[kictanet] State Wrong on Internet Exchange Point , is it true

Wed Dec 30 16:27:55 EAT 2009

Hi Barrack,

Speaking for self; let me try and explain some of the factors that
trigger the establishment of IXPs. (am intentionally staying clear off
the issue and instead providing background to hopefully enrich the
follow-up discussions)

1) Cost of transit - if there's always going to be a high cost of buying
transit capacity (buying access to other networks). This is often what
we call buying upstream service from an international service provider

2) More than 3 ISPs in a market - If there's more than 3 providers in
one market, it makes economic sense to bring down their costs by
exchanging traffic destined to each in a place thats within their
control and not controlled by their upstream provider. Most upstream
providers wont give you a local and international bill for traffic
exchange. If there are two providers, direct interconnects would be
sufficient.

3) Open Other opportunities - Content hosting, content delivery and
content creation will most often take root if theres a local market.
IXPs serve as a local market where folks with content can access
eyeballs at "peering" costs.  Peering is often referred to as exchange
of traffic between two networks at no fee.

If the cost of transit is lower than that of peering and there are less
than 3 providers in a market, the economic benefit of an IXP will be
negligible.

The question here is; what triggers more than one IXP in a market and
whats defines the market boundary. IMHO the market boundary is often
defined by the cost of leasing circuits. For instance the point where
the cost of leasing a circuit of 1mbps from one point to another within
a certain radius point A is often value X, beyond to radius point B the
cost is value Y hence radius point A boundary is the market boundary.

Having defined the market boundary and a prime example is Nairobi, would
be worth to start the discussion on what would trigger having more than
1 IXP.

At this stage, its worth getting to know that there are two main types
of IXPs i.e non-profit and commercial IXPs. In addition, its important
to know that ISPs' have is called peering policies. The peering policies
basically define the rules of engagement for members at the IXP. The
most common one are bilateral or multi-lateral peering policy. In
bilateral it means each member will have to negotiate with every member
at an IXP (IXP is not involved) in order to establish peering. In a
multi-lateral it means every member is subject to the same peering
policy i.e peer with everyone at the IXP. As you may have guessed it the
bilateral peering is most popular and preferred peering policy at most
IXPs. This allows for ISPs to selectively decide whom to peer with.

As a result, in any market there are small and large players. The large
players have the ability to selectively decide whom to peer with will
leads to the small players buying transit from the large players. Since
the cost of transit is often higher than that of peering, other IXPs
will emerge to cater for those who cannot easily fit into the existing
IXP. The key aspect for this new IXP will be the peering policy in place
to attract membership. The new IXP can also take the shape of commercial
or non-profit.

Please take a look at www.pch.net and see how dense IXPs are in Europe.
Some cities like Paris and London have more than 1 IXP operated by
different organizations or competing IXPs and both commercial and
non-profit.

Consequently, IXPs in different markets are never directly
interconnected and for this very same reason, competing IXPs in the same
market would not be interconnected. As a result, its common to find an
ISP participating in IXPs in different markets i.e in Nairobi and in
London and also in other competing IXPs in the same market. For an ISP
the more they can peer the lower their costs.

Countries like South Africa have more than one IXP in different markets
i.e one in Jburg, another in Grahamstown, and the recently
re-established Capetown IXP. Other countries have many players (large
and small) and with lots of content being created in those markets
having more than one IXP is often feasible each attracting a different
group of ISPs based on their peering policies.

In conclusion, the key point to remember is that an IXP will remain
viable if its able to reduce the cost of delivering packets from one
network from another. If its expensive then the cost of delivering
packets from one network to another will be same as transit costs and
therefore diminish its own value. Unless the IXP is providing access to
premium content that would otherwise be accessible via transit costs its
less likely to attract membership. Lastly, an IXP is just a Switch. The
question to have in mind is what attracts folks to come connect to your
switch where there are more than one.

I hope that gives sufficient background and info to enrich this
discussion further.

Regards,

Michuki.

Barrack Otieno wrote:
> Listers,
> 
> I was  going through an article in the East African standard with the
> above mentioned title by the CEO of  KIXP published on Sunday Dec  2009
> in the East African Standard, am  wondering is there a business case
> other than the security considerations highlighted  that justifies the
> GIXP and if we could be allowed to access it or is it classified info?,
> in addition the writer has mentioned the Strengths of KIXP however i
> would also appreciate to know the weaknesses since they might justify
> the need for GIXP bearing in mind the fact that we are just coming out
> of the woods in so far as monopolies are concerned as i try to digest
> the article in any case having gone through
> http://www.bgp4.as/internet-exchanges it is evident there are a few
> developed countries with more that one Exchange point eg South Africa ,
> the US, Sweden, Singapore, Russia and the UK, somebody educate me.
> 
> Regards
> 
> -- 
> Barrack O. Otieno
> Administrative Manager
> Afriregister Ltd (Ke)
> P.o.Box 21682
> Nairobi 00100
> Tel:
> +254721325277
> +254733206359
> +254202498789
> Riara Road, Bamboo Lane
> www.afriregister.com <http://www.afriregister.com>
> www.afriregister.co.ke <http://www.afriregister.co.ke>
> ICANN accredited registrar.
> Skype: barrack.otieno
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> kictanet mailing list
> kictanet at lists.kictanet.or.ke
> http://lists.kictanet.or.ke/mailman/listinfo/kictanet
> 
> This message was sent to: michuki at swiftkenya.com
> Unsubscribe or change your options at http://lists.kictanet.or.ke/mailman/options/kictanet/michuki%40swiftkenya.com