On 12 May 2024, two major submarine cables, SEACOM and EASSy, were damaged off the coast of KwaZulu-Natal, South Africa. The incident significantly reduced Internet connectivity across several East African countries.
Yet no country experienced a total Internet outage.The reason was not a single backup cable. Other submarine and cross-border terrestrial routes remained available, while Internet Exchange Points (IXPs) and local content caches helped keep part of the traffic within the region.
The incident highlighted a fundamental point for Network Planning, Peering and OTT/cloud infrastructure teams: A resilient backbone is measured by what remains connected when one of its routes disappears.
That resilience depends on two things working together: the physical paths carrying traffic and the interconnection architecture determining where that traffic goes.
That requires looking beyond individual cables and considering the entire architecture: IP transit, peering, BGP, terrestrial backhaul, landing stations, PoPs, data centres and content localisation.
IP transit and peering serve different purposes
IP transit and peering are complementary, not competing, components of Internet connectivity.
IP transit allows a network to reach destinations it cannot access directly. For an operator or regional network, it therefore remains essential for connectivity beyond its directly interconnected peers.
Peering follows a different model: two networks exchange traffic directly. At an Internet Exchange Point (IXP), multiple networks can interconnect their Autonomous System Numbers (ASNs) and exchange traffic locally through bilateral sessions or multilateral peering arrangements.
The practical distinction lies in where traffic can be exchanged locally and where international transit remains necessary.
A resilient architecture can consequently combine:
- local peering for traffic that can be exchanged directly;
- IP transit for destinations that are not reachable through peering;
- multiple physical routes to reduce dependence on any single infrastructure path.
This distinction matters even as interconnection models evolve. A 2026 analysis by RIPE/Namex describes the peering market as being in transformation, with private network interconnects, in-network caching and edge-oriented architectures increasingly complementing traditional public peering.
Traffic tromboning: When local traffic takes an international detour
Consider two networks located within the same region. If they lack an appropriate local interconnection path, traffic between them may travel through an international hub before returning to its destination.This is commonly described as traffic tromboning.
The problem is not simply additional distance. Regional traffic can consume international transit capacity even when the destination is geographically nearby. Geographic proximity does not necessarily mean network proximity. A regional connection can still depend on an international path if the right interconnection is not available closer to the source and destination. Local peering can avoid this detour when a direct interconnection is available.
The historical experience of the Kenya Internet Exchange Point (KIXP) and Internet Exchange Point of Nigeria (IXPN) illustrates the potential impact.
Between 2012 and 2020, KIXP’s peak traffic increased from 1 Gbit/s to 19 Gbit/s, while the associated cost savings were estimated at approximately US$6 million per year. Over the same period, IXPN grew from 300 Mbit/s to 125 Gbit/s of peak traffic, with estimated annual savings reaching approximately US$40 million in 2020.
These are historical figures specific to two mature African IXP ecosystems. They should not be interpreted as a guaranteed saving for every operator.
Their value is elsewhere: they demonstrate how regional interconnection can keep traffic local instead of sending it unnecessarily through international transit.
Why peering also matters for network resilience
Traffic localisation is not only an economic consideration. It can also reduce dependence on international routes for traffic that remains within a country or region.
The May 2024 East Africa outage provides a useful example. When SEACOM and EASSy were damaged, traffic was rerouted through other available infrastructure. But the Internet Society also observed that functioning submarine routes quickly became congested because they were not necessarily capable of absorbing the additional load.
This is a critical distinction for Network Planning. A backup route is not genuinely resilient simply because it appears on a topology diagram. Its available capacity, physical independence, shared dependencies, routing policies and behaviour under failure all matter.
Redundancy should therefore be assessed under failure conditions, not only under normal operating conditions.
Physical route diversity: Two cables can still share the same risk
The other side of regional interconnection is physical path diversity. Two submarine cables may look independent on a network map while sharing a landing station, terrestrial corridor or upstream infrastructure.The same issue continues inland. Two apparently separate routes may eventually converge on the same backhaul, PoP or terrestrial corridor. A single physical failure can then affect both.
The March 2024 West Africa outage provides another illustration. A suspected underwater rock slide off Ivory Coast affected ACE, SAT-3, WACS and MainOne. Traffic had to be rerouted through other available infrastructure.
Having two cables is only the starting point. The more useful exercise is to test what happens if a landing area, terrestrial corridor or shared piece of infrastructure becomes unavailable. The same principle applies to interconnection: diversity on the physical layer only creates value when traffic can use those alternative paths at the routing and service layers.
That is the difference between nominal redundancy and effective path diversity.
IXPs, CDN and PoPs: Keeping content closer to users
Peering is only one component of the wider regional interconnection ecosystem.
IXPs can also provide access to CDN caches, allowing frequently requested content to be served locally rather than retrieved repeatedly over an international route. The Internet Society’s analysis of the 2024 East Africa outage specifically highlights the role of IXPs and local content caches in mitigating the impact of submarine cable outages.
User location is only part of the equation. Where content, applications and network infrastructure are interconnected matters just as much. For OTT and cloud networks, this also affects where traffic can be served, exchanged and rerouted when a regional path becomes unavailable. The closer these functions are to users and to diverse network paths, the less dependent the service becomes on a single international route.
Recent analysis of the peering ecosystem also suggests that IXPs are evolving rather than becoming obsolete, with resilience, redundancy and local interconnection remaining important functions even as content delivery models change.
In practice, this brings several layers together:
IXPs + carrier-neutral data centres + PoPs + CDN infrastructure + diversified IP transit
What does this mean for East Africa and the Indian Ocean?
Building resilience means looking at several interconnected layers:
- local peering: exchange directly with relevant networks, content providers and digital platforms available through regional IXPs.
- IP transit: maintain connectivity to networks and destinations that cannot be reached through direct peering.
- diversified regional routes: maintain multiple paths towards major African connectivity hubs instead of relying on a single international corridor.
- physical diversity: assess whether submarine cables, landing stations, terrestrial backhaul and PoPs genuinely provide independent paths.
- CDN and local caching: keep frequently accessed content closer to users and reduce unnecessary dependence on international connectivity.
- BGP and traffic engineering: translate physical diversity into routing behaviour that can actually be used when network conditions change.
This layered approach is particularly relevant across the Indian Ocean, where island geography and dependence on a limited number of submarine corridors can make route diversity a particularly important network-planning consideration. For island markets, the dependency on a limited number of submarine and terrestrial paths makes the relationship between physical diversity and regional interconnection particularly important. Madagascar and other Indian Ocean markets illustrate why additional routes and local interconnection need to be considered together rather than separately.
Wholesale connectivity adds another layer to this architecture, providing additional paths between networks and major connectivity hubs while complementing local peering and IXP infrastructure.
This is where Silver Links operates across Africa and the Indian Ocean. Its role is to provide additional connectivity paths that can complement local interconnection and contribute to greater route diversity.
Regional interconnection shapes the network architecture
The evolution of African connectivity is making the distinction between individual services less useful.
Transit, peering, IXPs, CDN caches, PoPs, terrestrial networks and submarine systems increasingly interact as parts of the same connectivity architecture.
The interconnection landscape is becoming more distributed, with public peering, private interconnection, caching and edge infrastructure increasingly operating alongside one another. RIPE/Namex’s 2026 analysis describes this evolution as a transformation of the peering market rather than a decline.
This shifts the focus from capacity alone to the paths traffic actually takes. Capacity still matters, but so do the points where traffic enters and leaves the network, the routes it follows, and what happens when one of those routes disappears.
Resilience is built path by path
The events of 2024 demonstrated both the value and the limits of network redundancy.
When SEACOM and EASSy were damaged, alternative submarine and terrestrial routes, IXPs and local content caches helped prevent a complete Internet outage in affected East African countries. At the same time, some alternative routes became congested under the additional load.
The lesson is straightforward. A backup route only adds resilience when it is physically independent, operationally usable and sufficiently provisioned to carry traffic when the primary path fails.
Local peering can keep regional traffic local. IP transit provides access to the rest of the Internet. CDNs bring content closer to users. Diverse physical routes provide alternative paths. And BGP turns these components into an operational routing strategy.
A resilient backbone is therefore not a collection of independent services. It is the architecture that connects them.
