top of page

Why one undersea cable break can slow South Africa’s Internet to a crawl


Although it seems logical to blame Internet service providers and platform operators like Disney+ for badly performing online services in South Africa, they aren’t the main culprits for the country’s current connectivity problems.


Internet users in South Africa are still experiencing poor performance on their broadband connections after a rock fall in the Congo Canyon caused a break in the WACS and SAT–3 undersea fibre cables.


The breaks occurred last Sunday (6 August) and have wreaked havoc on Internet connectivity in South Africa.


What is curious about this round of west coast subsea cable breaks, is that WACS is no longer the highest capacity link between South Africa and Europe.


Last year, the massive Google-owned Equiano and Facebook-backed 2Africa cables made their first landings in South Africa.


To put the relative sizes in perspective, until recently, the West Africa Cable System (WACS) represented around 39% of all available undersea fibre cable capacity landing in South Africa.


It is bigger than the Seacom and EASSy east coast cables, and much bigger than Telkom’s west-coast SAT–3 cable.


However, since 2021, it has been overtaken by the ACE and PEACE cables and been completely dwarfed by Equiano and 2Africa.


When the capacities of all these cables are added to the equation, WACS now only contributes around 3% of South Africa’s total available undersea capacity.

Equiano and 2Africa are the country’s new subsea kings, contributing around 33% and 40% of unlit capacity.


This raises the question — if WACS and SAT–3 have come to represent a vanishingly-small proportion of the international Internet capacity in South Africa, why is it having such a significant impact on people’s experience?


The answer is that it is not a technical limitation, but a financial one.

Nowadays, much of the content Internet users in South Africa access is stored locally.


However, the content owners and distributors still require international capacity to transfer the videos, audio, and images to South African content caches.

Video streaming services like YouTube, Netflix, Disney+, and Amazon Prime Video all have local content delivery nodes.


Social networks, especially multimedia-heavy ones like Facebook, Instagram, and TikTok, cache content on local servers.


Some operate their own content delivery networks, while others use third parties like Akamai and Lumen (formerly Level3).


Many websites also use Cloudflare to defend against distributed denial-of-service attacks, which also offers caching and local content delivery.


The increasing use of caching and content delivery networks, and Teraco-owned NAPAfrica’s push for free peering in South Africa is part of the reason fast, uncapped broadband became affordable.


After conducting testing throughout the week, MyBroadband identified two general issues causing slowdowns.


  1. A provider doesn’t have sufficient backup capacity on alternative cables. This causes slowdowns when the remaining international capacity it does have saturates. This causes packet loss and degraded performance. Examples: Cloudflare, your Internet service provider.

  2. After its backup capacity floods, a CDN provider quietly redirects South African traffic to international locations, making it the ISP’s problem and exacerbating the first issue. Example: Akamai.

This is why South Africans may notice that network performance is fine outside peak periods, but slows to a crawl during the evening peak when everyone is trying to stream online video.


Some services, like Netflix and YouTube, continued to perform well throughout our testing, indicating that they are either not dependent on WACS or have sufficient backup capacity.


Considering that Google owns YouTube and Equiano, it isn’t surprising that they don’t have international capacity issues.


Responding to MyBroadband’s questions about why the WACS outage is having such an outsized impact on Internet speeds, Cloudflare explained the financial problem.


“Not all cables are created equal,” a Cloudflare spokesperson said. “Some are significantly more expensive than others.”


Cloudflare explained that transit providers don’t typically have capacity on all subsea cables — and may only purchase capacity on one or two.


“If these cables are cut, there’s no backup. Internet service providers can buy protected circuits, which means the circuit provider will failover to a different cable, but these circuits come at a significantly higher cost.”


On whether Cloudflare is taking additional steps to ensure sites remain performant despite the cut, the company provided the following statement:

“We actively monitor our performance to ensure we provide the best service to our customers. We are actively working with our partners in South Africa to ensure Cloudflare can provide a reliable and performant service while the cable operators work on restoring service.”


While it could be true that Equiano, 2Africa, EASSy, and Seacom are more expensive than WACS, that is only part of the story.


Agreements for capacity on these cables are usually locked in long-term, and companies don’t want to pay for excess capacity they won’t use except in the case of an extended cable break (or another outage).


MyBroadband has also learned that companies who bought a fibre pair from Google on Equiano are paying the equivalent of billions of rands for it.

All things considered, this is actually a steal.


Each fibre pair has 12 terabits per second capacity — almost as much as the entire WACS, which cost $650 million to build (over R12 billion in 2023). Equiano comprises twelve such fibre pairs.


Bad news about WACS repair


While the cable-laying ship Léon Thévenin has been mobilised for deep water repair, it will be around a month before WACS and SAT–3 will be fixed.


MarineTraffic reported the ship landed in Mombasa, Kenya last Sunday after a 10-day voyage from Cape Town.


The Léon Thévenin left Mombasa yesterday morning (12 August) and is returning to Cape Town. Its estimated time of arrival is 22 August 2023.


According to an industry source, the estimated time to repair for WACS is 8 September, subject to weather conditions.


View original post by Mybroadband: Post Link

23 views0 comments

Comentários


bottom of page