WO1989008363A1

WO1989008363A1 - Telecommunication system for transmitting information between subscribers connected to a bus system

Info

Publication number: WO1989008363A1
Application number: PCT/SE1989/000094
Authority: WO
Inventors: Erik Ernst Hagersten; Lars Gösta GAUFFIN
Original assignee: Telefonaktiebolaget L M Ericsson
Priority date: 1988-03-02
Filing date: 1989-03-02
Publication date: 1989-09-08
Also published as: SE8800745L; SE460750B; SE8800745D0

Abstract

The invention relates to a telecommunication system for transmitting speech and data information in time divided form through high speed buses (DB) to which are connected a plurality of nodes, each of which is in turn connected to a plurality of subscribers who can have an extremely varied need of bandwidth. Bridges connect the buses with each other so that a matrix network (M) is obtained in which all subscribers have the opportunity of communication with each other, and in which each node has the possibility of being connected to at least two different buses in the network. Information on the buses is divided into static and dynamic time slots. Each node is allocated at least one static time slot so that every node is given fixed connection to every other node on the common bus. Dynamic time slots are also allocated to the nodes in proportion to their particular bandwidth needs. If required, each node can be allocated unutilised dynamic time slots from other nodes. Each node includes a node handler (NH) which converts and exchanges information under the control of a node processor (NP) between subscribers, nodes and buses. For transmitting information between buses a node handler (NH) is connected as a bridge at each of the intersection points in said matrix network.

Description

TELECOMMUNICATION SYSTEM FOR TRANSMITTING INFORMATION BETWEEN SUBSCRIBERS CONNECTED TO A BUS SYSTEM

TECHNICAL FIELD

The invention relates to a telecommunication system in which information is transmitted between subscribers connected to a common, high speed bus for use from small, up to very large bandwidths.

BACKGROUND ART

It is known from such as the US Patent No 4383315 to transmit information between terminals through common buses. Systems of the mentioned type are utilised to a large extent for packet switching of data.

Packet switching involves the transmission of information in the form of shorter or longer packets. Depending on the network load, these packets can select different paths to their destination.

A further switching type is so-called circuit switching. Circuit switching involves establishing a connection the whole way between transmitter and receiver. This connection is kept established during the entire transmission time, even if there are pauses in the transmission of information.

DISCLOSURE OF INVENTION

A problem in the known art in this field is that it is not possible to transmit . information requiring very varying bandwidths through one and the same medium. In telephony technique, telephone exchanges of different kinds are also usually used for distributing information between subscribers via a group selector. Neither circuit nor packet switching satisfies the needs of integrated switching of speech, data and interprocessor communication.

Packet switching suits the speech part very badly, and in addition considerable bandwidth is lost due to collisions on the bus, which result in retransmissions. Circuit switching suits the speech part very well. On the other hand, its static nature suits data transmission very badly. If each node were assigned a static bandwidth, the bus capacity would be poorly utilised, and the capacity of the bus to suit different types of system would be heavily restricted.

The apparatus in accordance with the invention, which solves the above- mentioned problems, is characterized by the claims and includes a bus system including very fast optical buses for transmitting speech and data information between subscribers without utilising a group selector. A bus of the kind mentioned can be used, e.g. in a distributed telephone system, for connecting a plurality of distributed nodes connected to the bus. In addition, several buses can be connected to each other via so-called bridges, and thus form a complete exchange. The bus replaces group selectors and also regional processor buses in known exchanges, and expedites speech and data communication as well as interprocessor communication between the nodes. Control of the information yield between nodes takes place with the aid of a special communication protocol (DUPER). The protocol enables exchanging speech and data through a high speed bus, e.g. 2,4 GHz, with a dynamically variable bandwidth where the individual node only needs to handle its own part of the bandwidth.

BRIEF DESCRIPTION OF DRAWINGS

The communication system in accordance with the invention is described in more detail below with the aid of an embodiment example and with reference to the accompanying drawings, where

Figure 1 is a block diagram of a system in accordance with the invention, Figure 2 is the division of a bus into static and dynamic time slots, Figure 3 illustrates the format for a packet in accordance with the protocol according to the invention,

Figure 4 is a matrix of a network set up in accordance with the invention, Figure 5 is a block diagram of an example of the construction of a node in the network. BEST MODE FOR CARRYING OUT THE INVENTION

A communication system is illustrated in Figure 1 where a plurality of communication nodes N1-N4 are connected to a common high speed bus HB1. In turn, the bus is connected via a bridge unit B to a further high speed bus HB2, which in turn is connected to a plurality of communication nodes BN1-BN3. Subscribers with extremely varying requirements as to bandwidth, from ordina¬ ry telephony to transmission of television images are connected to nodes Nl- N4, which also applies to the nodes BN1-BN3.

By connecting buses via bridges, geographically very separated subscribers can

Φ- be connected together in large networks. The exchange of the future will still cover ordinary telephone traffic to a major extent, even if the greatest communication bandwidth will be occupied by data communication. According¬ ly, the bus must be able to handle simply and cheaply both static calls with a small bandwidth and data sent in bursts and requiring larger bandwidths for the bursts. For a long while, transmission of speech has taken place on time-divided buses where samples from several calls are transmitted during a first frame, the next sample during the next frame etc. Time-divided buses have also been used as time switches, where several unit on the same bus communicate with each other by reading and writing in special time slots.

The rapid fibre buses operating at 2,4 Gbit/s, open completely new fields. The high bit rate would make the hardware unreasonably expensive if traditional protocols were used for controlling the information flow on the bus, and the protocol in accordance with the invention is therefore adapted to high bit rates, and can also be utilised for connecting wider bandwidths than required for speech in the same way as a speech connection is set up.

Traditionally, all units are allocated the same bandwidth, whether it is required or not. In accordance, with the invention, the different units on the bus are allocated bandwidths of different magnitudes. The allocation can also be changed during operation with a very large dynamic response. The dynamic response of the bus therefore makes it suitable to transmit data and inter- processor communication as well as speech. The protocol in accordance with the invention is based on dividing the bus up into a plurality of time slots, each of which contains a command word and a plurality of data words, where each data word can correspond to a call. These time slots are what are time-switched on the rapid bus, not each call by itself. The command word gives changes, e.g. connection or disconnection of a call, which can comprise one or a plurality of data words. According to the protocol, each communication frame of 125 rt_^s is divided into a plurality of static and a plurality of dynamic time slots, see Figure 2.

Each unit will have a fixed time slot connected towards every other unit on the bus, which constitutes the static part of the frame. The rest of the frame is the dynamic part. These dynamic time slots are divided between the units at the initiation, so that different units can be assigned differently large transmitter- allocated bandwidths. Using a command word, e.g. in the static part, the transmitting unit can give the receiving unit an order to begin listening to a new dynamic time slot. The protocol also includes a distributed resource allocation, whereby one unit can ask the other units for more bandwidth when it has used all its own dynamic time slots.

The principles for time multiplexing and PCM are not accounted for in this application since these principles are known.

The invention relates to a bus protocol which can be characterized as a dynamic time slot protocol. The mechanism itself will be presented below. According to the example, the number of nodes is limited to 24. A change of the number of nodes only changes the relationship between the static and dynamic parts of the bus.

A static as well as a dynamic bus has advantages in certain situations. The apparatus in accordance with the invention is a combination which puts together the good properties of the buses. The bandwidth is divided into shares in the proportion of about 1/4 static and about 3/4 dynamic.

In order not to make the dynamic allocation of bandwidth ineffective, the bus is divided into larger bits, so-called time slot allocated packets. As illustrated in

Figure 3, each packet comprises 12 8-bit words (two command words, eight data words and two fault correcting words). On initiating the system, all nodes are given a static time slot connected to all other nodes. This is the so-called static part of the bus, which is fixed and takes up approximately 1/4 of the bandwidth. If the number of nodes is 24, the number of static slots will be 23 x 24 = 552. The dynamic part will then be about 1500, i.e. 2k slots per frame. With the bandwidth in question, the frame lenght would be 125 fl s. See Figure 2.

A node's static part of the bus will be sufficient for transmission of 8 calls to all other nodes, or 0,5 Mbit/s. Connection and disconnection is controlled by the command word of the static slot. If more than 8 calls, or a wider data communication of the packet type is required, the node can ask the receiving node to begin to listen to a specific dynamic packet, and use this for the transmission. Connection of this packet takes place by the command word "Begin to listen to a packet" being sent on the static packet wich goes to the node in question. In this way, the bandwidth to other nodes can be increased as required with an extra dynamic packet per frame (changing the bandwidth by 4

Mbit/s per ms). Increasing the bandwidth by 10 Mbit/s towards another node would take 2.5 ms.

When initiating the system, the dynamic packets are divided evenly between the nodes, i.e. in this case each node has about 64 dynamic packets. If a node's dynamic packets begin to be used up, a request is sent to the node with the nearest higher number to send an unused slot. The answer, with the number of the slot in question arrives after a couple of frames. In this way there will be a flow of unoccupied slots which can circulate, if required.

It should be emphasised that resource distribution only takes place when there begins to be a lack of slots at some node.

The bus thus adjusts itself to a suitable bandwidth distribution between the nodes.

The dynamic time slots have the same appearance as the static ones, with two command words, eight data words and two fault words. The dynamic command word can make one connection per frame within its own packet. A connection can apply to one or up to all eight words in a packet, i.e. 64 kbit/s to 0.5 Mbit/s. The dynamic slot will thus always control itself. This results in that the possibility of a node to control communication increases at the same rate as Its bandwidth. This also results in that control takes place at two levels, one which adjusts the bandwidth and another which sets up the data bits.

The command words which control connection and disconnection in the static and dynamic packets have the following format:

<"l"><START(3)><WIDTH(2)><DESTαθ)

Where START denotes which of the eight slots which is to be the channel's first slot. WIDTH denotes whether the channel shall embrace 0, 1, 4 or 8 consecutive words and DEST (destination) denotes what individual on the node which is to be the receiver of the packet. DEST is used as global node address.

The command words in a static packet which change the bandwidth between different nodes can have the following format:

<"0"><ORDER(3)><SPARE(l)xNUMBER(ll)>

Where ORDER encodes one of the commands:

"Begin listening to packet <NUMBER>" "Finish listening to packet <NUMBER>" "I want an unoccupied packet" "Here is an unoccupied packet <NUMBER>" "I am serivceable <STATUS CODE>"

"ERROR CODE" <ERROR CODE>

SPARE can be used for increasing one of the fields, as required. The NUMBER fields are used to denote a numeral 0-2k. The STATUS CODE and ERROR CODE advise other processors as to how hard the load on a node and its sub levels are, and which type of fault or data error has afflicted the node. This results in that the other nodes can be rapidly updated if anything serious is in the process of occuring. The advantages with the proposed solution are, inter alia: STATIC STRUCTURE. The bus has a static structure, where communication between specific nodes takes place in specific time slots. A certain (dynamic) part of these slots can be permitted to change slowly. This "dynamic-static" structure enables demands on the rapid part of the hardware (GaAs) to be kept at a reasonable level.

CONSISTENT FORMAT. Static and dynamic slots have the same format.

DYNAMIC BANDWIDTH. The bandwidth can be changed by increasing or decreasing the number of dynamic time slots between two specific nodes. A connected bandwidth is, however, not disconnected until some other node needs the resource. This is an important property in data communication where data is sent in bursts.

MODULAR CONTROL AND DATA RELATIONSHIP. The control bandwidth increases at the same rate as the data bandwidth.

HIERARCHIC CONTROL. Two control levels have been introduced. A top level which handles basic communication and adjustment of bandwidths, and a local level which handles connection and disconnection within a packet.

ERROR DETECTION. All nodes transmit at least one packet to every other node during a frame. If there is nothing to transmit, the code "I am serviceable" is sent, which results in that a faulty node will be discovered by all the other nodes within one frame.

ERROR CORRECTION. Error correction has been introduced, which results in . that nearly all communication errors can be detected and corrected as well.

DISTRIBUTED RESOURCE MANAGER. A simple concept for distributing the control of free slots.

SELFADJUSTMENT OF BANDWIDTH. A bandwidth which has been set by the resource manager will remain until the needs have been changed. FREE NUMBER OF NODES. An alteration of the number of nodes results in that the ratio of static to dynamic slots is changed. If freedom from congestion is desired, only a small number of nodes is connected. On the other hand, if cost effectiveness is important, a larger number of nodes can be connected.

Information through the buses comprises a serial bit phase-coded bit stream containing clock information and data as well as sync pulses. Sync pulses set the bus times so that a time-divided bus is obtained. The time part or frame is divided up into static and dynamic time slots. According to the example, a frame consists of 16 bit control words divided into two 8 bit words and a subsequent, selectable number of 8 bit PCM samples. The protocol format according to the example includes 8 PCM samples per frame.

Example of frame construction:

N 3 2 2 1

The frame length (bit content) is dimensioned by the node handler's implemen¬ tation. The form of the control word (COM) is determined by the properties in the envisaged protocol. All appear to be indentical in size.

The time slots on the bus are divided, as mentioned, into a static group and a dynamic group. The nodes are then allocated an array of static slots which have a destination defined in the requirements for reaching remaining nodes.

An example with three nodes is as follows:

The first NODE one time slot to to the second node and one to the third node, second NODE one time slot to the first and one to the third, the third NODE one time slot to the first and one to the second.

Dynamic time slots are allocated to the nodes but have no destination given. With the aid of the protocol these free slots can be given destinations to nodes as required.

Example:

Nodel Node 2 Node3

b = Static time slots that have been given destinations.

A status word "I am serviceable" is contionously transmitted on the bus and shows that the transmitting node's behaviour is under control. If no command information is carried, the communication remains in its previous state.

A a network with a very large capacity is created by tying together several buses in a grid pattern with the aid of bridge nodes.

Via the static time slot each node can now select what node it wants to communicate with on its two buses. This is internal information which never needs to leave the node, since it reaches all nodes along both X and Y axes via the static time slot. The receiving NODE knows which node is sending, due to the fixed part of the frame being defined by its time position.

If the connection of a node to a bus applies to one (1) PCM channel, or the corresponding bandwidth within the frame, the message can be received and the subscriber in the receiving node can be pointed by 10 bits in one operation along its own buses. If it is not a question of the local bus nodes, allocation and over- bridging to the next bus can be catered for by ORDER.

Since not all bits are utilised in pointing out the node (only 10 of them), a number of the further 6 bits can be used for pointing out subscribers, and in the example, this increases the pointing capacities from 1000 subscribers per node to 32,000 subscribers per node.

In setting up a network a maximum configuration can theoretically be obtained with an ORDER word and a STATUS word of 16 bits each, 5 + 10 bits being used for pointing out where this configuration is:

32*32 nodes * 32000. 32 million subscribers 24*24 nodes * 32000. 18 million subscribers

This requires two 125 Λ- s frames. If it is desired to manage the setting up on one frame, 1000 pointings per node can be selected. As previously mentioned, the protocol is to give a node, the computer power of which does not need to correspond to the speed of the fibres, since only a small part of transmitted data is received at the node. The re-allocateable bandwidth gives the ability of flexibly utilising the transmission capacity of the bus. The protocol allows distribution of bandwidth sparingly, since there is the facility of having all nodes listening on the same time slot or an optional number of time slots. Another advantage is that via only two buses there is the possibility of reaching any other node in the network and of always having two paths to it. The network may be regarded as a distributed switch or an intelligent transmission network. Advantages such as reduncancy and modularity are also achieved by the invention.

The system in accordance with the invention is designed as a communication mechanism in a distributed telephone system, but has also been found utilisable within mulitprocessing.

The properties which have been sought after are:

A network with a simple addressing algorithm, i.e. a routing table will not be needed.

As short path between all nodes, i.e. it is needed to pass only a minimum of

BRIDGES when a channel is connected between two optional nodes.

There shall be several optional paths, for reasons of safety.

Rapid and simple way of connecting a channel between two nodes. This is also for meeting the demands of interprocessor communication through the network.

Large bandwidth dynamic between nodes, without using a central control.

Each node is equipped with at least two ports to different buses, so that a single bus error will not be able to make several nodes unserviceable. Very little hardware is required for allowing these buses to communicate with each other via a direct channel also. A NODE HANDLER is required to enable this and the function of the handler will be described in more detail in connection with

Figure 5.

There is only one type of node handler in the system. The node handler has two ports towards the mentioned DUPER buses, and one port towards a node processor. This means that the node handler can serve as a BRIDGE between two buses, simultaneously as there is an unoccupied port for a node processor.

As will be seen from Figure 4, the network is built up as a matrix network M max 24 x 24 with a node handler at each intersection point. This gives a system with a maximum of 576 nodes, where connection between two optional nodes can be set up along two different paths which only pass one bridge on the way. The two paths are obtained by first going to the right X co-ordinate and then to the right Y co-ordinate, or vice versa. Of course, such a network is not dependent on routing tables or the like.

The system is built up round 24 horizontal Y buses and 24 vertical X buses of the DUPER type, denoted DB. These are zero numbered from the left and from below, respectively and are denoted, e.g. "X bus 1" for the second horizontal bus from below. The node names are formed by their X and Y bus numbers (X and Y co-ordinates) and are denoted (1,14) for node 14 on the X bus 1 and node

1 on the Y bus 14. In the network, the code name is given in 10-bit binary form, where the first 5 bits denote the X and the last 5 bits the Y co-ordinate. Node (1, 14) is given in binary form as 0000101110.

Each intersection point between the buses is controlled by a node handler, denoted NH. NH transfers data words from one bus DB to the other bus DB or to the node's local node processor, denoted NP.

Communication through DB takes place in packets of 12 words. There are two Command (COM) words, 8 Data words (DATA) and two words which are CRC checks (CRC). A whole packet is addressed to the same node handler NH. The different data words can be addressed to different destinations, however.

Approximately 2000 PACKETS are sent within a FRAME of 125 / s on each bus, i.e.

16000 calls.

Node processor NP leads and distributes the work at the nearest lower level, which is assumed to be a bus of the DB type, although slower. Each NP serves about 500 subscribers.

Messages which are not addressed to the node processor NP are passed on to the second bus DB by the node handler NH. This passing on action is called the "bridge function". It should be noted that a bridge connection always involves a right-angular change in course. In Figure 5 there is illustrated the construction of a node handler NH. As will be seen from the Figure it is possible, for increasing the degree of freedom, to connect more than one node handler. The example shows, with reference to the matrix network, two horizontal node handlers and two vertical node handlers. Input logic IL from the fibre bus DB to the node handler NH is implemented in

Gallium Arsenid (Ga As) and includes, inter alia, an opto/electronic converter for converting the fibre bus's optical signals to electrical signals and a serial- to-parallel converter for converting the serial information flow of the bus to information in parallel form. A clock unit CL provides the system with necessary timing signals. The receiver R receives incoming signals and sends them to a demultiplexer D, which in turn sends the information to a first buffer memory BM1 serving as a buffer between the incoming fast information flow and a slower internal bus IB. A second buffer memory BP adjusts the information flow between internal bus IB and node processor NP. A third buffer memory 3L adjusts the information flow between the internal bus IB and a local bus LB, to which the subscribers associated with the node are connected.

In the other direction, when information goes from the subscriber through the node NL, out on the bus DB for connection to other nodes on the common bus or for bridging over to another bus, the information passes the third buffer memory BL and, in accordance with the example a further buffer memory BM2.

The information is supplied to the bus DB via a multiplexer M, a transmitter T, parallel-to-serial conversion and electric-to-opto conversion in the unit IL, which thus also includes output logic. A bypass unit FK allows bypassing the send logic of the node handler when the node itself does not have anything to transmit. The node processor NP, which may be a microprocessor type M 68000 made by MOTOROLA, conventionally controls the information yield through the node. Up-to-date data on the information flow is stored and read out from the control memory CM. A time slot counter C serves as a pointer for writing into and reading out from the memories. As will be seen from Figure 5, it is possible to connect several node handlers to the node, and for example the node handlers

1 and 2 are assoicated with horizontals in the previously mentioned matrix network M, and node handlers 3 and 4 are associated with verticals.

Some methods of routing through the system are described below. Due to the simple structure and numbering of the network, there will be no need for routing tables. Both two-stages X-Y routing paths will be tried first. If these give errors or congestion, a three-stage routing can be used, which gives a further 64 possible paths.

CONNECTION OF TWO-STAGE ROUTING

When a node processor NP attempts to set up a connection towards a node, the logical name of the receiver is translated to the physical X-Y name. The NP sends the channel inquiry including the physical name as well as the bandwidth in question and an identifier unique for the node to the node handler NH. After a given time, the node handler NH will answer with the same identifier and a clearing sign for commencing transmission. The bandwidth for a channel can be

64 kbit/s - 500 kbit/s. If a larger bandwidth is desired, the node processor NP must request more channels than it administrates itself.

The node handler NH investigates whether it has a sufficient bandwidth towards the indicated BRIDGE. If this is not the case, it connects one of its unoccupied packets towards the bridge during the next frame.

THREE-STAGE ROUTING

Three-stage routing can take place via any of the other nodes along the two DB to which the node handler NH is connected. From old status reports the node handler can see which node which appears to be best suited. The "geographical" location of the nodes will not make any difference. If necessary, a new slot is set up towards the selected node, and a channel is set up in the usual way by a connection command being sent together with a 10 bit destination code. The receiving node will discover that the channel cannot be set up to its own node processor NP, and will connect the channel further at right angles. The same destination code will be used, which means that a normal two-stage routing is started at this point.

A network in accordance with the invention has very good safety against bus errors. Occasional errors will be put right by the CRC word. If a bus fails completely, it does not stop a single node even so, since all nodes on the bus understand that the bus is unserviceable due to the absent "I am serviceable" messages, and use the alternative rout.

If two X-buses fail, this does not cut out any node. If an X-bus and an Y-bus should fail, only the node in the interception point is cut out.

As will be seen from the description, the telecommunication system in accordance with the invention presents an unique solution to the problem of transmitting information of extremely varying bandwidth, moreover without utilising group selectors, as in today's telephone systems. In addition, the hardware costs are kept to a very low level by suitable selected parameters.

Claims

C L A I M S

1. Telecommunication system in which speech and data information in time divided form is transmitted through buses to which are connected a plurality of nodes, each of which is in turn connected to a plurality of subscribers with varying bandwidth needs, and where bridges connect the buses with each other so that a matrix network is obtained, in which all subscribers communicate with each other, and in which each node has connection possibilities to at least two different buses in the network, characterized in that for enabling transmission of information with small bandwidth needs as well as information with very large bandwidth needs, the information transmission on the buses (DB) is controlled by a protocol with a format adapted to the transmission, in that each node (N) is allocated time intervals including static and dynamic time slots, said time slots including at least one command word and a plurality of data words, where said words can each correspond to a call or other information in an established connection, in that at least one static time slot is allocated to each node so that each node is given fixed connection to every other node on the bus common to the nodes, in that said nodes are also allocated dynamic time slots in relation to their particular bandwidth needs, and in that each node is allocated as required, unutilised dynamic time slots from other nodes.

2. Telecommunication system as claimed in claim 1, in which speech and data information in time divided form are transmitted through buses to which are connected a plurality of nodes, each of which is in turn connected to a plurality of subscribers with varying bandwidth needs, and where bridges connect the buses with each other so that a matrix network is obtained in which all subscribers can communicate with each other, and in which each node has possibilities connection to at least two different buses in the network, charac¬ terized in that each node contains a node handler (NH) which converts and exchanges information under the control of a node processor (NP) between subscribers - nodes - buses or - buses - nodes - subscribers, and in that a node handler (NH) is connected as a bridge at each of the intersection points in said matrix network (M) for transmittng information between buses.

3. Telecommunication system as claimed in claim 1, characterized in that said buses (DB) are high speed buses implemented as optical fibre buses.

4. Telecommunication system as claimed in claim 2, characterized in that sending information further via a bridge always involves a right-angular change of direction.