Title: Africa: Local IXPs and Regional Carriers Russell Southwood, Balancing Act http:www'balancingactafric
1Africa Local IXPs and Regional CarriersRussell
Southwood, Balancing Acthttp//www.balancingact-a
frica.com
2Why Im here
- We run a free weekly e-letter on telecoms,
internet and computing that goes out to 6,400
subscribers in private sector and government - Consultancy, specifically RXP study for
Connectivity Africa - Viewpoint Regional internet industry
3The Starting Point
- AfrISPAs policy statement - the Halfway
Proposition - US400m a year in foreign exchange exported from
Africa to foreign carriers to carry traffic from
one African country to another
4Crude cost comparisons
Cost of transporting local traffic is 17 times
lower than International traffic. Impact of
monopoly incumbent telcos pricing power
5The growth of local IXPs
- Arguments reduction in operating costs
reduction in end-user costs increase in local
hosting and service. - Mantra Keep local traffic local
- Currently nine IXPs .SA, .MZ, .ZW, .EG, .KE, .NG
(Lagos soon on-stream), .TZ,.UG .CD
6The impact of a local IXP
- Kenyan IXP online in 2002. Started with 4 ISPs,
now 10 are connected - Usually 200-900 millisecond delay per hop
internationally. 30-60 millisecond latency
locally - 20-25 of all traffic and growth of streaming
services
7The next step Regional Exchange Points
- Connecting up the IXPs to exchange
inter-continental traffic and creating regional
traffic to be peered internationally - Background to the debates earlier proposal for
PAVIX - Connectivity Africa study Two options - PAVIX
revisited and regional carrier/s - Regulatory issues (VSAT in SA)
8African internet exchange - the birth of regional
carriers
- Meeting at iWeek in Joburg in September.
Decision to issue a Request For Service from IXPs
from six countries through AfrISPA - Considerable interest from potential regional
carriers - Likely to be meshed VSAT network.Little or no
inter-country fibre but beginning to change
9Six countries issuing RFS
- Chosen IXPs, scale of traffic, willing. Others
to follow - 1. Johannesburg Internet Exchange JINX
Johannesburg, South Africa - 2. Kenya Internet Exchange Point KIXP Nairobi,
Kenya - 3. Mozambique Internet Exchange MOZ-IX Maputo,
Mozambique - 4. Uganda Internet Exchange Point UIXP
Kampala, Uganda - 5. Tanzania Internet Exchange Point TIXP Dar
es Salaam, Tanzania - 6. Lagos Internet Exchange Lagos, Nigeria soon
to be operational
10What marks out this approach
- Self-help - driven by the internet industry and
its needs and those of its customers - Can be encouraged by regulation but beyond basic
open door does not require it - Does not require governments to be anything but
facilitating and supportive - Relatively fast decision-making
11Support from DFID and others
- CATIA - DFID supported by others including
Canadian Govt - 1C Creating a policy environment to bring about
key changes required - 1F Support for lobbying at an international
level. Recognising shift from ITU to WTO. - Practical assistance from PCH and self-created
institutions like AFNOG
12Here comes troublePart 1
- Internet and telecoms inextricably linked
- Lack of effective competition in telecoms in
Africa a big part of the problem - Monopoly incumbent telcos keep prices up. Their
influence on SAT3 pricing. Int vs regional
half-circuits 11 vs 14 by Telkom SA
13Here comes troublePart 2
- Impact of monopoly telco ISPs on growth of the
business. Worst example SONATEL in Senegal. 80
of market. De-facto, legal monopoly. No effective
anti-competition laws. Almost no IXPs in
Francophone Africa. - Growth of VOIP. Undercuts existing incumbents but
is way forward (Telkom SA) Feared by Govt and
regulators. Regional data networks may achieve
what monopoly incumbents failed to do
14What role for the ITU and why?
- Accounting rate system is all but dead. ITU is
slow to make decisions. Byzantine representative
structure. Developing world discussions dominated
by monopoly incumbents. Has little internet
expertise. - Be humble. Work with those who are already
addressing these problems regionally.