patent Application 0 054 077 a technique was disclosed for accommodating synchronous traffic on such rings. In-between asynchronous frames, synchronous tokens or frames are issued at regular intervals for "collecting" all messages that must be synchronously transmitted.

If synchronous messages are to be transmitted across the network between different rings, there are two basic possibilities. The first is to completely synchronize the whole network and provide TDM switches at the ring interconnections, like in a PCM telephone network. This, however, is not desirable for the asynchronous token rings. Another possibility is to provide buffers at the interconnections between rings of different data rates, and extract the data from the buffers under independent control of the individual rings, as described in above mentioned patents. This may, however, cause difficulties for the periodic transfer of synchronous information, in particular when asynchronous token rings are interconnected.

It is an object of present invention to devise a mechanism which enables efficient exchange of messages on a periodic or synchronous basis in a network of interconnected asynchronous token rings.

It is also an object to present a method to synchronize, on a system-wide basis, that portion of the operation of interconnected rings which deals with the transmission of synchronous information, despite the fact that no slot or frame synchronization is provided and each ring may have a different bit rate or clock than the other rings.

It is a further object to achieve the particular synchronization and the efficiency of synchronous-information transmission across rings with a minimum in hardwear or protocol effort.

A method and communication network for achieving these objects are presented in claims 1 and 8.

The method respectively has the advantage that it allows to interconnect rings of different purpose and design, i.e. rings having completely unrelated bit rates, and nevertheless allow the transportation of information needing a regular sampling, e.g. voice, at a rate common to the whole network. No particular interconnections or channels need be provided between the different bridges for the synchronization, as the existing transmission rings and the frames that must anyway be exchanged are sufficient for this purpose in the inventive method.

An embodiment of the invention will be described in connection with drawings. The drawings show the following: <ul id="ul0001" list-style="none"><li>Fig. 1 A network of token rings interconnected by bridges, in which the invention finds application.</li><li>Fig. 2 Frame format used for transport information in the system of Fig. 1.</li><li>Fig. 3 Details of one of the bridges shown in Fig. 1, interconnecting three rings.</li><li>Fig. 4 Block diagram of some details provided for synchronization in those bridge nodes comprising a synchronous bandwidth manager.</li><li>Fig. 5 Block diagram of some details provided for synchronization in bridge nodes without bandwidth manager, or in nodes for normal stations attached to a ring.</li><li>Fig. 6 Timing diagram showing the relation between synchronous token or frame starts and synchronizing pulses in the different rings and bridges of the system shown in Fig. 1.</li><li>Fig. 7A Timing diagram of a procedure for ensuring timely reservation of a transmission ring for synchronous information frames.</li><li>Fig. 7B Flow diagram of the precedure for ensuring timely reservation of a transmission ring for synchronous information frames.</li><li>Fig. 8 Timing diagram of a resulting sequence of synchronous and asynchronous information transmissions in periodic intervals with gaps resulting from early reservations in accordance with the procedure presented in Figs. 7A and 7B.</li><li>Fig. 9 Timing diagram showing the exchange of synchronous messages between different rings which have a phase shift in the periodic start of their synchronous frames to allow optimization of the overall transmission delay.</li></ul>

System Overview

A communication network in which present invention finds application is shown in Fig. 1. It comprises a plurality of transmission rings R1 ... R7 which are interconnected by bridges B1 ... B4. Some of the bridges are directly interconnected by a backbone ring BB. To each ring (except the backbone ring in present example) several data stations are attached via ring adapters or nodes 11. Only a few of the stations, i.e. ST-N, ST-M, ST-X, and ST-Y are identified in Fig. 1 for later reference.

Messages can be exchanged between rings via the bridges. For this purpose, each ring has a ring adapter or node 13 in the bridge to which it is connected. Backbone ring BB has similar node units 15 and 17 in the bridges it interconnects (node 15 having a master function to be explained later). Within each bridge, a message transfer unit 19 is provided for exchange of data between the nodes 13, 15, and 17. Bridge B4 is not connected to backbone ring BB, but rather to a normal ring R5 by another ring adapter or node unit 21.

The configuration shown in Fig. 1 is of course only one example. Rings could also be interconnected without a backbone ring, as e.g. in a string configuration (taking only R4-B2-R5-B4-R7) or in a tree configuration (if R3 is added to the string). It is also possible to attach stations directly to backbone ring BB via additional node units.

Communication between stations is effected by frames and access is regulated by a token mechanism. Details of token ring communication were already described in above mentioned publication by W. Bux et al. and in European Patent Application 0 054 077. Therefore, no detailed description of the basic mechanism is necessary here. A few essential points only should be mentioned: <ul id="ul0002" list-style="none"><li>(a) On each ring, one node or station has a monitor function. It maintains the bit clock for the respective ring and monitors the token to avoid situations with lost or duplicated tokens.</li><li>(b) On some or all rings, it is possible to transfer, besides the normal asynchronous messages, also synchronously occurring information such as digitized voice samples by a special mechanism disclosed in European Patent Application 0 054 077. For this purpose, each ring has particular means, called "Synchronous Bandwidth Manager" SBM in the following, which provide frames for transfer of synchronous information at regular intervals. In the present embodiment, this SBM function is incorporated in bridge node 13 for each normal ring, and the bridge node 15 for backbone ring BB.</li></ul>

It should be noted that rings R1 through R7 may have different, unrelated bit clocks, i.e. each ring can have an individual data transmission speed. It is an essential feature of present invention that it allows an ordely and efficient exchange of synchronously occurring information between rings via the bridges despite different transmission speeds on the rings.

Frame Format

The frame format used on the rings, which is known from above mentioned previous publication, EP-A-0 054077, is shown in Fig. 2. It comprises following essential elements: <ul id="ul0003" list-style="none"><li>- The actual message to be transferred between stations, including a destination address DA, an origin address OA, and the transported information INFO in a variable length field.</li><li>- A frame check sequence FCS.</li><li>- A transport control field TC comprising eight individual control bits of which three are of particular interest in connection with present invention:</li></ul><ul id="ul0004" list-style="none"><li>TK = Token Bit</li></ul>

Indicates whether this is a free/token (TK = 0) or not (TK = 1 ).- <ul id="ul0005" list-style="none"><li>PI = Priority Indicator</li></ul>

Signifies normal asynchronous transmission mode (PI = 0) or a reservation mode (PI = 1) in which no new asynchronous token must be issued to allow timely issuance of a token for synchronous information. <ul id="ul0006" list-style="none"><li>TM = Traffic Mode Indicator</li></ul>

Used to distinguish frames and tokens for asynchronous information (TM = 0) from those for synchronous information (TM = 1). <ul id="ul0007" list-style="none"><li>- A start delimiter S and an end delimiter E are added for completing the frame.</li></ul>

As indicated in Fig. 2, the two bytes S (start delimiter) and TC (transport control field) will together be designated as a token (or sometimes as "free token") if the token bit TK = 0. Only if a node sees such a free token it can set TK = 1 and then append the remainder of a frame containing the information it wants to send. If, however, the token bit TK is already "1" when it is received, the two first bytes S and TC do not represent a token but only the leading portion of a complete frame.

It should be noted that synchronous information can be transported between stations in two different ways: <ul id="ul0008" list-style="none"><li>(a) A single frame for synchronous information is issued by the SBM at regular intervals. It comprises in the INFO field a plurality of consecutive slots each having space for a busy/free indicator, a destination address, and a data byte. Each authorized station can seize and use one slot, and when the synch information frame circulated twice around the ring it is discarded.</li><li>(b) A first synchronous free token is issued by the SBM at regular intervals. It can be used by one single authorized station which after transmitting its synchronous data issues a new free token for synchronous information. This goes on until all authorized stations were served once.</li></ul>

Only one. of these different transport mechanisms is used in any system. For the present embodiment, method (b) was chosen, i.e. a sequence of synchronous information frames each used by one station.

It should further be noted that duplex transmission between any pair of stations or nodes is possible with a single frame: E.g., in the network of Fig. 1 a frame may first carry a data item UUU from ST-N to ST-M. The latter extracts UUU from the frame and instead inserts data item VVV which then travels to station ST-N.

Bridges

Important elements in the present multiple ring network are the bridges, and the invention is concerned with synchronization of SBM's toallow effective transfer of synchronous information across these bridges. Therefore, a block diagram of a bridge is shown in Fig. 3 and will now be explained, as well as the mechanism of data transfer between rings.

Fig. 3 shows some details of bridge B1 which interconnects rings R1 and R2 and the backbone ring BB. Bridge nodes 13 and 15 each include a ring adapter RA into which a synchronous bandwidth manager SBM is integrated, and which has inputs/outputs for asynchronous information (AS) and for synchronous information (SY). Each bridge node also includes a set of four FlF0 buffers 23A, 23B, 25A, 25B (FIFO = first-in, first- out) for separately storing asynchronous and synchronous data items for incoming and outgoing directions.

Instead of separate FIFO buffers, a common addressable buffer storage could be provided with a suitable addressing and access mechanism which allows the data items or messages to be stored and read separately by direction and information category.

The buffers are all interconnected by a message transfer unit 19. This unit is a TDM switch that can selectively transfer data items from the buffer of one node to the buffer of another node, depending on a destination address associated with the data item and in accordance with a given routing algorithm. Other implementations are possible, e.g. a bus with suitable access control to allow selective transfer of data items between buffers.

As mentioned earlier, bridge node 15 comprises the master SBM for the whole network; the two other bridge nodes 13 comprise normal SBM's. A direct connection 27/27' is provided between the three bridge nodes to enable a transfer of synchronization pulses from the master SBM to the normal SBM's. The other bridges B2, B3, and 84 do not comprise a bridge node with a master SBM, but each has a bridge node 17 or 21 which includes a deputy master or synch extractor from which, via line 27/27' which is provided in each bridge, synch pulses can be transferred to the SBM or SBM's in the other nodes of the respective bridge. The deputy master or synch extractor is not an SBM because each ring can have only one SBM.

Operation of the bridge nodes and the transfer mechanism are as follows: <ul id="ul0009" list-style="none"><li>Each bridge node (13,15,17,21) operates like a normal ring node (11 in Fig. 1) with the difference that instead of a data station that is connected to a normal ring node, a bridge node has the bridge circuitry attached to it which accepts and furnishes data items.</li></ul>

Transfer from ring to bridge: The bridge node watches for frames having a destination address .designating a location on another ring to which the respective bridge is the correct path. If it detects such an address, it extracts the actual message, i.e. destination address, origin address, and information field from the frame and inserts it either into buffer 23A or 25A depending on the status of the traffic mode indicator TM.

Transfer from bridge to ring: <ul id="ul0010" list-style="none"><li>(a) Asynchronous information: The bridge node waits for a free asynchronous token (i.e. with TM = 0), and if it detects one, extracts the next message, i.e. DA, OA and INFO field from buffer 238 and transmits a complete frame by adding the appropriate frame check sequence FCS as well as delimiter E.</li><li>(b) Synchronous information: The SBM in a bridge node issues a synchronous token (i.e. with TM = 1) at regular intervals, basically as described in above mentioned European Patent Application 0 054 077. The SBM further knows how many synchronous "connections" are existing because it must first authorize each additional connection to limit synchronous traffic to some reasonable fraction of total channel capacity. The bridge node may itself first use the synchronous token for transporting data items (synchronous messages) which are waiting in its buffer 25B. The bridge node then sequentially transmits as many frames as synchronous connections exist between this ring and other rings so that all data items or messages which arrived during the last synchronous period are transmitted on the respective ring at the beginning of the next period. Thereafter, i.e. when no more synchronous messages are to be dispatched from buffer 25B, a synchronous free token will be dispatched so that authorized stations on the ring can send their synchronous messages to other stations on the same ring, or to stations attached to other rings. If all authorized stations were served once, the SBM will recognize this and cause a new asynchronous token to be issued.</li></ul>

The synchronization of synchronous bandwidth managers SBM, and of normal station nodes, under control of the master SBM will be described in a separate section in connection with Figs. 4, 5, and 6.

Transfer between buffers: <ul id="ul0011" list-style="none"><li>The message transfer unit 19 cyclically scans those of its inputs which are the outputs of buffers 23A holding asynchronous information. It extracts the next message (if there is one) from each buffer just scanned and interrogates the destination address. Under control of a routing algorithm which associates addresses with rings the message is gated into a buffer 23B of either one of the two other bridge nodes. In the same way, the message transfer unit scans qll buffers 25A holding synchronous information, and gates each message to a buffer 25B in another bridge node, depending on the destination address. Thus, all buffers are served cyclically, and synchronous information is kept separate from asynchronous information during transfer between buffers within the bridge. The message transfer unit must of course operate fast enough to ensure that no synchronous message must wait more than one synchronous period in a buffer.</li></ul>

As an alternative, a single scanning/switching unit can be provided for both, synchronous and asynchronous messages instead of two separate such units. In this case, synchronous traffic buffers must be scanned with highest priority to ensure dispatching of each synchronous message within one period, whereas asynchronous traffic has lower priority.

Common Timing of Synchronous-information Intervals in Multiple-Ring Network

It is important that the periods for synchronous information, i.e. the intervals between release of synchronous tokens, are the same throughout the network. If these periods were different for different rings, an overflow or underflow of buffers in bridges would occur which would lead to loss of information.

One possibility to achieve uniform intervals would be a complete synchronization of the whole network. This is, however, not desirable. Different data speeds and clock rates must be allowed for different transmission rings to enable integration of various subnetworks in a single network system and to maintain flexibility for future developments.

The invention guarantees uniform periods for transfer of synchronous information despite different individual clock rates on the various rings. How this is achieved will now be explained in connection with Figs. 4, 5, and 6. Essentially, all synchronous bandwidth managers SBM are synchronized from one master SBM, but the paths for the synchronizing information are different for the various SBM's, depending on their location.

Fig. 4 is a simplified block diagram of the bridge nodes including either the master SBM or a normal ring SBM, showing in some detail only the elements for timing the synchronous intervals. The respective bridge nodes carry reference numbers 13 or 15 in Fig. 1. The SBM's in these bridge nodes must release each first synchronous token in proper relationship to both, the local bit clock and the overall system clock for synchronous information. The ring bit clock as derived from data received on the ring is provided on line 29. The clock for synchronous intervals is furnished by clock unit 31. This clock unit is synchronized itself by a very exact master oscillator 33 in the bridge node 15 representing the master SBM. Instead of having its own master oscillator, clock unit 31 in the master SBM may use a very exact external clock source (e.g. the clock of a public network to which present ring network is connected). In the bridge nodes comprising a normal SBM, clock unit 31 is synchronized by timing pulses on line 27' which are furnished by another bridge node within the same bridge. Switch SI in the position as shown represents the situation in node 15 (master SBM); in its alternate position, it would represent the situation in nodes 13 (normal SBM's).

Each output pulse (t,) from clock unit 31 is used to set a flipflop 34 indicating start of the next synch period. The setting of flipflop 34 will condition AND gate 35 which will then gate the next pulse of the ring bit clock to line 37. In each of the nodes 13 and 15, this gated pulse (t_R) will be used to control release of the first synchronous token from the respective bridge node to the associated ring. In bridge node 15 (comprising the master SBM), indicated by the closed position of SII, the pulse on line 37 is also furnished via line 27 to the other nodes/SBM's in the same bridge. Token release pulse t_R may be delayed for an interval Ts - 2 D_max by unit 38 for obtaining during each period T_s a pulse tp that leads t_R by two maximum frame durations 2 D_max. This lead interval is used for timely setting priority indicator PI as will be explained in connection with Fig. 7A.

Fig. 5 is a similar simplified block diagram with some detail for timing elements of bridge nodes 17 or 21 which are used as deputy master or synchronization extractor. These nodes recognize when they receive an asynchronous token or frame (TM = 0) or a synchronous token of frame (TM = 1), and indicate this by a pulse on lines 39 or 41, respectively. The first synchronous token/ frame pulse after a series of asynchronous token/ frame pulses will be gated by AND gate 43 to line 45, thus indicating the start of a new period for synchronous information. Due to setting of flipflop 47 by the first synch token/frame pulse, its complement output Q will become inactive. And gate 43 us then inhibited after a short delay A and any subsequent synch token/frame pulse will not be gated. The single pulse on line 45 is transferred as synch pulse SP over line 27 to other bridge nodes in the same bridge, for synchronizing their SBM's.

The same mechanism can also be used in these normal station nodes 11 which are allowed to send or receive synchronous information. Switch Slll which in its lower position, as shown in Fig. 5, represents the situation in bridge nodes 17 or 21, would be set to its upper position in a normal station node 11. The pulse on line 45 will then be transferred to clock unit 49 for synchronizing it, and the output of clock 49 is furnished in line 51 to the other circuitry of station node 11. Thus, it needs no expensive exact clock unit, as its clock unit 49 is resynchronizing at the beginning of each synchronous-information period, by the first token or frame in which TM = 1.

A survey of the timing relationship is now given with reference to Fig. 6 and Fig. 1. First in each synchronous information period, at time to, the master SBM in node 15 of bridge B1 starts the release of a synchronous token or frame SF for ring BB and simultaneously furnishes a synch pulse SP on a line to both other S8M's in bridge B1. Bridge nodes 13 of rings R1 and R2, respectively, receive the synch pulse and release, at their respective next bit clock time (or with a given delay at a later bit clock time) a synchronous token or frame SF to their respective rings R1 and R2.

The synchronous token/frame SF on backbone ring BB propagates meanwhile to bridge B2. Bridge node 17 in B2, when it recognized reception of a first synch token/frame SF, will furnish a synch pulse SP on a line to the three other bridge nodes in B2. The delay D1 between to and release of the synch pulse corresponds to the propagation time of token/frame SF from B1 to B2, plus the recognition time for SF in bridge node 17. The three bridge nodes 13 of rings R3, R4, and R5, respectively, receive the synch pulse and release, at their respective next bit clock time (or with a given delay at a later bit clock time) a synchronous token or frame SF to their respective rings. Bridge node 17 in B2 thus acts as a deputy master or synch extractor, i.e. it synchronizes all SBM's in its bridge upon receipt of the first synchronous token or frame SF.

Token or frame SF further propagates to bridge 83 where, in the same way as described above, bridge node 17, upon reception of SF, issues a synch pulse which synchronizes the SBM in bridge node 13 of ring R6 and causes release of a synchronous token or frame SF on ring R6. Delay D2 is the sum of the propagation delay between B1 and B3 plus the recognition time in bridge node 17 of B3.

Though bridge B4 is not directly attached to the backbone ring, synchronization of the SBM in ring R7 is effected in a similar way as for all other SBM's. The synchronous token/frame released from bridge B2 to ring R5 propagates to bridge node 21 in bridge 84. This bridge node will also act as a deputy master or synch extractor, i.e. upon reception and recognition of the first token/ frame SF, it furnishes a synch pulse SP on a line to bridge node 13 of ring R7. The latter releases, after receiving this pulse, a synchronous token or frame SF on ring R7 at its next bit clock time (or with a given delay at a later bit clock time). The delay D3 corresponds to the propagation time of a token/frame SF on ring R5 from bridge B2 to bridge B3, plus the recognition time in bridge node 21.

The dashed area under the lines for ring R1 ... R7 in Fig. 6 indicates that, prior to release of each first synchronous token or frame SF, idling patterns are transmitted in the respective rings so that no additional variations can be introduced by the fact that the end of a current asynchronous frame of unknown length has to be awaited. The modified method for this early reservation of each ring will be described in connection with Figs. 7A/ B and 8.

The first synchronous token or frame SF on a ring is recognized by all nodes 11 attached to that ring, and it will synchronize the synchronous-information clock in each such node, as already explained in connection with Fig. 5.

The pattern shown in Fig. 6 is repeated at the beginning of each period for synchronous information. Thus, the synchronous-information clocks of all nodes in the multiple-ring network have all the same period which is determined by one master SBM. There are different phase shifts of the various clocks with respect to the master SBM clock, but these phase shifts remain essentially constant (i.e. the pattern of Fig. 6 does not change). The only variations possible for a given configuration, .due to the independency of the system-wide synchronous-information clock period and the ring bit periods of the individual rings, are not greater than one bit period of each respective ring.

Reservation of Ring for Timely Release of Synchronous Tokens

To enable release of a synchronous token, the SBM in each ring sets the priority indicator PI in the TC field of an asynchronous frame.

In the system disclosed in European Patent Application 0 054 077, a synchronous period start signal is issued at regular intervals, and the respective unit (CS monitor) sets the priority indicator (PI bit) in the next frame header it sees after the start signal, and then waits for the end of the current asynchronous frame before it releases a synchronous token. Thus, the start of the first asynchronous token in each period varies with respect to the actual synchronous period by less than two maximum frame durations.

To avoid these variations, a modified method is introduced in the present invention to allow issuance of synchronous tokens at almost equidistant time instants. This modified method will be explained in connection with Fig. 7A, 7B, and 8.

A watch interval is introduced whose length is equal to two maximum frame durations 2 O_max. The maximum frame duration is the time between start delimiter and end delimiter of a frame which has an information field INFO of maximum allowed length, e.g. 256 bytes. The watch interval is marked by three time instants (Fig. 7A): <ul id="ul0012" list-style="none"><li>(a) tp = start time of watch interval for setting priority indicator</li><li>(b) t_R = release time for synchronous token</li><li>(c) t = intermediate time in the middle between tp and t_R.</li></ul>

These time instants delimit two time intervals 1₁ and 1₂ of equal duration Omax.

If the SBM during intervals 1₁ and 1₂ detects an asynchronous frame (TM = 0, TK = 1), it sets the priority indicator PI in the same frame to "1". If, thereafter, it detects an end delimiter within interval 1₁, enough time is left for transmitting a complete asynchronous frame. Therefore, the SBM issues a new asynchronous free token (TM = 0, TK = 0). To avoid usage of this token by an arbitrary station,, and for ensuring that only stations downstream of the station which transmitted can seize the token, the SBM may set the priority indicator (PI = 1) in the new token thus generating a "return" token, and only the station which had already seen a PI = 1 in the-returning frame header of its own message, will reset the PI bit in the "return" token (PI = 0) to thus allow its usage by a subsequent station.

If the SBM, during interval 1₂, detects an end delimiter, not enough time is left for another asynchronous frame of maximum duration. Therefore, the SBM sends idle characters during the remainder of 1₂ until t_R, and then promptly issues a synchronous token.

If the SBM detects, during interval 1₂, an asynchronous free token (TM = 0, TK = 0), it removes the token and thereafter transmits idle characters until t_R. It then promptly issues a synchronous token.

Fig. 7B shows the respective steps and operations in a flow diagram. In this diagram, R designates the time left until t_R.

A resulting sequence of asynchronous and synchronous transmission intervals on a ring is shown in Fig. 8. The synchronous transmission intervals start at almost equidistant instants t_R. T_s is the synchronous information period controlled by the master SBM in the network. Each synchronous transmission interval is immediately followed by an asynchronous transmission interval, but a gap of variable length G1, G2, G3 ... will occur between the end of the asynchronous interval and t_R. The gap length depends on the arbitrary start time and length of the last asynchronous frame. These idle gaps are the price to be paid for the exact timing of synchronous tokens.

It should be noted that in the timing of t_R still some micro-variations occur which are at most of a duration of one bit period of the respective ring. This can be seen from the diagram of Fig. 4 where times t_R and tp have been indicated at the lines where respective pulses occur. For obtaining a pulse for t,, a further delay of duration Omax could be provided at the output of delay element 38 in Fig. 4.

Duplex Transmission of Synchronous Information Over Bridges

For duplex transmission in synchronous frames, the following should be remembered (as disclosed in European Patent Application 0054 077): <ul id="ul0013" list-style="none"><li>- To transmit synchronous information, a station has to be authorized, i.e. it must request permission to send from the SBM which grants only as many simultaneous permissions for synchronous information transmission as are tolerable on its ring;</li><li>- In a synchronous "connection" between two stations, one is the primary station which seizes the synchronous frame or slot and inserts its message, and upon return extracts the message from the other station, whereas the other station is the secondary station which does not seize a frame or slot but only extracts a message addressed to it from a frame/slot and instead inserts its own message for the primary station.</li></ul>

For the presently disclosed multiple-ring network, following rules were chosen: <ul id="ul0014" list-style="none"><li>(a) For local duplex traffic, i.e. traffic between two stations at the same ring: the calling station is always the primary station, and the called station is the secondary (same rule as already known from single-ring system of EP Application 0 054 077).</li><li>(b) For remote duplex traffic, i.e. traffic between two stations at different rings: The bridge node is always the primary, whereas an actual station, regardless whether it called or was called, is the secondary. This is necessary because in the bridge node buffers, the messages appear in random sequence and not in the sequence of currently authorized stations around the ring. On the backbone ring, which has to exchange messages between bridge nodes only, no duplex transmission is attempted because of the different random sequences of messages in the buffers involved.</li></ul>

Fig. 9 shows the exchange of synchronous messages between different rings. It is assumed that a first synchronous connection exists between stations ST-M and ST-X (Fig. 1), and a second synchronous connection between stations ST-N and ST-Y. As can be seen, on each ring appear sequences of synchronous frames at regular intervals, and between these sequences room is left for asynchronous transmissions.

The first sequence on ring R1 comprises four frames: M-X for the connection between ST-M and ST-X, N-Y for the connection between ST-N and ST-Y, and two frames LOC for local connections, i.e. between stations on the same ring. Frame M-X first transports a message sent by ST-X from the ring BB to ST-M on R1, then takes the message from ST-M to the ring BB. On the backbone ring, it travels towards ring R3 (first frame with a downward arrow). On ring R3, frame M-X transports first the message sent by ST-M to station ST-X, and then takes the message from ST-X back to ring BB, and so forth.

As can be seen, frames are used in duplex mode on rings R1, R3, R6 etc., whereas on backbone ring BB, each frame transports only one message (simplex mode). This is favorable because of the random sequence of messages in buffers, as was mentioned earlier. As indicated by the smaller width of the frames on ring BB a faster transmission speed can be provided for the backbone ring to compensate for the simplex mode (greater number of frames) and for the possibly higher volume of traffic with respect to normal rings.

Alternatives and Modifications

Phase shift between synchronous frames on different rings: <ul id="ul0015" list-style="none"><li>As shown in Fig. 9, the sequences of synchronous frames start simultaneously on rings R1, R3 and R6, but with a phase shift of one-half synchronous period on backbone ring BB. This is an alternative to the solution shown in Fig. 6 where the synch frames start simultaneously (except for short delays) on all rings.</li></ul>

The alternative solution with phase shifts can be introduced to minimize the total transmission time between source and destination. As shown in Fig. 9, a message which travelled on ring R1 during period n is transported during the same period on backbone ring BB, and it travels already during the next period n + 1 on its destination ring R3. If synch frames start simultaneously on all rings, the transmission on backbone BB would require an extra period, i.e. a message travelling on R1 during period n would travel on R3 only during period n + 2.

In general, the delay between neighbour rings in a series of interconnected rings, i.e.those interconnected by a bridge, should be one-half synchronous-information period, and the frames for synchronous information should occupy less than half of the total transmission capacity or time (i.e., no overlap of synchronous frames, as shown in Fig. 9). For the network of Fig. 1, the group of rings R1, R2, R3, R4, R5, R6 should start synchronous frames simultaneously, whereas backbone ring BB and ring R7 should start with a phase shift of one-half synchronous period.

Location of Master SBM: <ul id="ul0016" list-style="none"><li>In the embodiment described, the master SBM was the SBM of the backbone ring. This will probably ensure the shortest delay for the synchronization information that must be propagated from the master SBM to all other SBM's. However, any one of the SBM's of the normal rings R1 ... R7 could be chosen to be the master SBM. In a series of interconnected rings without a backbone, which was mentioned earlier as a possibte configuration in which the invention can be used, it is anyway necessary that one of the normal ring SBM's is chosen to be the master SBM.</li></ul>

Connection to public network: <ul id="ul0017" list-style="none"><li>One of the rings, e.g. the backbone ring BB, may be connected through a gateway to a public network. A gateway is an interconnection unit similar to a bridge but probably with more general functions. The respective ring will have a gateway node in the gateway, as it has a bridge node in each bridge to which it is attached. If the public network is a fully synchronous network, e.g. a PCM telephone network, the master SBM of the interconnected rings is best located in the gateway node, and it should be synchronized itself from the public network via the gateway. Synchronization within the multiple-ring network is then possible from the master SBM in the same way as was described above.</li></ul>