CONCEPT EXPLAINED

5.4.1 General Physical Topology

The physical topology of the LambdaPON network is based on a bi-directional ring. The rationale for considering bi-directional ring networks are:

(i) Ring networks can be incrementally scaled up from a smaller ring to a large ring with relative ease; by deploying and interconnecting more PRNs (within limits, as discussed later on in Chapter 8 and 9).

(ii) Routing in rings is simple, offering only the clockwise and anticlockwise

possibilities (assuming the ring is bi-directional). This simplifies the control algorithms, control software and the network node interfaces.

(iii) Various standards built around the use of rings such as the Token Ring and

Fibre Distributed Data Interface (FDDI) protocols are already in existence.

(iv) The reliability! of bi-directional rings has been shown to be high [84][85]

because two alternative routes exist between every pair of nodes.

(v) Several of the future networks planned for deployment involve rings as the

standard topology e.g. SDH rings [86].

The network consists of a number of connected network access nodes (from now on referred to as Passive Routing Nodes). Each Passive Routing Node (PRN) is completely passive and unpowered allowing greater reliability to be achieved. The network control functions are located at a centrally shared Network Routing Manager (NRM) and also at the peripheries of the network (within the user's terminals). The Passive Routing Nodes are interconnected via dynamically tunable filters whose sole task is to block wavelength pairs or groups of wavelength pairs in real time, depending on connection requests received from users (as shown in Figure 5.2). The tunable filters are remotely controllable by the Network Routing Manager. Users are connected to the PRNs via a broadcast star topology employing two fibres (one for upstream and one for downstream traffic). Each PRN is capable of serving several customers (of the order 256). This number is dependent on the components used within the PRN which effect the power budget and bit-error rate of the system. The significance of the numbers will become apparent once the results, chapter 8, and implementation issues, chapter 9, have been explained).

To clarify the concept of wavelength reuse within a LambdaPON, the interconnection of eight Passive Routing Nodes using two unidirectional fibre rings; one for clockwise and one for anticlockwise traffic is shown in Figure 5.2. The establishment of five unrelated independent circuit-switched calls^ between ten users (A to A', B to B', a to a ' etc.) using only four wavelengths is shown. Assume a user A wishing to establish a connection with user A'. User A transmits a control packet to the NRM3,

^The reliability of the LambdaPON network is discussed in greater detail in Chapter 9

^ The term "call" refers to any traffic generated which is transmitted into the network whether it be straightforward voice-telephony type traffic or some future multi-media service.

3 Details regarding the transmission of "signalling information and packets" to and the receipt of packets from the NRM are given in section 5.4.2

requesting the use of a wavelength pair operating at a bit-rate suitable for the service required. If the control packet is successfully received by the NRM and a wavelength

pair (consisting of the two wavelength and A.4) is allocated, by turning on filters

A1 and A8 to absorb out wavelength X^ and X^, the two wavelengths are localised to

that PON and hence can be reused elsewhere. In the event that two paths overlap a new pair of wavelengths to interconnect the users is required; as shown by the

connection between a and a '. Note that the diagram only shows the users involved in

the calls. In practice, several users can be connected to each PON and all the information entering or leaving a PON is broadcast to every user on that PON (the inherent nature of the PON design).

N etwork Routing M anager

1

B'

D*

W avelength X^ 0 Filter Not Turned On

-► W avelength X^ W avelength Xc ^ # W avelength X^ ^ Blocks X^ and X^ Blocks X^ and Blocks Àg, and X^

Passive Routing Node

For the network shown in Figure 5.2, the maximum number of sim ultaneous calls which could occur using just two wavelengths (simplex transm ission is assumed) would be 8. This would occur only if all 8 calls were routed through only one of the Passive Routing Nodes such as that by A to A' and C to C i.e. the source and destination users are served by the PRU. It is im portant to stress that whilst the concept has been applied to a ring topology it is equally applicable to any arbitrary network design as shown in Figure 5.3.

Figure 5.3 Wavelength Reuse Applied To Arbitrary Network Topologies

The eventual design and topology of the network will be dependent on a number of issues such as the w avelength routing and allocation algorithm s used, the power losses induced by com ponents within the network and the noise efficiency of the optical am plifiers (if used). More details about the form er are raised again in Chapters 6 and 8, whilst the functionality and a possible design of the PRN, and a few suitable com ponents which can be used as dynam ic tunable filters are given in Chapter 9.

5.4.2 Initiating/Receiving A Call And The Dynam ic Allocation O f W avelengths

Users wishing to initiate a call are required to first access the N etw ork Routing M anager to obtain

(i) the status the destination user(s) to be called e.g. engaged, out of order (ii) a wavelength pair, to be used when transmitting and receiving information.

Accessing the NRM is accomplished by transmitting a control channel packet on a

common control wavelength which, from now on, will be designated as wavelength 1 (À%). The Passive Routing Node is constructed such that it passively routes only control packets to the NRM.

One Network Routing M anager is allocated to serve each LambdaPON. Each NRM contains the following information about the LambdaPON:

(i) a routing table which keeps a record of the various possible routes between users; in a bi-directional ring network this would only be two.

(ii) the wavelength pairs in use on each of the routes; (iii) the state of every user (engaged / available);

(iii) the current state of the dynamically tunable filters (on / off);

In addition, the NRM has a direct link to each of the dynamic filter enabling it to control and configure them according to each connection established as necessary, to filter out the necessary wavelength(s).

In place of the single centralised NRM, several distributed NRMs could be used to control the network (one for each PRU). This would offer the advantages of improved performance since each NRM would be shared between fewer customers reducing contention. Distributed NRMs also offer improved reliability and expandability. However, the consequent trade-offs include increased management and control complexity and additional costs due to hardware replication. The advantages and disadvantages of using distributed database systems have been intensively studied over the years and for further information the reader is directed to reference [82].

If the NRM receives the control packet successfully (i.e. the packet is not corrupted or contended with packets transmitted by other users sharing the NRM), the requested information is retrieved and sent downstream from the NRM database to the users using a "status" packet on the same control wavelength channel X\.

To reduce contention for X\ another wavelength could be dedicated solely for the use

of the NRM to transmit information downstream to the users (called "status"

wavelength (X,2)). Contention resolution for X\ and the requirement of an additional

status wavelength X2 will be discussed in section 5.6 and calculated later in Chapter 7.

However, for now it will be assumed that A,i is used to both transmit information to and receive information from the NRM. Once the users have completed their call, they are required to inform the NRM by transmitting a call termination packet allowing the tunable filter to be reset and the wavelengths to be allocated to future connection request.

The control, status and call termination packet all consist of two parts, a header and a payload. The header section contains information which is accessible by everyone connected to the LambdaPON i.e. it is public domain information. The payload, on the other hand, is readable only by the intended party. Figure 5.4 shows the different packets and the information contained within the header and payload sections.

Control Packet Term ination Packet Î Header i '--- ;----!--- !--- Î Header I Address of user to be called

Service & Bit-rate required t Payload

4

W avelengths to be released Address of users in call

t Payload

4

Status Packet

Calling Party Addmsa Called Pâity Address . Misceilaneoiis e.g. ^ Wavelength allocated

Î

1

Figure 5.4 Packets Sent During Call Set Up And Completion

5.5 APPLYING THE LAM BDAPON W AVELENGTH REUSE CONCEPT