In Part One, we looked at how AI may change network demand.  The impact is not always a simple increase in bandwidth.  AI can alter where traffic flows, how long transfers last and how the network behaves during periods of peak demand.

We also considered the limits of Quality of Service, prioritisation and logical separation.  These techniques can make better use of available capacity, but they cannot create more of it.

Once the constraint has been identified, the next question is what to do about it.

Optimise what is already there

The most sensible starting point is often the existing network.

QoS can protect delay-sensitive services.  Traffic shaping can control how bandwidth is shared.  Routing changes may prevent traffic from taking an inefficient path.  Logical separation can also reduce interference between applications, departments or customers.

The main strength of this approach is that it may improve performance without major infrastructure investment.  Changes can be introduced gradually, monitored and adjusted as traffic patterns develop.

However, optimisation has limits.

QoS changes which traffic receives preferential treatment.  Traffic shaping controls demand by delaying or restricting some flows.  Logical separation improves control and visibility, but does not increase the physical capacity underneath.

These techniques work best where the issue is contention, poor configuration or uneven use of resources.  They are less effective where links regularly operate near capacity, or where most traffic is considered business-critical.

At that point, optimisation may improve how congestion is experienced without removing it.

Target the real bottleneck

Where the constraint is clearly understood, a targeted upgrade can be highly effective.

This could mean replacing a low-capacity uplink, increasing an interface speed, upgrading a switch or adding a higher-capacity circuit between two locations.

The advantage is precision.  Investment is directed at the part of the network causing the problem.  This can be faster, less disruptive and more economical than a wider redesign.

The risk is that the bottleneck simply moves.

A faster access link may expose a slower aggregation layer.  A higher-capacity data-centre connection may reveal limitations within switching or storage.  Increasing WAN capacity may move congestion closer to the application.

For that reason, targeted upgrades should be assessed across the complete end-to-end path.

Add circuits or fibre

Sometimes the simplest answer is to add physical capacity.

A second circuit may increase bandwidth, improve resilience or allow traffic to be divided between services.  Additional fibre can provide a clean route for expansion and greater design freedom.

The strength of this approach is simplicity.  Separate physical paths can be easier to understand and troubleshoot than a complex shared architecture.

Availability is often the deciding factor.  Spare fibre may not exist.  Installing more can be costly, disruptive or impractical.  Additional leased circuits also introduce recurring costs, lead times and new contractual commitments.

A second link does not automatically provide resilience either.  The services must follow genuinely diverse routes and avoid shared ducts, entry points, equipment and power sources.  Otherwise, apparent diversity may still contain a common point of failure.

Increase capacity over existing fibre

Where fibre is available, but fibre count is limited, wavelength-based technologies may offer a more scalable answer.

CWDM and DWDM allow multiple services or wavelengths to be carried over the same fibre pair.  This can increase capacity without the cost and disruption of installing new fibre.

CWDM can provide a relatively simple and cost-effective way to add several channels over suitable distances.  It is often well suited to straightforward point-to-point requirements where capacity needs are moderate and future growth is reasonably predictable.

DWDM supports many more channels and higher capacities over longer distances.  It may also provide a stronger platform for expansion, protection and service management.

The strength of both approaches is fibre efficiency.  The limitation is that the design becomes more dependent on optical engineering.

Distance, fibre loss, connector performance, passive losses and optical power budgets all need to be considered.  Longer-distance or higher-capacity systems may also require amplification, active monitoring or more advanced protection.

The choice should therefore follow the application.  CWDM and DWDM are not simply ways to add more bandwidth.  They are tools for creating a practical capacity and growth path over limited fibre resources.

Make the passive network work harder

Passive Optical Networks provide another example of how more value can be extracted from the same fibre infrastructure.

In a PON, multiple users or services share a common feeder fibre and passive splitter network.  Dynamic Bandwidth Allocation can distribute upstream capacity according to demand, but it is still managing a finite resource.

A more strategic opportunity comes from wavelength coexistence.

GPON and XGS-PON can operate over the same Optical Distribution Network using different wavelength bands.  Existing GPON services can remain in place, while selected higher-demand customers move onto XGS-PON.

This allows migration to follow demand.  The fibre, splitters and customer distribution infrastructure remain in service, while the active OLT and ONT equipment changes.

It can also create separate capacity domains over the same physical network.  A customer moved from GPON to XGS-PON no longer consumes capacity from the original GPON system.  The outside plant is shared, but the transmission capacity is not one common pool.

Looking further ahead, 50G-PON extends the same principle.  GPON, XGS-PON and 50G-PON can coexist over a suitably designed common fibre and splitter infrastructure, subject to the wavelength plan, equipment compatibility and available optical-loss budget.

50G-PON has not yet reached the wider adoption of GPON or XGS-PON.  Its early use is likely to remain selective, particularly for high-capacity business services, mobile transport and other demanding applications.

Its wider significance is architectural.  A passive network installed today may support several generations of active technology without repeatedly rebuilding the outside plant.

That opportunity depends on decisions made at the design stage.  Coexistence components introduce additional loss.  Split ratios, distance, connector condition and available optical margin can all affect whether future overlays remain practical.

The ODN should therefore be viewed as a long-life infrastructure platform, rather than something tied to one generation of PON electronics.

Avoid solving only today’s problem

One of the most common risks is designing around the immediate requirement alone.

A low-cost point-to-point solution may meet today’s need.  However, if it has no practical upgrade path, it may need to be replaced when the next service, site or capacity step is added.

At the other extreme, it is possible to over-engineer.  A large modular platform may offer impressive capability, but much of it may never be used.

The aim is not to choose the biggest solution or the cheapest one.  It is to choose an architecture with a proportionate growth path.

That means considering the next interface speed, the next wavelength, the next PON generation and the next resilience requirement.

There is no universal point at which QoS should give way to more capacity.  Nor is there one optical solution that suits every organisation.

Where traffic is bursty and priorities are clear, optimisation may be enough.  Where one component is constrained, a targeted upgrade may provide the best return.  Where fibre is plentiful, physical expansion may be simplest.  Where fibre is limited and demand is growing, CWDM, DWDM or a multi-generation PON strategy may provide the more scalable route.

AI may be the reason organisations are asking the question today.  The underlying challenge is broader.

How can the network meet the current requirement without limiting the organisation’s next step?

Discuss your network requirements

To discuss your current network requirements, capacity challenges or future expansion plans in more detail, contact the FTL technical sales team.

Telephone: +44 (0) 1344 752 222

Email: sales@fibre.co.uk