[kictanet] KIXP reaches 1Gbps of traffic

[Michuki Mwangi]

Hi Walu,

On 7/16/11 7:07 PM, McTim wrote:
> 
>     But as Francis says, plse put it in perspective by comparing with
>     the previous years (or give us the url for the graphs we do the
>     comparisons ourselves). 
>

I have some slides i can share, i would probably want to pull them
together. However in general KIXP has been growing at an average of 100%
every year for the past 7 years and higher (150 - 300%) in the last 3 years.

Prior to Google Cache KIXP had an average of (95th percentile) of
120Mbps of local traffic.

> 
> 
>     Also, throw in the ratio between Local vs International traffic over
>     the years since 1G local traffic is good but compared to 20G
>     International traffic (CCK May2011 report?) may imply we might still
>     have some time to go.
> 
> 
> Well then your ratio is 20:1, no?
> 
> KIXP probably doesn't collect data on international traffic, as that
> doesn't go over their switch.
> 

As McTim put it, its difficult to measure. Since the traffic is going
through different carriers (Cable and satellite). Also its worth
mentioning that what we quote at KIXP is actual traffic against what is
purchased capacity.

For instance a growing number ISPs at KIXP have 1Gbps circuits. We are
measure the actual traffic passing and not the ISP Capacity to the facility.

>     In layman terms this means that we are Net IMPORTERS of content and
>     EXPORT very little content.
> 
> 
> As are most countries.
>

This is exacerbated by reimportation of content (content created here
and hosted abroad).

>     Put differently very few chaps out there are interested in our
>     content down here (coz it doesnt exist - even our very own trace
>     route to nation.co.ke <http://nation.co.ke> hosts abroad...) 
>

The hard realities that we face today are as follows;

1) We have almost no hosting facilities locally.

2) The hosting market in Europe, Asia and US is mature and very competitive

3) As an English speaking community - most of the content already exists
in one way or another. Therefore the unique and binding language theory
for growing local content such as what enabled Korea, Japan, China,
France and Brazil grow their local content will be less successful here
even with Swahili. But am glad to be proven wrong.

4) The media influence on what we consume on the Internet has left us
more consumers than contributors on the Internet.

As a result, unless by some ingenuity we are able to develop the next
killer application (i fear that even if we do its going to be hosted
abroad). For instance the last time there was an outage on MPESA
communication that came from Safaricom indicated that their servers are
not in Kenya.

For the above reasons, a change in strategy is necessary if we are to
access content at peering rates and not transit rates. IMHO i believe
Africa has the largest unexplored eyeball network in the world.
Fortunately this large eyeball network speaks the two main languages of
the world (English and French) that makes up majority of the available
Internet content.

Therefore the only way we can change the position we are in is by
aggregating our many eyeball networks (in significantly larger eyeball
networks)  in such a way that no content provider/Distributor will fail
to recognize and find ways of tapping into it.

The way to aggregate the eyeball networks is having as many of the
connect to strategic Internet Exchange Points.

Asia and Europe have strategically managed to do so and all content
providers/Distributors have already built infrastructure in these
regions in order to have a direct access to their target markets
(eyeball networks).

If we take a similar approach, and thus far its working (i.e with the
likes of Google building infrastructure into the region as they have
into Nigeria and South Africa). We can get the remaining ones to follow
suit i.e Facebook, yahoo, amazon, akamai, limelight, Microsoft, etc.

But this has to be a conscious choice for everyone. As this process is
taking place, repatriate the relevant local content - Kenya-airways,
nation, standard, open-gov, etc.

Then maybe we can start looking at different challenges at that time
such as net neutrality issues.

>     ...a good target would be to work towards 50:50 ratio of local to
>     international traffic or 20G local and 20G international...all the
>     said and done. 
> 
> 
> Did you just pull these numbers out of your hat?  Why would we want to
> keep int'l traffic at current levels?  
>

Probably what McTim is trying to say is that if you manage to make a
shift in what traffic is accessed through local peering (free locally)
vs transit (paid and reachable via other networks), you will find that
the percentage of peered traffic is likely to get to above 50%.

For instance any ISP in Kenya peering at the London Internet Exchange
Point today will peer (for free) at least 60% of their traffic and only
look to buy transit for about 40% of the traffic.

If we managed to get all the networks that are peering (that own the 60%
we need) in Kenya, that means we would only require to find someone to
carry the 40% to Europe. Of that 40% is traffic to other African
countries (since most African ISPs dont peer in Europe either). So
assuming that Intra Africa traffic is about 10% because of the intra
trading block business communications,etc. Then essentially if we
interconnect within the region we are looking for transit of less than 30%.

This is what we should be aiming for.

So for us at KIXP - Google is halfway in the bag not just yet. Until we
have them peering fully in Kenya.

We have afew more on target list. They are hard to sell to since they
buy numbers just like most businesses :). So if you are able to assist,
we would be happy to partner and strategize with you.

HTH.

Michuki.