P2P vs. PON

Figure 4.8 - Upstream optical path penalties for XG-PON.

In the upstream directions, problems only arise if ONTs use FP lasers beyond 10 km. These results for the upstream, along with those for GPON, are expected when we consider that the fibre used, G.652, has zero dispersion around 1310 nm. For XG-PON, the upstream is at 1270 nm so the dispersion coefficient is still relatively low, which means the penalties only become menacing at larger distances. For the downstream direction, however, results show that XG-PON might have some difficulties. The dispersion coefficient at 1577 nm is around 19 ps/(nm*km). The main consequence of this, coupled with a line rate of 10 Gbps, is that the OLT now requires external modulation even for short distances.

Furthermore, even with the use of DFB lasers at the OLT, the penalty for spectral width seems to limit the PON’s range to just over 6 km.

Observing the results above, one can conclude that under these specifications, the migration to XG- PON must include the use of DFB lasers at the OLT. Also, while urban scenarios where the ONTs are only a few kilometres away (at most) from the OLT should be able to cope with the change, possible rural scenarios need different specifications.

between PON and P2P and how they affect the cost structures. Then the model developed for P2P planning is presented and the comparative results are shown.

4.3.1 Architectural Differences

Both topologies have been schematized in Figure 1.5. Some differences are fairly obvious when considering both designs. The first is that a P2P network does not require the passive splitters used by PONs. The fibre runs, however, are bound to be much higher than in PONs since there is a dedicated transmission medium for each terminal. What P2P vendors argue is that the infrastructure costs are similar since they use the same ducts, only P2P might need larger cables to support more fibres in the feeder section. One aspect where PONs clearly have the upper hand is the equipment at the CO. While a single OLT port supports 64 users in a PON, an Ethernet interface must exist for each subscriber in P2P.

This is not only more expensive in the cost of equipment, but it also takes up more space and has a higher energy consumption. According to [44], a PON consumes 120 kW to service about 16 000 subscribers with 100 Mbps each. For the same bit-rate, a P2P network uses 320 kW for the same 100 Mbps. However, when offering 1 Gbps to subscribers, the energy consumption only rises to 390 kW.

Regarding the complexity of the terminals, P2P benefits from the maturity of Ethernet technology and does not need complex multiple access management so the P2P equivalents to ONTs are considerably cheaper. Finally, P2P has a less stringent power budget (because it has no power division) so it can use that advantage to have cheaper lasers with less input power or use less costly fibre with greater attenuation.

4.3.2 ILP P2P Model

The P2P paradigm can be thought of as the optical “version” of the xDSL networks. In that sense, the model tries to mimic the deployments of xDSL Access Networks. To enable a direct with PON scenarios, the splitters locations needed an equivalent in the P2P model. Looking at copper deployments, those locations can again be Junction Boxes, only now they do not house any sort of equipment. They are simply derivations on the main fibre cables (exemplified in Figure 4.9) to serve smaller areas.

Figure 4.9 - Junction derivation in fibre cable.

Optimizing networks in this context does not make as much sense as it does for PONs because there is no aggregation point. To have a clear comparison, the model was developed under the assumption that the possible locations for Junction Boxes are now the possible locations for derivations. In reality, a derivation is easier to install because in theory it will not require further access to it like splitters might.

The capacity constraint considered was the maximum number of fibres reaching or leaving a single derivation. This suffers from the same problem as the PON formulation because it does not account for paths between derivations and terminals being shared with other paths along the way (because a linear model cannot know them beforehand). The model for P2P networks can be defined with the following parameters, for ONTs and possible derivation sites:

- Cost of connecting ONTs in building to derivation ;

! - Cost of placing a derivation at location ;

- Cost of connecting derivations between and ; - Maximum number of fibres in path between and ;

† - Very small number to guarantee coherent constraints;

_G - Number of fibre connections (ONTs) in building .

The integrality conditions correspond to:

I 2 h/!%!0!%n" "'" k

. j"/!%$0!k 1 (4.4)

Furthermore, variables ‡ ˆ and denote the number of fibres between each connection.

The objective function is:

?

! ? I

?

v

(4.5) Subject to:

G

(4.6)

_[

v

[

v

(4.7)

k . (4.8)

I k . (4.9)

I S † (4.10)

The formulation is similar to the single-stage PON one. The first term of the objective function accounts for the costs of connecting ONTs to derivations, the second term for the costs of derivation themselves and the third term for the costs per fibre of the links between derivations. The variables indicate the number of fibres in each connection and I denotes the existence or not of a derivation. We impose that every building must have _G connections (4.6), that the number of connections that “leave” a derivation equals the number of connections that “enter” it (4.7) and that each connection cannot have more than fibres passing through it (4.8). Constraints (4.9) and (4.10) act to make sure the I variables are zero if there are no fibres passing through a derivation and one if there are. † is a very small constant to ensure that I is not greater than one.

Of course, OPEX costs are not considered in the ILP because they are constant for a given set of users served (the number of active interfaces at the OLT is always the same), but they must be factored in the total cost when comparing to the PON cost. The same applies to the terminals’ cost for both the PON and P2P formulations. The cost structure assumed is based on [29] and is present in Table 4.2. The considered cost for OPEX is twice that of a PON for a single user (because of additional energy consumption, space and equipment at the CO) and the cost for a PON ONT was 245 €.

Table 4.2 - Costs for P2P. Source: [29].

Element Cost [€]

OPEX per user 1 150

Terminal equipment 105

Fibre 115 [/km]

Laying Fibre ]\. [/km]

Derivation 100

4.3.3 Results

This section details the results obtained for the comparison in the P2P and PON models. Naturally, both were compared in the same scenarios. Specifically, Maps 7 and 10 from Chapter 2 were used for urban deployments models, and a rural scenario with 600 users was set with an 8 km maximum distance to the OLT/CO. Splitters or derivations were placed within 6 km of the OLT/CO in the latter. Figure 4.10 shows the costs obtained.

Figure 4.10 - Total network costs for PON and P2P.

As one would expect, PONs’ lower OPEX and less fibre-intensive structure makes it the cheaper option in these scenarios. This is even more noticeable in the larger scenarios, where the high amount of users makes variable costs (higher in P2P) offset the fixed ones.

However, the comparison is not entirely fair, in the sense that we are comparing the cost of a technology that shares 2,5 Gbps by 64 users against one that can offer a dedicated connection to each one. A more balanced comparison would be to factor in the bandwidth in the calculations. Therefore, Figure 4.11 shows the results for the cost per Mbps offered. For P2P, 100 Mbps or 1 Gbps connections were the most viable options. Naturally, 1 Gbps terminals are more costly and OPEX is higher as mentioned in 4.3.1.

Figure 4.11 - Network cost per Mbps offered for PON and P2P.

0 5 10 15 20 25

Urban 1 (Map 7) Urban 2 (Map 10) Rural

Network Cost [millions of €]

Scenarios

Cost per Topology

PON P2P

0 50000 100000 150000 200000 250000 300000

Urban 1 (Map 7) Urban 2 (Map 10) Rural

Cost per Mbps offered [€]

Scenarios

Cost per Mbps offered

PON P2P 100Mb P2P 1Gb

The picture changes dramatically. Factoring in available bandwidth, P2P is less expensive on a per bit basis, because a GPON can provide at best 39Mbps to each user (with a 1:64 ratio). The cost per bit for 1 Gbps is much lower than the alternatives, although perhaps market needs do not yet justify 1 Gbps in the access.