GB2343093A - Changing from a uni-directional ring mode to a bi-directional non-ring mode in the event of an interruption - Google Patents

Abstract

An optical network in which user equipment is connected to bi-directional optical links by transponders having a switching function. Each transponder in succession is enabled to transmit by a token forwarded from one transponder to the next. In the case of a non-circulating configuration, the enabled transponder transmits signals onto the links in both directions; the other transponders upstream and downstream receive these signals for local processing, and also forward them on to further transponders beyond them. The links may be coupled into a ring, providing two separate propagation rings. In this case, the enabled transponder transmits signals onto one the rings to start propagating round it; the other transponders receive the signals for further processing and forward them on in the same direction; the second ring may operate the same protocol independently; in the event of interruption of the ring the system may switch to operation in the non-circular configuration.

Description

Media access control protocols for multiple node networks Technical Field This invention relates to media access control protocols for multiple node networks, in particular for networks in which signals are communicated in the optical domain.

Background Art It is anticipated that new optical telecommunications networks will be implemented in metropolitan areas, and one possible interface between the telecommunications network and the end customer will be"Gigabit Ethernet" (or a higher speed derivative). In this scenario, the end customer's equipment (for example a router) will have a Gigabit Ethernet interface card, which may for example connect to the telecommunications network via a "transponder" (an optical frequency converter for converting an optical signal between the end-user system wavelength and the wavelength of an assigned optical channel in a wavelength division multiplexing-WDM-system). In initial implementations it is likely that an optical channel will be set up across the optical telecommunications network to a second corresponding transponder and then on to a second end customer's equipment. It will therefore create a point to point connection, across the metropolitan area telecommunications network, between the two pieces of customer's equipment. However in many cases this may be too inflexible.

WDM technology and the use of optical cross-connects (OXC-see for example US patents 5,699,462 and 5,732,168) are now being implemented in working products.

As well as the use of an optical layer as the highest layer in Synchronous Optical Network (SONET)/Synchronous Digital Hierarchy (SDH) technology, there is also very considerable interest in Internet Protocol (IP) and Asynchronous Transfer Mode (ATM) networks being implemented in a way which minimises or eliminates SONET/SDH support within the protocol stack. Specifically, it seems that in the long-haul networks, because of the considerable number of electrical repeaters already in place which are specific to SONET or SDH, it will be required in the short term to keep SONET/SDH framing, but the SONET/SDH multiplexing and cross-connect functions can potentially be eliminated.

In the metropolitan/municipal area, there is no requirement for electrical repeaters, and so both IP/ATM within SONET/SDH framing, and independent IP or ATM over WDM are likely to arise. Of particular interest will be the use of Gigabit Ethernet technology to implement IP over the WDM optical layer channel.

A parallel development is the realisation that the tremendous current and projected future demand for bandwidth will comprise packetised data rather than voice communications. The demand for data is projected to considerably exceed that of voice within very few years. It therefore makes sense to optimise the telecommunications network for data traffic rather than voice. There is also a strong movement towards transporting voice across IP networks (IP telephony), and this leads to the conclusion that IP will become the universal convergence layer.

It is an object of this invention to provide methods and apparatus for media access control which facilitate implementation of large area networks using optical technologies.

Disclosure of Invention According to one aspect of this invention there is provided a method of implementing media access control for an optical network comprising nodes linked by bidirectional optical links, comprising the steps of : enabling each node successively to transmit optical signals onto the network; determining within each node whether that node is currently enabled to transmit and: -if the node is so enabled, selectively coupling signals from that node to the links connected thereto for propagation in both directions from that node; -if the node is not so enabled, selectively coupling received signals propagating in either direction on said links onwards in said direction, and selectively accepting signals propagating in a predetermined one of said directions for local reception and processing.

According to another aspect of this invention there is provided a method of implementing media access control for an optical network comprising nodes linked by optical links in an endless ring for propagation of optical signals in a predetermined direction, comprising the steps of : enabling each node successively to transmit optical signals onto the network ; determining within each node whether that node is currently enabled to transmit and: -if the node is so enabled, selectively coupling signals from that node to the links for propagation in said predetermined direction from that node; -if the node is not so enabled, selectively coupling received signals to said links for propagation onwards in said predetermined direction, and selectively accepting said received signals for local reception and processing.

If a system operating according to this second method suffers an interruption of the ring, it may switch to operation in accordance with the first method.

Thus a logical network is created connecting several pieces of end customer equipment each at different locations across, for example, a metropolitan area. The issue of sharing bandwidth across, and allowing orderly access to, such a multiple node network can be solved, in one implementation, by superimposing a MAC protocol on top of the XON/XOFF scheme (implemented as the PAUSE frame) which is already provided by Gigabit Ethernet interface cards. The superimposed MAC is essentially a token-based protocol, potentially implemented at the transponder level.

In either case circuitry associated with each node may be enabled to supply data when the associated node is enabled to transmit optical signals onto the network, and disabled when that node is not so enabled; this circuitry may be an interface supporting XON/XOFF switching, in which case the interface can be enabled and disabled in accordance with XON and XOFF signals generated in the associated node.

Brief Description of Drawings A media access control protocol in accordance with this invention will now be described, by way of example, with reference to the accompanying drawings, in which: Figure 1 shows an architecture for interconnecting sites via multiple networks; Figure 2 shows another architecture for interconnecting sites via multiple networks; Figure 3 shows a modified form of the architecture of Figure 2; Figure 4 illustrates operation of one form of the media access control protocol; Figure 5 illustrates operation of another form of the media access control protocol; and Figures 6a and 6b illustrate use of the invention in compensating for component failure.

Best Mode for Carrying Out the Invention, & Industrial Applicability One problem with current proposals for optical networking within metropolitan areas involves the issue of granularity. If an optical channel is defined as, for example, 1 Gbit/s, then the bandwidth between two endpoints can only be allocated in integral units of 1 Gbit/s, and it is impossible to allocate the bandwidth in any fraction of 1 Gbit/s. In other words, if there is an optical channel connection from, for example, a corporate Site A to an Internet Service Provider (ISP) B (i. e. an optical wavelength conveying information between A and B), then that channel from A to B is unavailable to any other traffic. Even if there is little or no actual traffic for an interval between A and B, it is not possible for other traffic to make use of the capacity of that optical channel. Essentially, there is no channel sharing and the system is restricted to one connection (between two end points) per channel. So the number of clients that can be supported is very closely related to the number of WDM channels per fibre and the number of fibres per cable. IP over SONET, where SONET framing but not SONET add-drop multiplexer (ADM) or Cross Connect technology is utilised also has the restriction of one connection per channel.

By contrast, a full SONET system effectively allows optical channel sharing, so it is not restricted to one connection per channel and offers much greater granularity of bandwidth allocation. In other words, the optical bandwidth can be shared amongst the ADM's along the route. However, a full SONET system is very expensive, and one of the purposes of optical networking is to reduce costs as compared to a full SONET implementation. Also a SONET ADM can only provide fixed bit-rate channels, and so can be very inefficient in terms of bandwidth use when carrying variable bit-rate traffic.

For example, referring to Figure 1, five sites A to E require connection to an ISP via one or more metropolitan and municipal networks. These networks consist of, for example, OXC's, wavelength multiplexers and demultiplexers, optical amplifiers and wavelength converters interconnected by optical fibres to allow optical channels to be transmitted from one side of the network to the other. In the simplest implementation an optical channel occupies the same wavelength throughout its entire path, but it is also possible for the channel to be converted to a different wavelength at one or more points; however, the channel is transmitted at a single well-defined wavelength on any particular fibre. The invention is equally usable whether or not the optical channel is defined as having the same wavelength throughout its path.

As explained above, if there is no possibility of optical channel sharing there have to be five point-to-point bi-directional channels each connecting a respective site to the ISP. Furthermore, in the case where protection against the possibility of a failure in the path of the main optical channel is required, each site has to have a second, diverselyrouted, bi-directional channel back to the ISP, as shown in Figure 1.

A preferable system would not have the restriction of one connection per optical channel, but would also be much lower in cost than a full SONET system, and in contrast to SONET would more efficiently carry variable bit-rate traffic. Systems with these characteristics, in which channel sharing in particular is implemented, are shown in Figures 2 and 3.

Referring to Figure 2, the data to and from the ISP are'collected'in a logical network comprising the individual optical links between each component physical network.

It can be seen that there is a saving in optical channel occupation, as compared to each of the sites having an individual bi-directional connection as shown in Figure 1. A still greater saving is obtained if protection is required. As shown in Figure 3, a single extra link is inserted from site E to the ISP and this gives all of the sites protection against a single failure.

Figure 4 shows the essential features of a practical implementation of the invention, using Gigabit Ethernet interface cards 10,12,14,16 and 18, and associated transponders 20,22,24,26 and 28 (described in more detail below).

Referring to Figure 4, each interface card 10-18 is linked by a bi directional optical channel 30 to its respective transponder 20-28, which is in turn connected by bi-directional channels to the transponder of its nearest neighbour in the logical network. These channels, indicated by arrows in Figure 4, may pass through OXC's, wavelength multiplexers and demultiplexers and other similar equipment, which are omitted from Figure 4 for the sake of clarity. The transponders 20-28 include switches, operating for example on signals converted into the electrical domain, enabling them to implement the protocol described below.

A'token', i. e. permission to transmit on the channel, is passed from the furthest upstream (i. e. right-most as viewed in Figure 4) transponder 28, from one transponder to the next downstream. The token is, for example, transmitted from one transponder to another by sending a predetermined pattern of bits.

The token can be held by a transponder for a length of time up to a predetermined maximum period MAXTOKENHOLD. Other, more elaborate algorithms are also possible for determining how long the token can be held. When the token reaches the furthest downstream transponder 20, it is sent straight back on the reverse channel to the transponder 28 furthest upstream, being immediately repeated through the intermediate transponders 22-26 along the way. Therefore each transponder knows whether the token is upstream (to the right) of it, downstream (to the left) of it, or currently resident at that transponder itself.

In Figure 4 it is assumed that the token is at the transponder 24, the middle transponder. Thick black lines indicate data being sent from the associated originating interface card 14 through the transponders to all of the other interface cards. Dotted black lines show the internal connections within the transponders.

The essential rules to determine the state of these internal connections : 1) If the token has already been forwarded and is now downstream of a transponder, then that transponder's internal connection is set to take the upstream directed channel and repeat it both to the associated interface card and to the next transponder upstream. Thus in Figure 4, the transponder 26 and all transponders upstream of it (i. e. to its right) have their internal connections set as illustrated by the dashed lines in the transponder 26.

2) If the token has not been forwarded and is still upstream of a transponder, then its internal connection is set to take the downstream directed channel and repeat it both to the associated interface card and to the next transponder downstream. In Figure 4, the transponder 22 and all transponders downstream of it (i. e. to its left) have their internal connections set as illustrated by the dashed lines in the transponder 20.

3) If a transponder has possession of the token, then data received from the associated interface card are connected to both the downstream and the upstream outputs. In Figure 4 the transponder 24 has the token, and its internal connections are shown by the dashed lines within it.

Assuming the interface card to transponder connection is full duplex Gigabit Ethernet, the data flow between the transponder and the interface card is organised using the Full Duplex flow control, defined in IEEE 802.3x. Since Gigabit connections have the most potential to flood switched and router networks, it is envisaged that Gigabit Ethernet interface cards will be implemented to support 802.3x flow control, in which the XON and XOFF signals are in practice implemented by a'Pause'flow control frame.) If a transponder receives the token, it sends XON to the associated interface card.

If no frame emerges from the interface card in response, XOFF is sent to the interface card and the token is forwarded on. If frames do emerge, then the internal connection within the transponder ensures frames are repeated to the neighbouring transponders. When the MAXTOKENHOLD period expires, indicating that the token should be forwarded, or when frames from the interface card cease, then XOFF is sent to the interface card and the token is forwarded to the next transponder.

If the interface uses half-duplex Gigabit Ethernet, then an alternative mechanism to the use of XOFF is required, such as making the interface card sense carrier or collisions.

As noted above, the transponders 20-28 are used to translate client signals (e. g. from the interface cards 20-28) from loosely defined and unstable wavelengths to well defined and stable wavelengths as they are transferred into a WDM system. They are also enabled to perform the following functions: -decode incoming symbols (e. g. nB/mB) down to the n-bit data level, and re-code them when repeating to another transponder; -generate n-bit symbols which constitute the Pause flow control (i. e. XON/XOFF) frames; -sense from the decoded symbol stream when frames are being passed through; -generate and detect the bit patterns which constitute tokens.

It is also possible for the transponders 20-28 to support more than one interface card if desired. The transponders are not required to store frames or packets, i. e. they are symbol-level repeater devices, rather than store-and-forward devices.

Transponders also have some other advantages: -they can act as a demarcation device between client and carrier, with'keep alive'signal and test capabilities ; incorporation of BER measurement capabilities etc. is also possible; -they transmit continuously, and so optical amplifier disruption, caused by varying total power levels due to client equipment turning on and off, is avoided; -they ensure a consistent power level enters the optical network and so increase the margins/reach of the WDM network; also, a remote client becomes allowable; -transponders could also be convenient places from an architectural viewpoint to locate various other functions, besides BER measurement and verification, such as: -channel encryption; -Forward Error Correction (FEC); -optical power measurement; -line code conversion (e. g. between different block codes); -frame format conversion (e. g. from Gigabit Ethernet to some other format); -channel protection (i. e. duplication of optical paths so that a spare path is available in case the working path fails); -network management (e. g. using Simple Network Management Protocol (SNMP) or telecom management network (TMN)).

In the embodiments described above a dual (bi-directional) bus has been examined.

If there is a connection, as shown in Figure 3, between Site E and the ISP, then a ring can be formed, as shown in Figure 5. More particularly there would be two oppositelydirected and fairly independent rings, Ring 1 and Ring 2. Some sites would connect to just Ring 1 (shown in solid line in Figure 5), some to just Ring 2 (shown in dashed line) and some (e. g. the ISP) would attach to both. The sites which attach to both rings would do so with two separate interface cards. The two rings operate independently, so the operation of one only is examined below and illustrated in Figure 5.

Referring to Figure 5, the operating protocol for one ring is as follows: 1) If a transponder has possession of the token, then the signal from the associated interface card is connected to the (single) ring output which is connected to the next transponder around the ring.

2) If the transponder does not have possession of the token, then incoming data being received on the ring is both passed to the associated interface card and repeated to the next transponder around the ring.

In one possible implementation the dual ring configuration described above with reference to Figure 5 is used under normal conditions, as shown in Figure 6a, but if there is a fibre break or transponder failure which breaks the ring's continuity, then the transponders revert to the dual bus configuration of Figure 4 for the duration of the fault, as shown in Figure 6b.

Claims (7)

1. A method of implementing media access control for an optical network comprising nodes linked by bi-directional optical links, comprising the steps of : enabling each node successively to transmit optical signals onto the network; determining within each node whether that node is currently enabled to transmit and: -if the node is so enabled, selectively coupling signals from that node to the links connected thereto for propagation in both directions from that node; -if the node is not so enabled, selectively coupling received signals propagating in either direction on said links onwards in said direction, and selectively accepting signals propagating in a predetermined one of said directions for local reception and processing.

2. A method of implementing media access control for an optical network comprising nodes linked by optical links in an endless ring for propagation of optical signals in a predetermined direction, comprising the steps of: enabling each node successively to transmit optical signals onto the network; determining within each node whether that node is currently enabled to transmit and: -if the node is so enabled, selectively coupling signals from that node to the links for propagation in said predetermined direction from that node; -if the node is not so enabled, selectively coupling received signals to said links for propagation onwards in said predetermined direction, and selectively accepting said received signals for local reception and processing.

3. The method of claim 2, wherein said optical links enable propagation of optical signals in a second direction opposite to said predetermined direction, and signals propagating in said second direction are controlled in a manner analogous to but independent of signals propagating in said predetermined direction.

4. The method of claim 2 or claim 3, wherein in the event of interruption of said endless ring said nodes change to operating in accordance with claim 1.

5. The method of any one of the preceding claims, wherein the signals are selectively coupled within transponders connected to said links.

6. The method of any one of the preceding claims, wherein circuitry associated with each said node is enabled to supply data when the associated node is enabled to transmit optical signals onto the network, and is disabled when that node is not so enabled.

7. The method of claim 6, wherein said circuitry is an interface supporting XON/XOFF switching, and the interface is enabled and disabled in accordance with XON and XOFF signals generated in the associated node.