[Afpif] Transit vs. peering Focust for Africa by 2021

[Mark Tinka]

On 10/Aug/15 08:11, Meoli Kashorda wrote:
> On the 60% local content, I am the one who misread the note from Michuki (have not read the actual report). It suggests we shall achieve 60% not 80% by 2021.


Well, the TG report suggests the majority of Africa's peered traffic
will rise to 60% by 2021. It does not say where this traffic will be
exchanged (inside or outside Africa), nor does it say how much of that
will affect the amount of content being served locally in Africa.

Michuki's goal for Africa is to have 80% of content served from within
Africa by 2020.


>
> I have no data to challenge their forecast. But I do have some anecdotal data on Kenya -  global CDNs are saving Kenyan operators about 50% of international circuits capacity! And that Google and Akamai caches are serving about 35-40% of traffic from Kenya. So we have probably achieved 40% local traffic? 


I'm not sure it is that straight forward.

Remember that folk like Google, Netflix, Facebook, Microsoft, e.t.c.
generate quite a lot of content. Then folk like Akamai, Limelight,
Fastly, e.t.c., all distribute it. Some of the content players (like
Google, Netflix and others) are also distributing their own content,
either fully by themselves or partially between themselves and select
pure-play CDN's like Akamai, Limelight and others.

Also, remember that not every content owner on the Internet pays Akamai
to distribute their stuff. In fact, a good portion of content owners do
not use CDN's for the distribution of their content, perhaps because
their target audience is in a specific country where latency is not a
problem and CDN's would only eat into their already-tight margins.

Also, some of the CDN's may serve different content in one region from
another depending on the kinds of deals they make with their customers,
the licensing regime the content they are serving falls under, the
amount of demand expected for that content from a specific region,
infrastructure resources available to that CDN in a specific region, e.t.c.

So it's quite tricky when you only have one or two CDN's on the
continent, who perhaps have only a certain % of market share, out of
which they can only serve a portion of that content within Africa.

Let's also remember that Google's GGC is primarily Youtube videos (so
all the other Google applications and services aren't included). Also,
not all videos will be played from a local GGC. How Google decide where
the actual video comes from and when the local GGC will cache it is
something not within your control, as a GGC host. So even in this
scenario, it's unlikely 100% of all Youtube videos are being served
locally, although the chance of that increases with a high volume of
hits for the same video clip. 

But remember that there are a bunch of other networks in Kenya whose
data you may not have. So when you take a global view of that (and
Africa in general), it is quite difficult to see how we're already at
40% of global content being served locally, especially considering that
as a whole, Google, Facebook and Netflix account for about 45% of total
Internet traffic, a metric, I'm certain, is mostly based around the
developed world.


>
> Maybe someone has already done a country by country study? We at KENET would appreciate a research grant to conduct EA study - anybody generous? 


I agree that a study in this area would be very useful indeed.

Mark.