GB2390927A - Multi-processor interconnection and management - Google Patents

Abstract

A multi-processor equipment enclosure 100 includes an interconnection backplane 190 for a plurality of processing modules 150, 160 and at least one management module 200. The backplane interconnection includes generic portions standardised over a range of processing modules and other portions specific to different processing modules. The management module senses the specific type of a processing module via protocols implemented in the generic portion and thereby controls the routing of communication and management signals via the backplane for each specific type of module. The type sensing protocols between processors 150, 160 may be implemented via geographic address lines in the generic portions of a compact PCI backplane. Also included is an independent claim directed towards a method of configuring multi-processor equipment to operate with a range of processing modules.

Description

ENCLOSURE FOR MULTI-PROCESSOR EQUIPMENT

INTRODUCTION

5 The invention relates to enclosures for multi-processor equipment, including in particular multi-channel instrumentation such as monitoring equipment for telecommunication networks.

10 In telecommunication networks, network element connectivity can be achieved using optical fibre bearers to carry data and voice traffic.

Data traffic on public telecommunication networks is expected to exceed voice traffic with Internet Protocol (IP) emerging as one data networking standard, in conjunction 15 with Asynchronous Transfer Mode (ATM) systems. Voice over IP is also becoming an important application for many Internet service providers with IP switches connecting IP networks to the public telephony network (PSTN). IP can be carried over a Sonet transport layer, either with or without ATM. In order to inter-operate with the PSTN, IP switches are also capable of inter-working with SS7, the common 20 signalling system for telecommunications networks, as defined by the International Telecommunications Union (ITU) standard for the exchange of signalling messages over a common signalling network.

Different protocols are used to set up calls according to network type and supported 25 services. The signalling traffic carries messages to set up calls between the necessary network nodes. In response to the SS7 messages, an appropriate link through the transport network is established, to carry the actual data and voice traffic (the payload data) for the duration of each call,. Traditional SS7 links are time division multiplexed, so that the same physical bearer may be carrying the signalling and the 30 payload data. The SS7 network is effectively an example of an "out of band" signalling network, because the signalling is readily separated from the payload. For ATM and IP networks, however, the signalling and payload data are statistically

multiplexed on the same bearer. In the case of statistical multiplexing the receiver has to examine each message/cell to decide if it is carrying signalling or payload data.

One protocol similar to SS7 used in such IP networks is known as Gateway Control Protocol (GCP).

The monitoring of networks and their traffic is a fundamental requirement of any system. The "health" of the network must be monitored, to predict, detect and even anticipate failures, overloads and so forth. Monitoring is also crucial to billing of usage charges, both to end users and between service providers. The reliability 10 (percentage availability) of monitonag equipment is a prime concern for service providers and users, and many applications such as billing require "high availability" monitoring systems, such that outages, due to breakdown or maintenance, must be made extremely rare.

15 A widely-used monitoring system for SS7 signalling networks is acceS57 _ from Hewlett-Packard. An instrument extracts all the SS7 packetised signals at Signalling Transfer Points (STPs), which are packet switches analogous to IP routers, that route messages between end points in SS7 networks. The need can be seen for similar monitoring systems able to cope with combined IP/PSTN networks, especially at 20 gateways where the two protocols meet. A problem arises, however, in the quantity of data that needs to be processed for the monitoring of IP traffic. In Intemet Protocol networks, there is no out of band signalling network separate from the data traffic itself. Rather, routing information is embedded in the packet headers of the data transport network itself, and the full data stream has to be processed by the 25 monitoring equipment to extract the necessary information as to network health, billing etc. Moreover, IP communication is not based on allocating each "call" with a link of fixed bandwidth for the duration of the call: rather bandwidth is allocated by packets on demand, in a link shared with any number of other data streams.

Accordingly, there is a need for a new kind of monitoring equipment capable of grabbing the vast volume of data flowing in the IP network bearers, and of processing it fast enough to extract and analyse the routing and other information crucial to the

J monitoring function. The requirements of extreme reliability mentioned above apply equally in the new environment.

Networks such as these may be monitored using instruments (generally referred to as 5 probes) by making a passive optical connection to the optical fibre bearer using an optical splitter. However, this approach cannot be considered without due attention to the optical power budget of the bearer, as the optical splitters are lossy devices. In addition to this, it may be desirable to monitor the same bearer marry times or to monitor the same bearer twice as part of a backup strategy for redundancy purposes.

10 With available instrumentation, this implies a multiplication of the losses, and also disruption to the bearers as each new splitter is installed. Issues of upgrading the transmitter and/or receiver arise as losses mount up.

The inventors have analysed acceSS7 network monitoring systems (unpublished at the 15 present filing date). This shows that the reasons for lack of availability of the system can be broken down into three broad categories: unplanned outages, such as software defects; planned outages, such as software and hardware upgrades; and hardware failures. Further analysis shows that the majority of operational hours lost are caused by planned and unplanned maintenance, while hardware failures have a relatively 20 minor effect. Accordingly, increasing the redundancy of disk drives? power supplies and the like, although psychologically comforting, can do relatively little to improve system availability. The greatest scope for reducing operational hours lost and hence increasing availability is in the category of planned outages.

25 In order to implement a reliable monitoring system it would therefore be advantageous to have an architecture with redundancy allowing for spare probe units that is tolerant of both probe failure and probe reconfiguration, and provides software redundancy. 30 Monitoring equipment designed for this purpose does not currently exist. Service providers may therefore use stand-alone protocol analysers which are tools really intended for the network commissioning stage. These usually terminate the fibre bearer, in place of the product being installed, or they plug into a specific test port on

the product under test. Specific test software is then needed for each product.

Manufacturers have alternatively built diagnostic capability into the network equipment itself, but each perceives the problems differently, leading to a lack of unifonnity, and actual monitoring problems, as opposed to perceived problems, may 5 not be addressed.

Further considerations include the physical environment needed to house such processing architecture. Such a hardware platform should be as flexible as possible to allow for changes in telecommunications technology and utilise standard building 10 blocks to ensure cross platform compatibility. For example, there exist standards in the USA, as set out by the American National Standards Institute (ANSI) and Bellcore, which differ from those of Europe as set by the European Telecommunications Standards Institute (ETSI). Versions of SS7 may also vary from country to country, owing to the flexibility of the standard, although the ITU standard 15 is generally used at international gateways. The USA Bellcore Network Equipment Building System (NEBS) is of particular relevance to rack-mounted telecommunications equipment as it provides design standards for engineering construction and should be taken into account when designing network monitoring equipment. Such standards impose limitations such as connectivity and physical 20 dimensions upon equipment and, consequently, on cooling requirements and aisle spacing of network rack equipment.

It is known that standard processing modules conforming for example to the cPCI standard are suitable for use in telecommunication applications. The further standard 25 H.110 provides a bus for multiplexing baseband telephony signals in the same backplane as the cPC1 bus. Even with Intel Pentium_ or similar processors, however, such arrangements do not currently accommodate the computing power needed for the capture and analysis of broadband packet data. Examples of protocols and their data rates to be accommodated in the monitored bearers in the future 30 equipment are for example DS3 (44 Mbit.s-'), OC3 (155 Mbit.s-'), OC12 (622 Mbit.s-') and OC48 (2.4 Gbit.s-'). Aside from the volume of data to be handled, conventional chassis for housing such modules do not also support probe architectures of the type currently desired, both in terms of processing capability and also to the

extent that their dimensions do not suit the layout of telecommunication equipment rooms such as may be designed to NEBS allowing them to coreside with network equipment. 5 For example, the typical general purpose chassis provides a rack-mounted enclosure in which a backplane supports and interconnects a number of cPCI cards, including a processor card and peripheral cards, to form a functional system. The cards are generally oriented vertically, with power supply (PSU) modules located above or below. Fans force air through the enclosure frown bottom to top for cooling the 10 modules. A peripheral card may have input and output (I/O) connections on its front panel. Alternatively, I/O connections may be arranged at the rear of the enclosure, using a special "transition card". Examples of rack widths in common use are l9 inch (483mm) and 23 inch (584mm). The siting of racks in telecommunications equipment rooms implies an enclosure depth should be little over 12 inches (305mm).

15 However, cPCI and VME standard processor cards and compatible peripheral cards are already 205mm deep (including mountings) and the conventional interface card mounted behind the back plane adds another 130 mm. Moreover, although parts of the connector pin-outs for cPCI products are standardized, different vendors use other connectors differently for management bus signals and for LAN connections. These 20 variations must also be adapted to by a dedicated interconnect, and designs will often assume that cards from a single vendor only are used.

It is noted at this point that the cPCI standard defines a number of physical connectors to be present on the backplane, but only two of these (J1, J2) are specified as to their 25 pin functions. Although many general-purpose processor cards are based for example on Pentium_ microprocessors, different card vendors use the remaining connectors differently for communication and management signals such as SMB and LAN connections. - - -

Accordin, to one aspect of the invention there is provided a multiprocessor equipment enclosure having a housing and a backplane interconnection for a plurality of processing modules and a management module, the backplane interconnection including generic portions standardised over a range of processing modules and other 5 portions specific to different processing modules within said range, wherein said management module is arranged to sense the specific type of a processing module by using protocols implemented by the modules via connections in the generic portion of the interconnection, and to control routing of communication and management signals via the backplane in accordance with the sensed specific type of each processing 1 0 module.

The type sensing protocols may be implemented via geographic address lines in the standardised portions of a compact PCI backplane.

15 According to another aspect of the invention there is provided a method of configuring a multi-processor equipment to operate with different ones of a range of processing modules, said equipment having a backplane interconnection for a plurality of processing modules and a management module, and the backplane interconnection including generic portions standardised over the range of processing 20 modules and other portions specific to different processing modules within said range, comprising the steps of: causing said management module to sense the specific type of a processing module by using protocols implemented by the modules via connections in the generic portion of the interconnection; and 25 controlling routing of communication and management signals via the backplane in accordance with the sensed specific type of each processing module.

A_ \

( 7 BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which: 5 Figure 1 shows a model of a typical ATM network.

Figure 2 shows a data collection and packet processing apparatus connected to a physical telecommunications network via a LAN/WAN interconnect.

10 Figure 3 shows the basic functional architecture of a novel network probe apparatus, as featured in Figure 2.

Figure 4 shows a simple network monitoring system which can be implemented using the apparatus of the type shown in Figure 3.

Figure 5 shows another application of the apparatus of Figure 4 giving 3+ 1 redundancy. Figure 6 shows a larger redundant network monitoring system including a backup 20 apparatus.

Figure 7 shows an example of a modified probe apparatus permitting a "daisy chain" configuration to provide extra redundancy and/or processing power.

25 Figure 8 shows an example of daisy chaining the probe chassis of Figure 7 giving 8+1 redundancy.

Figure 9 shows a further application of the probe apparatus giving added processing power per bearer.

Figure 10 shows a second means of increasing processing power by linking more than one chassis together.

Figure 11 shows a signal replicating device (referred to as a Broadband Bridging Isolator (BBI)) for use in a network monitoring system.

5 Figure 12 shows a typical configuration of a network monitoring system using the BBI of Figure 11 and several probe apparatuses.

Figures 13A and 13B illustrate a process of upgrading the processing power of a network monitoring system without interrupting operation.

Figure 14 is a functional schematic diagram of a generalised network probe apparatus showing the functional relationships between the major modules of the apparatus.

Figure ISA shows the general physical layout of modules in a specific network probe 15 apparatus implemented in a novel chassis and backplane.

Figure 15B is a front view of the chassis and backplane of Figure 15A with all modules removed, showing the general layout of connectors and interconnections in 20 the backplane.

Figure 15C is a rear view of the chassis and backplane of Figure 15A with all modules removed, showing the general layout of connectors and interconnections in the backplane, and showing in cut-away form the location of a power supply module.

Figure 16 shows in block schematic form the interconnections between modules in the apparatus of Figures 1 5A-C.

Figure 17 is a block diagram showing in more detail a cross-point switch module in 30 the apparatus of Figure 16, and its interconnections with other modules.

Figure 18 is a block diagram showing in more detail a packet processor module in the apparatus of Figure 16, and its interconnections with other modules.

Figure 19 is a block diagram showing in more detail a combined LAN and chassis management card in the apparatus of Figure 16.

s DETAILED DESCRIPTION OF THE EMBODIMENTS

Background

Figure I shows a model of a telecommunication network 10 based on asynchronous transfer mode (ATM) bearers. Possible monitoring points on various bearers in the 10 network are shown at 20 and elsewhere. Each bearer is generally an optical fibre carrying packetised data with both routing information and data "payload" travelling in the same data stream. Here "bearer" is used to mean the physical media that carries the data and is distinct from a "link", which in this context is defined to mean a logical stream of data. Many links of data may be multiplexed onto a single bearer.

15 These definitions are provided for consistency in the present description, however,

and should not be taken to imply any limitation of the applicability of the techniques

disclosed, or on the scope of the invention defined in the appended claims. Those skilled in the art will sometimes use the term "channel" to refer to a link (as defined), or "channel" may be used to refer to one of a number of virtual channels being carried 20 over one linlc, which comprises the logical connection between two subscribers, or between a subscriber and a service provider. Note that such "channels" within the larger telecommunications network should not be confused with the monitoring channels within the network probe apparatus of the embodiments to be described hereinafter. The payload may comprise voice traffic and/or other data. Different protocols may be catered for, with examples showing connections to Frame Relay Gateway, ATM and DSLAM equipment being illustrated. User-Network traffic 22 and Network-Network traffic 24 are shown here as dashed lines and solid lines respectively.

In Figure 2 various elements 25-60 of a data collection and packet processing system distributed at different sites are provided for monitoring bearers Ll-L8 etc. of a

telecommunications network. The bearers in the examples herein operate in pairs L1, L2 etc. for bi-directional traffic, but this is not universal, nor is it essential to the invention. Each pair is conveniently monitored by a separate probe unit 25, by means of optical splitters Sl, S2 etc. inserted in the physical bearers. For example, one 5 probe unit 25, which monitors bearers Ll and L2, is connected to a local area network (LAN) 60, along with other units at the same site. The probe unit 25 on an ATM/IP network must examine a vast quantity of data, and can be programmed to filter the data by a Virtual Channel (VC) as a means of reducing the onboard processing load.

Filtering by IP address can be used to the same effect in the case of IP over SDH and 10 other such optical networks. Similar techniques can be used for other protocols. Site processors 40 collate and aggregate the large quantity of information gathered by the probe units, and pass the results via a Wide Area Network (WAN) 30 to a central site 65. Here this information may be used for network planning and operations. It may alternatively be used for billing according to the volume of monitored traffic per 15 subscriber or service provider or other applications.

The term "probe unit" is used herein refer to a functionally selfcontained sub-system designed to carry out the required analysis for a bearer, or for a pair or larger group of bearers. Each probe unit may include separate modules to carry out such operations 20 as filtering the packets of interest and then interpreting the actual packet or other data analysis. In accordance with current trends, it is assumed in this description that the links to be

monitored carry Internet Protocol (IP) traffic over passive optical networks (PONs) 25 comprising optical fibre bearers. Connection to such a network can only really be achieved through use of passive optical splitters Sl, S2 etc. Passive splitters have advantages such as high reliability, comparatively small dimensions, various connection configurations and the fact that no power or element management resources are required. An optical splitter in such a situation works by paring off a 30 percentage of the optical power in a bearer to a test port, the percentage being variable according to hardware specifications.

1/ ' 11

A number of issues are raised when insertion of such a device is considered. For example, there should be sufficient bearer receiver power margins remaining at both test device and the through port to the rest of the network. it becomes necessary to consider what is the most economic method of monitoring the bearer in the presence 5 of a reduced test port power budget while limiting the optical power needed by the monitoring probe and if the network would have to be re-configured as a result of inserting the device.

Consequently, inserting a power splitter to monitor a network frequently requires an 10 increase in launch power. This entails upgrading the transmit laser assembly and installing an optical attenuator where needed to reduce optical power into the through path to normal levels. Such an upgrade would ideally only be performed once.

For these reasons it is not desirable to probe, for example, an ATM network more 15 than once on any given bearer. Nevertheless, it would be desirable to have the ability for multiple probing devices to be connected to the same bearer, that is, have multiple outputs from the optical interface. The different probes may be monitoring different parameters. In addition, however, any network monitoring system must offer a high degree of availability, and multiple probes are desirable in the interests of redundancy.

20 The probe apparatuses and ancillary equipment described below allow the implementation of such a network monitoring system which can be maintained and expanded with simple procedures, with minimal disruption to the network itself and to the monitoring applications.

25 Network Probe System - General Architecture Figure 3 shows the basic functional architecture of a multi-channel optical fibre telecommunications probe apparatus 50 combining several individual probe units, into a more flexible system than has hitherto been available. The network monitoring apparatus shown receives N bearer signals 70 such as may be available from N optical 30 fibre splitters. These enter a crosspoint switch 80 capable of routing each signal to any of M individual and independently-replaceable probe units 90. Each probe unit corresponds in functionality broadly with the unit 25 shown in Figure 2. An additional

external output 85 from the cross-point switch 80 is routed to an external connector.

This brings important benefits, as will be describe] below.

The cross-point switch 80 and interconnections shown in Figure 3 may be 5 implemented using different technologies, for example using passive optical or optoelectronic cross-points. High speed networks, for example OC48, require electrical path lengths as short as possible. An optical switch would therefore be desirable, deferring as much as possible the conversion to electrical. However, the optical switching technology is not yet fully mature. Therefore the present proposal is 10 to have an electrical implementation for the cross-point switch 80, the signals converted from optical to electrical at the point of entry into the probe apparatus 50.

The scale of an optoelectronic installation will be limited by the complexities of the cross-point switch and size of the probe unit. The choice of interconnect technology (for example between electrical and optical), is generally dependent on signal 15 bandwidth-distance product. For example, in the case of high bandwidth/speed standards such as OC3, OC12, or OC48, inter-rack connections may be best implemented using optical technology.

Detailed implementation of the probe apparatus in a specific embodiment will be 20 described in more detail with reference to Figures 14 to 19. As part of this, a novel chassis arrangement for multi-channel processing products is described, with reference to Figures 15A-15C, which may find application in fields beyond

telecommunications monitoring. First, however, applications of the multichannel probe architecture will be described, with reference to Figures 4 tol3.

Figure 4 shows a simple monitoring application which can be implemented using the apparatus of the type shown in Figure 3. The cross-point switch 80 is integrated into the probe chassis 100 together with up to four independently operating probe units. In this implementation of the architecture each probe unit (90 in Figure 3) is formed by a 30 packet processor module 150 and single board computer SBC 160 as previously described. There are provided two packet processors 150 in each probe unit 90. Each packet processor can receive and process the signal of one half-duplex bearer. SBC 160 in each probe unit has the capacity to analyse and report the data collected by the

! 13

two packet processors. Other modules included in the chassis provide LAN interconnections for onward reporting of results, probe management, power supply, and cooling modules (not shown in Figures 4 to 1 3B).

5 In this application example a single, fully loaded chassis 100 is used with no redundancy to monitor eight single (four duplex) bearers connected at 140 to the external optical inputs of the apparatus (inputs 70-1 to 70-N in Figure 3). The cross-

point switch external outputs 85 are shown but not used in this configuration.

Applications of these outputs are explained for example in the description of Figures

10 6, 8, 10 and 12 below.

Figure 5 shows an alternative application of the apparatus giving 3+1 redundancy.

Here three duplex bearer signals are applied to external inputs 140 of the probe apparatus chassis 100, while the fourth pair of inputs 142 is unused. Within the 15 chassis there are thus three primary probe units 90 plus a fourth, spare probe unit 120.

The cross-point switch 80 can be used to switch any of the other bearers to this spare probe unit in the event of a failure in another probe unit. Already by integrating the cross-point switch and several probe units in a single chassis, a scaleable packet processor redundancy down to 1:1 is achieved without the overhead associated with 20 an external cross-point switch. Since electrical failures cause only a very minor proportion of outages, redundancy within the chassis is valuable, with the added bonus that complex wiring outside the chassis may be avoided. One or more processors within each chassis can be spare at a given time, and switched instantaneously andJor remotely if one of the other probe units becomes inoperative.

Figure 6 shows a larger redundant system comprising four primary probe apparatuses (chassis 100-1 to 100-4), and a backup chassis 130 which operates in the event of a failure in one of the primary chassis. In this example of a large redundant system there are 16 duplex bearers being monitored. Each external input pair of the backup chassis 30 is connected to receive a duplex bearer signal from the external optical output 85 of a respective one of the primary apparatuses l 00-1 to 100-4. By this arrangement, in the event of a single probe unit failure in one of the primary apparatuses, a spare probe unit within the backup chassis can take over the out-of-service unit's function.

i' 14 Assuming all inputs and all of the primary probe units are operational in nonnal circumstances, we may say that 4:1 redundancy is provided.

Recognising that in this embodiment only one optical interface is connected to the 5 bearer under test, the chassis containing the optical interfaces can if desired have redundant communications and/or power supply units (PSUs) and adopt a "hot swap" strategy to permit rapid replacement of any hardware failures. "Hot swap" in this context means the facility to unplug one module of a probe unit within the apparatus and replace it with another without interrupting the operation or functionality of the 10 other probe units. Higher levels of protection can be provided on top of this, if desired, as described below with reference to Figures 15A-C and 16.

Figure 7 shows a modified probe apparatus which provides an additional optical input 170 to the cross-point switch 80. In other words, the crosspoint switch 80 has inputs I 5 for more bearer signals than can be monitored by the probe units within its chassis.

At the same time, with the external optical outputs 85, the cross-point switch 80 has outputs for more signals than can be monitored by the probe units within the chassis. These additional inputs 170 and outputs 85 can be used to connect a number

of probe chassis together in a "daisy chain", to provide extra redundancy and/or processing 20 power. By default, in the present embodiment, copies of the bearer signals received at daisy chain inputs 170 are routed to the external outputs 85. Any other routing can be commanded, however, either from within the apparatus or from outside via the LAN (not shown).

25 Figure 8 shows an example of daisy chaining the probe chassis to give 8:1 redundancy. The four primary probe chassis 100-1 to 100-4 are connected in pairs (100-1 & 100-2 and 100-3 & 100-4). The external outputs 85 on the first chassis of each pair are connected to the daisy chain inputs 170 on the second chassis. The external outputs 85 of the second chassis are cormected to inputs of the backup chassis 30 130 as before. These connections can carry the signal through to the spare chassis 130 when there has been a failure in a probe unit in either the first or second chassis in each pair. Unlike the arrangement of Figure 6, however, it will be seen that the backup chassis 130 still has two spare pairs of external inputs. Accordingly, the

f 15 system could be extended to accommodate a further four chassis (up to sixteen further probe units, and up to thirty-two further bearer signals), with the single backup chassis 130 providing some redundancy for all of them.

5 For applications that involve processor intensive tasks it may be desirable to increase the processing power available to monitor each bearer. This may be achieved by various different configurations, and the degree of redundancy can be varied at the same time to suit each application.

10 Figure 9 illustrates how it is possible to increase the processing power available for any given bearer by reconfiguring the probe units. In this configuration only two duplex bearer signals 140 are connected to the chassis 100. Two inputs 142 are unused. Within the cross-point switch 80 each bearer signal is duplicated and routed to two probe units 90. This doubles the processing power available for each of the 15 bearers 140. This may be for different applications (for example routine billing and fraud detection), or for more complex analysis on the same application. Each packet processor (150, Figure 4) and SBC (160) will be programmed according to the application desired. In particular, each packet processor, while receiving and processing all the data carried by an associated bearer, will be programmed to filter 20 the data and to pass on only those packets, cells, or header information which are needed by the SBC for a particular monitoring task. The ability to provide redundancy via the external outputs 85 still remains.

Figure 10 illustrates a second method of increasing processing power is by connecting 25 more than one chassis together in a daisy chain or similar arrangement. Concerning the "daisy chain" inputs 170, Figure 10 also illustrates how a similar effect can be achieved using the unmodified apparatus (Figure 3), providing the apparatus is not monitoring its full complement of bearer signals. The external inputs 140 can thus be connected to the external outputs 85 of the previous chassis, instead of special inputs 30 1 70.

In the configuration of Figure 10, two chassis 100-1 and 100-2 are fully loaded with four probe units 90 each. External signals for all eight probe units are received at 140

from a single duplex bearer. The cross-point switch 80 is used to replicate these signals to every probe unit 90 within the chassis 100-1, and also to the external outputs 85 of the first chassis 100-1. These outputs in turn are connected to one pair of inputs 144 of the second chassis 100-?. Within the second chassis, the same signals 5 are replicated again and applied to all four probe units, and (optionally) to the external outputs 85 ofthe second chassis 100-2.

Thus, all eight probe units are able to apply their processing power to the same pair of signals, without tapping into the bearer more than once. By adding further chassis in 10 such a daisy chain, processing power is scaleable practically to as much processing power as needed.

The examples given are for way of illustration only, showing how using the chassis architecture described it is possible to provide the user with the processing power 15 needed and the redundancy to maintain operation of the system in the event of faults and planned outages. It will be appreciated that there are numerous different configurations possible, besides those described.

For example, it is also possible to envisage a bi-directional daisy chain arrangement.

20 Here, one output 85 of a first chassis might be connected to one input l 70 of a second chassis, while the other output 85 is connected to an input 170 of a third chassis. This arrangement can be repeated if desired to form a bi-directional ring of apparatuses, forming a kind of "optical bus".

25 The probe apparatus described above allow the system designer to achieve N+l redundancy by using the cross-point switch 80 to internally re-route a bearer to a spare processor, or to another chassis. On the other hand, it will be recognised that some types of failure (e.g. in the chassis power supply) will disrupt operation of all of the processors in the chassis. It is possible to reduce such a risk by providing N+l 30 PSU redundancy, as will be/has been described.

( Broadband Bridging Isolator Figure 11 shows an optional signal replicating device for use in conjunction with the probe apparatus described above, or other monitoring apparatus. This device will be 5 referred to as a Broadband Bridging Isolator (BBI). Broadband Bridging Isolator can be scaled to di fferent capacities, and to provide additional fault tolerance independently of the probe apparatuses described above. The basic unit comprises a signal replicator 175. For each unit, an (optical) input 176 is converted at 177 to an electrical signal, which is then replicated and converted at 179 etc. to produce a 10 number of identical optical output signals at outputs 178-1 etc Also provided within BBI 172 are one or more standby selectors (multiplexers) 180 (one only shown). Each selector 180 receives replicas of the input signals and can select from these a desired one to be replicated at a selector optical output 182. An 15 additional input 186 (shown in broken lines) may be provided which passes to the selector 180 without being replicated, to permit "daisy chain" connection.

In use, BBI 172 takes a single tap input 176 from a bearer being monitored and distributes this to multiple monitoring devices, for example probe apparatuses of the 20 type shown in Figures 3 to 10. For reliability, the standby selector 180 allows any of the input signals to be switched to a standby chassis.

The number of outputs that are duplicated from each input is not critical. A typical implementation may provide four, eight or sixteen replicators 175 in a relatively small 25 rack mountable chassis, each having (for example) four outputs per input. Although the concepts here are described in terms of optical bearers, the same concepts could be applied to high speed electrical bearers (e.g. E3, DS3 and STMle).

The reasons for distributing the signal could be for multiple applications, duplication 30 for reliability, load sharing or a combination of all three. It is important that only one tap need be made in the operational bearer. As described in the introductory part of this specification, each optical tap reduces the strength of the optical signal reaching

the receiver. In marginal conditions, adding a tap may require boosting the signal on

the operational bearer. Network operators do not want to disrupt their operational networks unless they have to. The BBI allows different monitoring apparatuses for different applications to be connected, and removed and re-confgured without affecting the operational bearer, hence the name "isolator". The BBI can even be 5 used to re-generate this signal by feeding one of the outputs back into the network, so that the BBI becomes part of the operational network.

The number of bearer signals that are switched through the standby selector 180 will depend on the user's requirements - this number corresponds effectively to "N" in the 10 phrase "N+l redundancy". The number of standby selectors in each BBI is not critical. Adding more means that more bearers can be switched should there be a failure. The BBI must have high reliability as when operational in a monitoring environment 15 it an essential component in the monitoring of data, providing the only bridging link between the signal bearers and the probe chassis. No digital processing of the bearer signal is performed in the BBI, which can thus be made entirely of the simplest and most reliable optoelectronic components. When technology permits, in terms of cost and reliability, there may be an "all-optical" solution, which avoids conversion to 20 electrical form and back to optical. Presently, however, the state of the art favours the optoelectronic solution detailed here. The BBI can be powered from a redundant power supply to ensure continuous operation. The number of bearers handled on a single card can be kept small so that in event of a failure the number of bearers impacted is small. The control of the standby switch can be by an external control 25 processor.

Figure 12 shows a system configuration using BBIs and two separate probe chassis lOO-1 and 100-2 implementing separate monitoring applications. The two application chassis may be operated by different departments within the network operator's 30 organization. A third, spare probe chassis 130 is shared in a standby mode. This example uses two BBIs 172 to monitor a duplex bearer pair shown at Ll, L2, and other bearers not shown. Splitters Sl and S2 respectively provide tap input signals from Ll, L2 to the inputs 176 of the separate BBIs. Each BBI duplicates the signal at

its input 176 to two outputs 178, in the manner described above with reference to Figure 11. For improved fault tolerance, the two four-way BBIs 179 are used to monitor half duplex bearers Ll and L: separately. In other words, the two halves of the same duplex bearer are handled by different BBIs. Three further duplex bearers 5 (L3-L8, say, not shown in the drawing), are connected to the remaining inputs of the BBIs 172 in a similar fashion.

Using the standby selector 180 any one of the bearers can be switched through to the standby chassis 130 in the event of a failure of a probe unit in one of the main probe 10 chassis 100-1, 100-2. It will be appreciated that, if there is a failure of a complete probe chassis, then only one of the bearers can be switched through to the standby probe. In a larger system with, say, 16 duplex bearers, four main probe chassis and two standby chassis, the bearers distributed by each BBI can be shared around the probe chassis so that each probe chassis processes one bearer from each BBI. Then 15 all four bearers can be switched to the standby probe in the event of a complete chassis failure.

It will be seen that the BBI offers increased resilience for users particularly when they have multiple departments wanting to look at the same bearers. The size of the BBI 20 used is not critical and practical considerations will influence the number of inputs and outputs. For example, the BBI could provide inputs for 16 duplex bearers, each being distributed to two or three outputs with four standby outputs. Where multiple standby circuits are used each will be capable of being independently switched to any of the inputs.

Figures 13A and 13B illustrate a process of upgrading the processing power of a network monitoring system without interrupting operation, using the facilities of the replicating devices (BBls 172) and probe chassis described above. Figure 13A shows an example of an "existing" system with one probe chassis 100-1. Four duplex bearer 30 signals are applied to inputs 140 of the chassis. Via the internal cross-point switch 80, each bearer signal is routed to one probe unit 90. With a view to further upgrades and fault tolerance, a broadband bridging isolator (BBI) 172 is included. Each bearer signal is received from a tap in the actual bearer (not shown) at a BBI input 176. The

same bearer signal is replicated at BBI outputs 178-1, 178-2 etc. The first set of outputs 178-1 are connected to the inputs 140 of the probe chassis. The second set of outputs 178-2 are not used in the initial configuration.

5 Figure 13B shows an expanded system, which includes a second probe chassis 100-2 also loaded with four probe units 90. Consequently there are now provided two probe units per bearer, increasing the processing power available per bearer. It is a simple task to migrate from the original configuration in Figure 13A to the new one shown in Figure 13B: Step I Install the extra chassis 100-2 with the probe units, establishing the appropriate power supply and LAN communications.

Step 2 - Connect two of the duplicate BBI outputs 186 to inputs of the extra chassis 100-2. (All four could be connected for redundancy if desired.) 15 Step 3 - Configure the new chassis 100-2 and probe units to monitor the two bearer signals in accordance with the desired applications.

Step 4 - Re-configure the original chassis to cease monitoring the corresponding two bearer signals of the first set of outputs 178-1 (188 in Figure 1313). (The processing capacity freed in the original chassis 20 100-l can then be assigned to expanded monitoring of the two duplex bearer signals which remain connected to the BBI outputs 178-1.) Step 5 - Remove the connections 188 no longer being used. (These connections could be left for redundancy if desired.) 25 In this example the processing power has been doubled from one probe unit per bearer to two probe units per bearer but it can be seen that such a scheme could be easily extended by connecting further chassis. At no point has the original monitoring capacity been lost, and at no point have the bearers themselves (not shown) been disrupted. Thus, for example, a module of one probe unit can be removed for upgrade 30 while other units continue their own operations. If there is spare capacity, one of the other units can step in to provide the functionality of the unit being replaced. After Step 2, the entire first chassis 100-1 could be removed and replaced while the second chassis 100-2 steps in to perform its functions. Variations on this method are

practically infinite, and can also be used for other types of migration, such as when increasing, system reliability.

The hardware and methods used in these steps can be arranged to comply with "hot 5 swap" standards as defined earlier. The system of Figures 13A and 13B, and of course any of the systems described above, may further provide automatic sensing of the removal (or failure) of a probe unit (or entire chassis), and automatic re-

configuration of switches and re-programming of probe units to resume critical monitoring functions with minimum delay. Preferably, of course, the engineer would 10 instruct the re-prograrnming prior to any planned removal of a probe unit module. A further level of protection, which allows completely uninterrupted operation with minimum staff involvement, is to sense the unlocking of a processing card prior to actual removal, to reconfigure other units to take over the functions of the affected module, and then to signal to the engineer that actual removal is permitted. This will 15 be illustrated further below with reference to Figure 1 5A.

Multi-channel Probe Apparatus - Functional arrangement Figure 14 is a functional block schematic diagram of a multi-channel probe apparatus suitable for implementing the systems shown in Figures 4 to 13A and 13B. Like 20 numerals depict like elements. All of the modules shown in Figure 14 and their interconnections are ideally separately replaceable, and housed within a self-contained enclosure of standard rack-mount dimensions. The actual physical configuration of the network probe unit modules in a chassis with special backplane will be described later. A network interface module 200 provides optical fibre connectors for the incoming bearer signals EXT 1-8 (70-1 to 70-N in Figure 3), and performs optical to electrical conversion. A cross-point switch 80 provides a means of linking these connections to appropriate probe units 90. Each input of a probe unit can be regarded as a separate 30 monitoring channel CHI, CH2 etc. As mentioned previously, each probe unit may in fact accept plural signals for processing simultaneously, and these may or may not be selectable independently, or grouped into larger monitoring channels. Additional optical outputs EXT 9,10 are provided to act as "spare" outputs (corresponding to 85

in Figure 4). In the embodiment, each probe unit 90 controls the crosspoint switch 80 to feed its inputs (forming channel CHI, 9, 3 or 4 etc.) with a bearer signal selected from among the incoming signals EXT 1-8. This selection may be pre-programmed in the apparatus, or may be set by remote command over a LAN. Each probe unit (90) 5 is implemented in two parts, which may conveniently be realised as a specialized packet processor 150 and a general purpose single board computer SBC 160 module.

There are provided four packet processors 150 each capable of filtering and pre-

processing eight half duplex bearer signals at full rate, and four SBCs 160 capable of further processing the results obtained by the packet processors. The packet 10 processors 150 comprise dedicated data processing hardware, while the SBC can be implemented using industry standard processors or other general purpose processing modules. The packet processors 150 are closely coupled by individual peripheral buses to their respective SBCs 160 so as to form self-contained processing systems, each packet processor acting as a peripheral to its "host" SBC. Each packet 15 processor 150 carries out a high speed time critical cell and packet processing including data aggregation and filtering. A second level of aggregation is carried out in the SBC 160.

LAN and chassis management modules 230, 235 (which in the implementation 20 described later are combined on a single card) provide central hardware platform management and onward communication of the processing results. For this onward communication, multiple redundant LAN interfaces are provided between every SBC 160 and the LAN management module 230 across the backplane. The LAN management function has four LAN inputs (one from each SBC) and four LAN 25 outputs (for redundancy) to the monitoring LAN network. Multiple connections are provided as different SBC manufacturers use different pin connectors on their connectors. For any particular manufacturer there is normally only one connection between the SBC 160 and the LAN management module 230. The dual redundant LAN interfaces are provided for reliability in reporting the filtered and processed data 30 to the next level of aggregation (site processor 40 in Figure 2). This next level can be located remotely. Each outgoing LAN interface is connectable to a completely independent network, LAN A or LAN B to ensure reporting in case of LAN outages.

( 23 In case of dual outages, the apparatus has buffer space for a substantial quantity of reporting data.

The chassis management module 235 oversees monitoring and wiring functions via 5 (for example) an I2C bus using various protocols. Although l-C is normally defined as a shared bus system, each probe unit for reliability has its own I2C connection direct to the management module. The management module can also instruct the cross-point switch to activate the "spare" output (labelled as monitoring channels CH9,10 and optical outputs EXT 9,10) when it detects failure of one of the probe unit 10 modules. This operation can also be carried out under instruction via LAN.

The network probe having the architecture described above must be realised in a physical environment capable of fulfilling the functional specifications and other

hardware platform considerations such as the telecommunications environment it is to 15 be deployed in. A novel chassis (or "cardcage") configuration has been developed to meet these requirements within a compact rack-mountable enclosure. The chassis is deployed as a fundamental component of the data collection and processing system.

Multi-channel Probe Apparatus - Physical Implementation 20 Figures 15A, B and C show how the probe architecture of Figure 14 can be implemented with a novel chassis, in a particularly compact and reliable manner. To support the network probe architecture for this embodiment there is also provided a custom backplane 190. Figure 16 shows which signals are carried by the backplane, and which modules provide the external connections. Similar reference signs are used 25 as in Figure 14, where possible.

Referring to Figure 16 for an overview of the functional architecture, the similarities with the architecture of Figure 14 will be apparent. The network probe apparatus again has eight external optical terminals for signals EXT 1-8 to be monitored. These 30 are received at a network interface module 200. A cross-point switch module 80 receives eight corresponding electrical signals EXT 1'-S' from module 200 through the backplane 190. Switch 80 has ten signal outputs, forming eight monitoring

channels CH1-8 plus two external outputs (CH9,10). Four packet processor modules 150-1 to 150-4 receive pairs of these channels CHI,2, CH3,4 etc. respectively.

CH9,10 signals are fed back to the network interface module 200, and reproduced in optical form at external terminals EXT 910. All internal connections just mentioned 5 are made through the backplane via transmission lines in the backplane 190. Each packet processor is paired with a respective SBC 160-1 to 160-4 by individual cPCI bus connections in the backplane.

A LAN & Chassis Management module 230 is provided, which is connected to the 10 other modules by I2C buses in the backplane, and by LAN connections. A LAN interface module 270 provides external LAN connections for the onward reporting of processing results. Also provided is a fan assembly 400 for cooling and a power supply (PSU) module 420.

15 Referring to the views in Figure l5A chassis IOO carries a backplane 190 and provides support and interconnections for various processing modules Conventionally, the processing modules are arranged in slots to the "front" of the backplane, and space behind the backplane in a telecommunications application is occupied by specialized interconnect. This specialized interconnect may include 20 further removable DO cards referred to as "transition cards". The power supply and fans are generally located above and/or below the main card space, and the cards (processing modules) are arranged vertically in a vertical airflow. These factors make for a very tall enclosure, and one which is far deeper than the ideals of 300mm or so in the NEBS environment. The present chassis features significant departures from 25 the conventional design, which result in a compact and particularly shallow enclosure.

In the present chassis, the power supply module (PSU) 420 is located in a shallow space behind the backplane 190. The processing modules 150-1, 160letc. at the front of the backplane are, moreover, arranged to lie horizontally, with their long axes 30 parallel to the front panel. The cooling fans 400 are placed to one side of the chassis.

Airflow enters the chassis at the front at 410 and flows horizontally over the components to be cooled, before exiting at the rear at 412. This arrangement gives the chassis a high cooling capability while at the same time not extending the size of the

( 25 chassis beyond the desired dimensions. The outer dimensions and front flange of the housing allow the chassis to be mounted on a standard 19 inch (483mm) equipment rack, with just SU height. Since the width of the enclosure is fixed by standard racl; dimensions, but the height is freely selectable, the horizontal arrangement allows the 5 space occupied by the enclosure to be matched to the number of processor slots required by the application. In the known vertical orientation, a chassis which provides ten slots must be just as high as one which provides twenty slots, and additional height must be allowed for airflow arrangements at top and bottom.

10 Referring also to Figures 1SB and 15C, there are ten card slots labelled Fl-F10 on the front side of the backplane 190. There are two shallow slots Bl and B2 to the rear of the backplane l 9O, back-to back with Fg and F10 respectively. The front slot dimensions correspond to those of the cPCI standard, which also defines up to five standard electrical connectors referred to generally as J1 to J5, as marked in Figures 15 15B and 15C. It will be known to the skilled reader that connectors Jl and J2 have 110 pins each, and the functions of these are specified in the cPCI standard (version PICMG 2.0 R2. 1 (May 1 st 1998)).

Other connector positions are used differently by different manufacturers. I 20 Eight of the front slots (F1-F8) support the packet processor/SBC cards in pairs. The cards are removable using 'hot swap' techniques, as previously outlined, using thumb levers 195 to lock/unlock the cards and to signal that a card is to be inserted/removed.

The other two front slots F9 and F10 are used for cross-point switch 80 and LAN/Management card 230 respectively. Slots Fl to F8 comply with the cPCI insofar I 25 as connectors J 1, J2, J3 and J5 are concerned. Other bus standards such as VME could be also be used. The other slots F9 and F10 are unique to this design. All ofthe cPCI connections are standard and the connectivity, routing and termination reguirements are taken from the cPCI standard specification. Keying requirements are also taken

from the cPCI standard. The cPCI bus does not connect all modules, however: it is 30 split into four independent buses CPCI1-4 to form four self-contained host-peripheral processing sub-systems. Failure of any packet processor/SBC combination will not affect the other three probe units.

Each of the cards is hot-swappable and will automatically recover from any reconfiguration. Moreover, by providing switches responsive to operation of the thumb levers 195, prior to physical removal of the card, the system can be warned of impending removal of an module. This warning can be used to trigger automatic re 5 routing of the affected monitoring channel(s). The engineer replacing the card can be instructed to await a visual signal on the front panel of the card or elsewhere, before completing the removal of the card. This signal can be sent by the LAN/Management module 270, or by a remote controlling site. This scheme allows easy operation for the engineer, without any interruption of the monitoring functions, and without 10 special steps to command the re- routing. Such commands might otherwise require the co-ordination of actions at the local site with staff at a central site, or at best the same engineer might be required to move between the chassis being worked upon and a nearby PC workstation.

15 As mentioned above, the upper two front slots (F 10, F9) hold the LAN & Management module 230 and the cross-point switch 80 respectively. Slot Bl (behind F9) carries a Network Transition card forming network interface module 200, while the LAN interface 270 in slot B2 (behind F10) carries the LAN connectors. All external connections to the apparatus are provided by special transition cards in these 20 rear slots, and routed through the backplane. No cabling needs to reach directly the rear of the individual probe unit slots. No cabling at all is required to the front of the enclosure. This is not only tidy externally of the housing, but leaves a clear volume behind the backplane which can be occupied by the PSU 420, shown cut-away in Figure 1 5C, yielding a substantial space saving over conventional designs and giving 25 greater ease of maintenance. The rearslot positions Bl, B2 are slightly wider, to accommodate the PSU connectors 422.

The J4 position in the backplane is customised to route high integrity network signals (labelled "RI;" in Figure 15B). These are transported on custom corrections not 30 within cPC1 standards. Figure 1 5B shows schematically how these connectors transport the bearer signals in monitoring channels CHI etc. from the cross-point switch 80 in slot F9 to the appropriate packet processors 150-1 etc. in slots F2, F4, F6, F8. The external bearer signals EXTI'-8' in electrical form can be seen passing

( 97 through the backplane from the cross-point switch 80 (in slot F9) to the network interface module 200 (B1). These high speed, high-integrity signals are carried via appropriately designed transmission lines in the printed wiring of the backplane 190.

The variation in transmission delay between channels in the chassis is not significant 5 for the applications envisaged. However, in order to avoid phase errors it is still important to ensure that each half of any differential signal is routed from its source to its destination using essentially equal delays. To ensure this, the delays must be matched to the packet processors for each set for the backplane and cross-point switch combination. 10 It is important to note that these monitoring channels are carried independently on point-to point connections, rather than through any shared bus such as is provided in the H. 1 10 protocol for computer telephony.

The backplane also carries 12C buses (SMB protocol) and the LAN wiring. These are 15 carried to each SBC 160-1 etc. either in the J3 position or the J5 position, depending on the manufacturer of the particular SBC, as described later. The LAN interface module 270 provides the apparatus with two external LAN ports for communications to the next layer of data processing aggregation, for example a site processor.

20 Connectivity is achieved using two LANs (A and B) at 100 BaseT for a cardcage.

The LAN l/O can be arranged to provide redundant connection to the external host computer 40. This may be done, for example, by using four internal LAN connections and four external LAN connections routed via different segments of the LAN 60. It is therefore possible to switch any SBC to either of the LAN connections 25 such that any SBC may be on any one connection or split between connections. This arrangement may be changed dynamically according to circumstances, as in the case of an error occurring, and allows different combinations of load sharing and redundancy. Additionally, this allows the probe processors to communicate with each other without going on the external LAN. However, this level of redundancy in the 30 LAN connection cannot be achieved if the total data from the probe processors exceeds the capacity of any one external LAN connection.

An external timing port (not shown in Figure 16) is additionally provided for accurately time-stamping the data in the packet processor. The signal is derived from any suitable source, for example a GPS receiver giving a I pulse per second input. It 5 is also possible to generate this signal using one of the packet processor cards, where one packet processor becomes a master card and the others can synchronise to it.

The individual modules will now be described in detail, with reference to Figures 17 19. This will further clarify the inter-relationships between them, and the role of the 10 backplane 190 and chassis 100.

Cross-Point Switch Module 80 Figure 17 is a block diagram of the crosspoint switch 80 and shows also the network line interfaces 300 (RX) and 310 (TX) provided on the network interface module 200.

There are eight optical line receiver interfaces 300 provided within module 200. There Is are thus eight bearer signals which are conditioned on the transition card (module 200) and transmitted in electrical fond EXT 1'-8' directly through the backplane 190 to the cross-point switch card 80. Ten individually configurable multiplexers (selectors) M are provided, each freely selecting one of the eight inputs. Each monitoring charmer (CHI-8) and hence each packet processor 150 can receive any of 20 the eight incoming network signals (EXTI'-8').

The outputs to the packet processors (CHI-CH4) are via the backplane 190 (position J4, Figure 1 SB as described above) and may follow, amongst others, DS3/OC3/OC12/OC48 electrical standards or utilise a suitable proprietary interface.

25 Each packet processor module 150 controls its own pair of multiplexers M directly.

The external optical outputs EXT 9,10 are provided via transmit interface 310 of the module 200 for connecting to a spare chassis (as in Figure 8). These outputs can be configured to be any of the eight inputs, using a further pair of multiplexers M which 30 are controlled by the LAN/\lanagement Module 230. In this way, the spare processor or chassis 130 mentioned above can be activated in case of processor failure. In an alternative implementation, the selection of these external output signals CH9 and

( 29 CHIC can be perfonned entirely on the network interface module 200, without passing through the backplane or the cross-point switch module 80.

Although functionally each multiplexer M of the cross-point switch is described and 5 shown as being controlled by a respective packet processor 15(), in the present embodiment this control is conducted via the LAN & management module 230.

Commands or requests for a particular connection can be sent to the LAN & management module from the packet processor (or associated SBC 160) via the LAN connections, or I2C buses, provided in connectors J3 or J5.

10 Packet Processor Module 150 Figure 18 is a block diagram of one of the packet processor modules 150 of the apparatus. The main purpose of packet processor (PP) 150 is to capture data from the network interface. This data is then processed, analysed and filtered before being sent to a SBC via a local cPCI bus. Packet processor 150 complies with Compact PCI Hot 15 Swap specification PICNIG-2.1 R 1.0, mentioned above. Packet processor 150 here

described is designed to work up to 622Mbits/s using a Sonet/SDH frame structure carrying ATM cells using AAL5 Segmentation And Reassembly (SAR). Other embodiments can be employed using the same architecture, for example to operate at OC48 (2.4 Gbit.s-').

The following description makes reference to a single "half" of the twochannel

packet processor module 150, and to a single packet processor/SBC pair only (single channel). The chassis as described supports four such packet processor/SBC pairs, and each packet processor comprises two processing means to handle multiple bearer 25 signals (multiple monitoring channels).

It is possible for the packet processor 150 to filter the incoming data. This is essential due to the very high speed of the broadband network interfaces being monitored, such as would be the case for OC-3 and above. The incoming signals are processed by the 30 packet processor, this generally taking the form of time stamping the data and performing filtering based on appropriate fields in the data. Different fields can be

chosen accordingly, for example ATM cells by VPL{VCI (VC) number, IP by IP

t 30 address, or filtering can be based on other, user defined fields. It is necessary to

provide the appropriate means to recover the clock and data from the incoming signal, as the means needed varies dependent on link media and coding schemes used.

In a typical example using ATM, ATM cells are processed by VPI/VCI (VC) number.

5 The packet processor is provided with means 320 to recover the clock and data from the incoming signal bit stream. The data is then 'deframed' at a transmission convergence sub-layer 330 to extract the ATM cells. The ATM cells are then time-

starnped 340 and then buffered in a First In First Out (FIFO) buffer 350 to smooth the rate of burst type data. Cells from this FIFO buffer are then passed sequentially 10 to an ATM cell processor 360. The packet processor can store ATM cells to allow it to re-assemble cells into a message - a Protocol Data Unit (PDU). Only when the PDU has been assembled will it be sent to the SBC. Before assembly, the VC of a cell is checked to ascertain what actions should be taller, for example, to discard cell, assemble PDU, or pass on the raw cell.

Data is transferred into the SBC memory using cPC1 DMA transfers to a data buffer 38.This ensures the very high data throughput that may be required if large amounts of data are being stored. The main limitation in the amount of data that is processed will be due to the applications software that processes it. It is therefore the 20 responsibility of the packet processor 150 to carry out as much pre-processing of the data as possible so that only that data which is relevant is passed up into the application domain.

The first function of the packet processor 150 is to locate the instructions for 25 processing the VC (virtual channel) to which the cell belongs. To do this it must convert the very large VPIJVCI of the cell into a manageable pointer to its associated processing instructions (VC # key). This is done using a hashing algorithm by hash generator 390, which in turn uses a VC hash table. Processor l SO, having located the instructions, can then process the cell.

Processing the cell involves updating status information for the particular VC (e.g. cell count) and forwarding the cell and any associated information (e.g. 'Protocol Data Unit (PDU) received") to the SBC 160 if required. By reading the status of a

! 31 particular VC, the processor can vary its action depending on the current status of that VC (e.;,. providing summary information after first cell received). Cell processor 360

also requires certain configurable information which is applicable to all of its processing functions regardless of VC (e.g. buffer sizes) and this 'global' 5 configuration is accessible via a global configuration store.

A time stamping function 340 can be synchronised to an external GPS time signal or can be adjusted by the SBC 160. The SBC can also configure and monitor the deframer' (e.g. set up frame formats and monitor alarms) as well as select the optical 10 inputs (EXT 1-8) to be monitored. Packet processor 150 provides all of the necessary cPCI interface functions.

Each packet processor board 150-1 etc. is removable without disconnecting power from the chassis. This board will not impact the performance of other boards in the 15 chassis other than the associated SBC. The microprocessor notifies the presence or absence of the packet processor and processes any signal loss conditions generated by the packet processor.

Single Board Computer (SBCJ Modules 160 20 The SBC module 160 is not shown in detail herein, being a general-purpose processing module, examples including the Motorola CPV5350, FORCE CPCI-730, and SMT NAPA. The SBC 150 is a flexible, programmable device. In this specific embodiment two such devices may exist on one cPCI card, in the form of "piggyback" modules (PMCs). The 100 BaseT interfaces, disk memory etc. may also 25 be in the form of PMCs. As already described, communications via the cPCI bus (Jl/J2) on the input side and via the LAN port on the output side and all other connections are via the backplane at the rear, unless for diagnostic purposes for which an RS-232 port is provided at the front.

30 LAN & Chassis Management Module 230 Figure 19 is a block diagram of the combined LAN and chassis management card for the network probe as has been described. Module 230 performs a number of key management functions, although the probe units 150/160 can be commanded

! 32 independently from a remote location, via the LAN interface. The card firstly provides a means for routing probe units SMB and LAN connections, including dual independent LAN switches 500A and 5 00B to route the LAN connections with redundancy and sufficecnt bandwidth to the outside world.

s On the chassis management side, a Field Programmable Gate Array (FGPA) 510

within this module performs the following functions: (520) I2C and SMB communications, with reference to chassis configuration storage registers 530 10 (540) 'magic packet' handling, for resetting the modules remotely in the event that the higher level network protocols "hang up"; (550) environmental control and monitoring of fan speed and PSU & CPU temperatures functions to ensure optimal operating conditions for the chassis, and preferably also to minimise unnecessary power consumption and fan noise; A hardware watchdog feature 560 is also included to monitor the activity of all modules and take appropriate action in the event that any of them becomes inactive or unresponsive. This includes the ability to reset modules.

90 Finally, the management module implements at 580 '4Multivendor interconnect", whereby differences in the usage of cPCI connectors pins (or whatever standard is adopted) between a selection of processor vendors can be accommodated.

As mentioned previously, the chassis carries at some locations, cPCI processor 25 modules from a choice of selected vendors, but these are coupled via cPCI bus to special peripheral cards. While such cards are known in principle, and the processor-

peripheral bus is fully specified, the apparatus described does not have a conventional interconnect arrangement for the broadband signals, multiple redundant LAN connections and so forth. Even for the same functions, such as the LAN signals and 30 I2C/SMB protocol for hardware monitoring, different SBC vendors place the relevant signals on different pins of the cPCI connector set, particularly they may be on certain pins in J3 with some vendors, and on various locations in J5 with others.

Conventionally, this means the system designer has to restrict the user's choice of

SBC modules to those of one vendor, or a group of vendors who have adopted the same pin assignment for LAN and SMB functions, besides the standard assignments for J I, and J2 which are specified for all cPCl products.

5 To overcome this obstacle a modular Multivendor Interconnect (MVI) solution may be applied. The MVI module 580 is effectively four productspecific configuration cards that individually route the LAN and SMB signals received from each SBC 160-

I etc. to the correct locations on the LAN/Management cards. One MVl card exists for each processor. These are carried piggyback on the LAN/management module 10 230, and each is accessible from the front panel of the enclosure. The backplane in locations J3 and J5 includes sufficient connectors, pins and interconnections between the modules to satisfy a number of different possible SBC types. Needless to say, when replacing a processor card with one of a different type, the corresponding MVI configuration card needs exchanging also.

An alternative scheme to switch the card connection automatically based on vendor ID codes read via the backplane can also be envisaged. In a particular embodiment, for example, the "Geographic Address" pins defined in the cPCI connector specifications may be available for signalling (under control of a start-up program)

20 which type of SBC 160 is in a given slot. The routing of SMB, LAN and other signals can then be switched electronically under control of programs in the LAN & management card 230.

Conclusion

25 Those skilled in the art will recognise that the invention in any of its aspects is not limited to the specific embodiments disclosed herein. In particular, unless specified in the claims, the invention is in no way limited to any particular type of processor, type of network to be monitored, protocol, choice of physical interconnect, choice of peripheral bus (cPCI v. VME, parallel v. serial etc.), number of bearers per chassis, 30 number of bearers per monitoring channel, number of monitoring channels per probe unit.

Tile fact that independent processor subsystems are arranged in the chassis allows multiple data paths from the telecommunications network to the LAN network' thereby providing inherent redundancy. On the other hand, for other applications such as computer telephony, reliability and availability may not be so critical as in the 5 applications addressed by the present embodiment. For such applications, a similar chassis arrangement but with H. 1 10 bus in the backplane may be very useful.

Similarly, the cPCI bus, I-C bus and/or LAN interconnect may be shared among all the modules.

10 Each aspect of the invention mentioned above is to be considered as independent, such that the probe functional architecture can be used irrespective of the chassis configuration, and vice versa. On the other hand, the reader will recognise that the specific combinations of these features offers in a highly desirable instrumentation system, which provides the desired functionality, reliability and availability levels in a 15 compact and scalable architecture.

In the specific embodiments described herein, each probe unit comprising first and second processor modules (the packet processor and SBC respectively) is configured to monitor simplex and duplex bearers. The invention, in any of its aspects, is not 20 limited to such embodiments. In particular, each probe unit may be adapted to process one or more individual bearer signals. In the case of lower speed protocol signals the bearer signals can be multiplexed together (for example within the cross-point switch module 80 or network interface module 200) to take full advantage of the internal bandwidth of the architecture.

Attention is directed to co-pending UK patent application no. 99 23 143.3 (publication no. 2 354 883) relating to aspects of the above description as defined in the claims of

that application.

Claims (3)

! 3s CLAIMS

1. A multi-processor equipment enclosure having a housing and a backplane interconnection for a plurality of processing modules and a management module, the 5 backplane interconnection including generic portions standardized over a range of processing modules and other portions specific to different processing modules within said range, wherein said management module is arranged to sense the specific type of a processing module by using protocols implemented by the modules via connections in the generic portion of the interconnection, and to control routing of communication 10 and management signals via the backplane in accordance with the sensed specific type of each processing module.

2. An enclosure as claimed in claim 1, wherein said type sensing protocols are implemented via geographic address lines in the standardised portions of a compact 15 PCI backplane.

3. A method of configuring a multi-processor equipment to operate with different ones of a range of processing modules, said equipment having a backplane interconnection for a plurality of processing modules and a management module, and 20 the backplane interconnection including generic portions standardised over the range of processing modules and other portions specific to different processing modules within said range, comprising the steps of: causing said management module to sense the specific type of a processing module by using protocols implemented by the modules via connections in the 25 generic portion of the interconnection; and controlling routing of communication and managemcut signals via the backplane in accordance with the sensed specific type of each processing module.