AU657176B2 - Method and circuit for controlling access to an asynchronously oeprated network - Google Patents

Description

P100/011 28/5/91 Regulation 3.2 657 17

AUSTRALIA

Patents Act 1990 r r r r~ r r r *r

ORIGINAL

COMPLETE SPECIFICATION STANDARD PATENT Invention Title: "METHOD AND CIRCUIT FOR CONTROLLING ACCESS TO AN ASYNCHRONOUSLY OPERATED NETWORK" The following statement is a full description of this invention, including the best method of performing it known to us:- This invention relates to a method and circuit arrangement for controlling the access to an asynchronously operated first network which allows access to a further network by means of a networking unit.

This invention arose during investigations on "Local and Metropolitan Area Networks" according to the Standards Proposal IEEE 802.6. The method of operation of these networks is similar to that with Asynchronous Transfer Mode (ATM). With ATM, the transmission of data takes place in such a way that during a connectionestablishment phase, first a connection path, and then a defined traffic capacity along this path, is reserved, and that subsequently the incoming data is separated into equallength blocks and transmitted over this path. Each such block has a "cell header" inserted before it and, together with this, forms a "cell". The cell header contains the necessary information to characterise that this cell belongs to a particular connection.

The cells of a particular connection are transmitted, together with cells of other 15 connections, with empty cells, and with control cells, over the same transmission S medium. A fixed allocation does not exist within the total data stream between the cells of a given connection and particular time positions; contrary to the customary timedivision multiplex, the transmission takes place asynchronously. This operating method is particularly suited to data transmission, for which the traffic often occurs only in bursts.

This traffic occurrence in bursts is quite well equalised in higher network levels, where the traffic from very many subscribers is combined, so that there a clear advantage is gained by asynchronous operation, compared to synchronous operation.

However, in lower-level networks closer to the subscribers, no such equalisation S 25 applies.

For the reservation of the required traffic capacity during the connectionestablishment phase, either a time mean value or a maximum value is usually specified.

If capacity is reserved everywhere for the maximum value, then this is very uneconomic; if the mean value is chosen, more or less lengthy overloading can occur.

Such overloading leads to losses, unless they are trapped by buffer stores. But such buffer stores are undesirable.

On the other hand, the above-mentioned proposal for a standard has no provision for a connection-establishment phase. On the contrary, it assumes that the possible traffic conditions between the subscribers or groups of subscribers are predetermined by the network controller. There is no provision for a switching function.

However, this is irrelevant for the present invention; it is just as suitable for ATM systems with switching facilities.

For both systems a common problem emerges from the above comments: because of the random behaviour of the information source, it can happen at particular places in the system that an overload situation occurs, because either the physically available capacity or a predetermined threshold is exceeded. In this regard, the problem areas lie in the transitions between two networks, parts of networks, or network levels.

In the previously mentioned example of the "Local and Metropolitan Area Networks" in accordance with IEEE 802.6, the subscribers are connected to Metropolitan Area Networks (MANs), either directly or via Local Area Networks (LANs).

The subscribers ccnnected to the same LAN can communicate with each other without S the use of MANs. The MANs themselves consist of interconnected sub-networks.

Transitions into the public network and into data networks are also provided.

0 Every one of these sub-networks (TNs) consists of two bus systems by which the LANs and the subscribers are connected to each other, and connected to transitions into other networks. The TNs in principle do not distinguish between connections to LANs and to (individual) subscribers. Like a TN, a LAN can contain two bus systems.

The bus systems of a TN carry the whole of its traffic which comprises both internal and external traffic. Access to the bus systems of the TN is controlled by a network access procedure which is currently described in the above-mentioned Standards Proposal IEEE 802.6. This network access procedure, however, only guarantees that the total traffic in this TN does not exceed the given limits, not even briefly. Since at network interworlking units in general the total traffic cannot be transferred into the other network, it can happen that a portion of the external traffic has access to the TN, 25 but cannot leave it at the network interworking unit, and is thus lost.

In principle, the same problem also occurs with respect to the traffic destined for an individual subscriber (or an individual LAN), since each subscriber can fundamentally be connected to several other subscribers which can be transmitting (quasi)simultaneously to that subscriber, where the TN is not overloaded but the individual subscriber is briefly overloaded.

With other systems also, where several transmitters and receivers are connected via a common transmission medium, for example a multi-processor system connected to a common bus, the same problem can occur, in so far as these systems are not operated synchronously, that is with indivioual channels of fixed predetermined capacity.

The solution to be described is usable with all systems of this type. The concepts used here, "first network", "further network", and "network interworking unit" are to be understood in a very general sense.

According to the invention there is provided a method for controlling the access to an asynchronously operated first network which allows access to a further network by means of a network interworking unit, wherein equipment via which the first network is accessed is informed about the current loading of the network interworking unit.

According to a further aspect of the present invention there is described a circuit arrangement for controlling access to an asynchronously operated first network which allows access to a further network by means of a network interworking unit, wherein the circuit arrangement has a register (RQ, CD) for holding information about the loading of the network interworking unit.

.:i The concepts "first network", "further network", and "network interworking unit" are to be understood in a very general sense.

The basic concept consists of making the access to a transmission medium dependent not only on the current loading of the transmission medium itself, but also on the loading of the target receiver, or intermediate receiver. For this purpose, all accesses to a sub-network are continuously informed about the current loading at a network interworking unit leading out from this sub-network.

The invention will be described with the example of the DQDB System (DQDB Distributed Queue Dual Bus) as proposed in Standards Proposal IEEE 802.6. As claimed in the invention, the subscriber terminals of this system separately identify requests for time slots for external traffic.

In order that the invention may be readily cai-ried into effect, an embodiment thereof will now be described in relation to the accompanying drawings, in which: Figure 1 shows an example of a Metropolitan Area Network; Figure 2 shows the basic construction of a DQDB network Figure 3 shows the data structure of a DQDB network, where Figure 3a shows a Slot Format, Figure 3b an Access Control Field, Figure 3c a Segment and Figure 3d a Segment Header; Figure 4 shows the access control for a bus, Bus A, of a DQDB network; Fig 5 shows the process of access control, where Figure 5a shows the ongoing monitoring state, and Figure 5b an access procedure; Figure 1 shows the Metropolitan Area Network MAN given as an example in the Standards Proposal mentioned above. This MAN has several sub-networks TN. These sub-networks, TN1, TN4 are of the DQDB type. A Work station WS, a Host Computer HS and a LAN LR, for example a Ring Network, are connected to these TNs.

The TNs are interconnected, partly direct and partly via a Multiport Bridge MB. In addition there are transitions to a (the) public line switching network PCS and to a (the) public packet switching network PPS. Not shown is the fact that several subscribers are connected to each individual TN, where there is in principle no difference between the connection of a subscriber and the transition to another network.

Figure 2 shows the basic structure of a DQDB network, that is a simplified form of the TNs TN1, TN4. The dual bus, consisting of the two busses Bus A and Bus B, can be seen. Also four subscriber equipments TEl, TE4 and a network node NK are S shown. A network interworking unit NI is also shown. All subscriber equipments and the network node have access to both busses, with the two busses being driven in opposite directions. The network node does not necessarily have to be at the end of the 15 busses, being of equal ranking with the subscriber equipments in this regard. The network node only differs from the subscriber equipments insofar as the traffic to or from it is transferred to a network of equal or higher hierarchy level, wheras the subscriber equipments can only be considered as networks of a lower hierarchy level.

Because of the random occurrence of the traffic, it is possible, both for the network node as well as for the subscriber equipments, that they cannot accept the traffic S destined for them, at least for a short period, although it does not overload the TN S itself, i.e. the Bus A and Bus B.

Figure 3 is subdivided into Figure 3a to 3d and shows the data structure of a DQDB LAN. Figure 3a shows the basic structure of the smallest transmissible 25 information unit, denoted Slot Format by IEEE. Each slot comprises 53 octets, the first octet being a field for access control, Access Control Field ACF, while the remaining 52 octets are called a Segment, which serves for the transmission of information.

Figure 3b shows the more detailed structure of the Access Control Field ACF.

The first bit, BUSY, shows the busy status of the Segment following the ACF. The second bit, SL TYPE, together with the first bit shows the type of occupancy of the corresponding Segment. The following bit, PSR, is not relevant to the present invention.

This is followed by two not yet allocated bits, RESERVED, as well as three bits, REQUEST, for requesting free slots.

As Figure 3c shows, a Segment consists of a four-octet header, Segment Header, and 48 octets for useful data, Segment Payload. The Segment Header serves to associate the whole Segment with an arbit'ary procedure, perhaps a control process or a specified connection. The Segment Header is shown in more detail in Figure 3d. The first 20 bits represent a kind of connection number, the Virtual Channel Identifier VCI.

Four bits follow for characterising a specific data type, PAYLOAD TYPE, and priority, SEGMENT PRIORITY. The last eight bits, Header Check Sequence HCS, serve for error detection and error correction.

As with some other systems, the DQDB system provides the possibility of permitting synchronous as well as asynchronous connections. Synchronous connections assune a periodically recurring frame structure for the data stream. With the DQDB system, the frame structure is predetermined by those subscriber equipments or network nodes from which the busses start. In the example of Figure 2 this is NK for SBus A and TE4 for Bus B. For syncihronous connections, the network controller provides slots and allocates them, and thus possible traffic conditions are established and, later, °o:i preset for each frame, within the frame, by the unit generating the frame structure, here TE4 and NK. For this purpose, in the Access Control Field of this slot the BUSY bit and the SL TYPE bit are set to 1, and the Segment Header is filled. The stations taking part in such a synchronous connection can now identify the capacity allocated to them, either by the time position within the frame or by using the Virtual Channel Identifier VCI, and can insert the information being sent into the appropriate position in the data stream, or extract it at the receiving end. Since for synchronous traffic conditions definite time slots are reserved everywhere, in both sender and receiver as well as in S the network and network interworking units, no collisions are possible, and no losses can occur due to overloading.

With asynchronous traffic conditions, slots must be allocated according to the 25 current requirements. This can cause collisions at various places. With the DQDB system, collisions when accessing a network are avoided as described below. The present invention provides a proposal for a solution for overloading with network interworking units and during reception.

Figure 4 shows the access control for Bus A of a DQDB system. The access control for Bus B operates similarly. In a somewhat different representation to Figure 2, it shows the two busses, Bus A and Bus B, as well as the three subscriber terminals TE3, TE2 and TEl. The data flow on Bus A is shown at the top, the data flow on Bus B at the bottom. Two slots on Bus A are shown, and two on Bus B. The BUSY bits B in the slots on Bus A and the REQUEST bits R in the slots on Bus B, are relevant for the access control for Bus A. Only these are shown in Figure 4. The fact that three REQUEST bits per slot are actually provided, can be ignored here in the explanation of their function (and is also ignored at this stage in Standard Proposal P802.6).

Each terminal (TE3, TEl) learns on Bus B how many slots have been requested by thr. terminals preceding it on Bus B. These slots requested for Bus A must now be kept frEe on Bus A. Of the slots still free on Bus A, as many must therefore be left free as have been requested by a subsequent terminal on Bus A. Only then can a terminal occupy, for itself, one of the slots passing by.

With the aid of Figure 5, the process will be described in more detail using the example of subscriber terminal TE2. The whole process is divided into an ongoing monitoring state, Figure 5a, and the access process itself, Figure 5b. The subscriber terminal TE2 contains a Request Counter RQ which is incremented by requests on Bus B (REQUEST and is decremented by free slots on Bus A (BUSY Thus it becomes known at any time how many open requests still remain at the terminals following on Bus A. However, the subscriber terminal TE2 does not have to wait until there are no outstanding requests left at the following terminals. It is in fact sufficient that there are no older requests outstanding. If the subscriber terminal TE2 has a request of its own, it sets a still un-set REQUEST bit on Bus B, transfers the current count of the Request Counter RQ into a Supervisory Counter CD and resets the Request Counter RQ to zero. The Supervisory Counter CD is now decremented by the free slots (BUSY 0) pass.,g by on Bus A. When it reaches a count of zero, TE2 can occupy the next free slot. Then control returns to the Request Counter RQ, which has in the meantime already been registering further requests.

The process just described also takes place independently for access to Bus B, for which each subscriber terminal is provided with another Request Counter and Supervisory Counter.

With the use of different priorities, which has not so far been considered, the three REQUEST bits can be used to distinguish between three, or even seven, different priorities. For each of these priorities, a separate access control, as described in connection with Figure 5, is then required per bus. In this case, every request with a higher priority is also counted as a request for every lower priority. Furthermore, every higher-priority request that is fulfilled is also counted as fulfilling a request with lower priority. The counters with the same priority, arranged along one of these two busses and allocated to it, act as a distributed queue DQ, with the behaviour of a FIFO.

According to the invention, the same access procedure is used twice in parallel, namely once for internal traffic and once for external traffic. For this another set of the same counters is required. The slots leaving the network via the network node NK must be identified as such during their request. This can be achieved by using one of the two RESERVED bits in the Access Control Field, or one of the REQUES i bits, as an EXTERNAL bit. Then the internal and external traffic can be requested separately through the remaining REQUEST bits on the one hand, and the EXTERNAL bit on the other hand or the EXTERNAL bit can specify the type of request and the REQUEST bits their priority.

With the assumption that only one new slot is requested per slot, i.e. by all three REQUEST bits together, overloading on the Busses A and B is not possible. Therefore, it is sufficient for the basic procedure that the request only occurs on the bus which is opposite to the bus which is to be accessed. For the separate requests for slots for 0: external traffic, many more requests can occur than can be fulfilled. These requests must therefore be distributed over all subscribers, that is over both busses, so that further requests can be suppressed during overloading, or some other control can be exercised.

A distinction must be made between the two types of traffic, not only during the formation, but also in the processing, of the two request queues. For this also, one of the two RESERVED bits can be used. For example, the subscriber terminal TE4, at which Bus B starts (which only allows outgoing external traffic), can reserve individual slots for external traffic by setting a RESERVED bit, within the framework of the network node capacity and the current requests.

By means of the procedure described, each of the subscriber terminals TEl, TE4 is continously informed, implicitly or explicitly, concerning the current loading of the network interworking unit in the network node NK, even if only by the message "still free" or "no longer free". In order to actually prevent overloads, this information must naturally be processed in such a way that external traffic is only transmitted when both the TN and the network interworking un;t can accept it.

Another possibility for informing the individual terminals connected to the network, consisting here of the busses A and B, about the current loading of the network interworking unit, for example, lies in having a slot travel completely around both busses once per frame, into whose Segment Payload all connected terminals write their requests for the next frame, and from which all terminals can deduce the requests of the other terminals. If a terminal knows the requests of all the terminals, and if a clear distribution rule applies, then each terminal can determine when any terminal, in particular itself, can transmit.

9 In a synchronous travelling slot quite a lot of information about the current loading of this network can be distributed. In this way, not only can the overloading be avoided of an individual network interworking unit or the network itself, but also the overloading of additional network interworking units or individual subscriber terminals.

Finally, in order to avoid the overloading of an individual subscriber terminal which is connected to a network, all other terminals which could contribute to the overload, must be informed about the current loading. By evaluating this information with the aid of a suitable distribution rule, an overload state can then be avoided.

0 too 00 o0 o 0 e

Claims (4)

1. A method for controlling the access to an asynchronously operated first network which allows access to a further network by means of a network interworking unit, wherein all equipment via which the first network is accessed is informed about the current loading of the network interworking unit

2. A method as claimed in claim 1, wherein the access to the first network is controlled by two like methods for network access contrc' in such a way that always the access to the first network and, if necessary, also the interworking with a further network, must be guaranteed before the access.

3. A circuit arrangement for controlling access to an asynchronously operated first network which allows access to a further network by means of a network interworking unit, wherein the circuit arrangement has registers or holding information about the loading of the network interworking unit.

4. A method substantially as herein described with reference to Figures 1 5 of the accompanying drawings. DATED THIS SEVENTH DAY OF DECEMBER 1994 ALCATEL N.V. 4 S n- 64 o 4 4 6 ABSTRACT This invention relates to a method and c.ouit arrangement for controlling access to an asynchronously operated network. Known communication networks compriseequal-level, higher-level and lower-level sub-networks with network interworking units. The external traffic of a sub-network is limited, either physically or by protocol, to a value which is smaller than the internal traffic. In sub-networks near the subscribers, there is no smoothing of the random traffic. Network interworking units dr'? either designed for traffic peaks and generally severely underloaded, or designed for the mean value and then suffer losses during traffic peaks. According to the invention, all accesses to a sub-network (TN) are continuously informed about the current loading of a network inter-working unit which leads out of 6 this sub-network. The subscriber terminals of a DQDB system indicate requests for time slots separately for external traffic. FIGURE 1. ft 00. 0