By now I’m sure you’re aware of the multiple undersea outages affecting
WACS and SAT-3. In addition to this, ACE was also damaged during July,
resulting in higher latencies on that path. [Mybroadband]
mybroadband.co.za/news/fibre/503568-bad-news-about-break-in-undersea-cables.html
I want to provide a comprehensive view of the above cable breaks; and
how they affect the local internet ecosystem and the potential impact
these outages will be over the next few weeks.
From a cable point of view, the Léon Thévenin (the ship that can fix
this fault) is only expected to reach the site of the break in the
2nd-3rd week of September, and please keep in mind, repairs are complex
in this region and may take additional time. So we can safely assume
that these cable systems will only return to service later in September
or early October. ACE’s repairs have been delayed until late October due
to the location of the fault, and adverse weather conditions in the area
hampering safe access to the location of the break.
Equiano is in production for some networks (two of our upstream partners
have lit Equiano capacity). But the cable system is not fully in
production yet for everyone (despite all the media fanfare), with
hardware vendor delays or terrestrial capacity delays between Portugal
and London, leading to only a fraction of the total cable system
capacity being online. This is important as we explore the local
internet landscape.
Latencies to some international destinations will be generally higher
during this time.
Locally, South Africa has one of the richest peering ecosystems in the
world. With over 550 networks present at NAP JHB alone. This means that
many large data entry and exit points are locally peered, including big
names like Netflix, Microsoft, Amazon, Facebook, Cloudflare, and other
clouds and content providers. Some of these networks are part of those
either waiting for additional capacity (eg. hardware vendor delays for
Equiano, or 2Africa) and/or struggling with sufficient capacity on
re-routed traffic, with so many active cable systems currently affected
by faults/breaks.
Peering exchanges play a pivotal role in the South African network
ecosystem. However, some content providers may currently lack the
capacity they require to adequately transport traffic as they normally
do. We’ve already seen providers such as Cloudflare (one of the largest
content networks on the planet) re-route traffic around these
constraints www.cloudflarestatus.com/, as best as they can – but
the delays in loading Mybroadband show that this is not ideal. These
changes are, however, more fluid than static. Changes are constantly
happening based on demand and other network factors. De-peering these
networks is also not an option due to the need for access to their local
resources (eg. AWS and Azure local instances). As such, certain traffic
may be less than ideal as these peered networks grapple with these
constraints.
We have taken measures to ensure we proactively monitor these networks,
and make our best efforts to work around any potential issues. Bear in
mind these are incredibly complex network structures (Cloudflare, for
example, peer with us at 4 DCs on 10 different interfaces nationally)
and managing changes is a complex task.
–
Pretoria Wisp Team
www.pretoria-wisp.co.za