Ethiopia Peering Day - Welcome and Opening Remarks
5:33AM Jan 24, 2025
Speakers:
Keywords:
Ethiopian Peering Day
Internet Society
Internet exchange point
local content
network interconnection
transit vs. peering
regulatory framework
local traffic
global infrastructure
data centers
fiber connectivity
subsea cables
regional peering
engineering expertise
cost reduction.
Cultural day to day.
So we'll we'll have to deal with that
anyway. So. This will not be bored,
by the way, for the locals. Is it a wedding? Or is that cultural day? Or is it A Day National
Better, 121212121212121
So sorry for the delay. We are trying to find a solution,
because we there is no way we can have this our meeting with this noise, so we are trying to find a solution as quickly as possible so that we can proceed. Sorry about that. You
but I think most importantly is the discussions that we need
to have. So I will invite you to come forward this way, and then we can do the photo group. Then we can proceed.
Okay, so in change of program, we are going to have the tea break now.
Now as they are moving equipment, and then we can look at photo after the photo tea break, and then we can proceed after the break. Thank you very much. So photographer, you can lead us to what you want. I hope you had a good tea break, at least we we we had to find a good use of time as we were setting up. And really thank you for the team that has been able to pull it off in few minutes and be able to have a room that we can use this morning. So my name is Gilan Hira mugawa. I'm from the Internet Society, and I welcome you for the Ethiopian peering day. It's the first in its kind, and we look forward to have good discussions today. We have a long agenda. I hope we will be able to cover most of the points, but the most important thing is to have the discussion going, even if we have some slides, those are supporting elements, supporting documents, supporting topics. The most important thing is to have the conversation going. So take advantage of the lunch break. Take advantage of the tea break. We already had one, and at the end of the day we will have also a tea tea breaks, slash social event, talk to someone, get to know them, ask questions. I am aware that some of us may not know exactly.
Feel free to note down and ask any question
in English or in local language, Dawit has offered to act as our translator for today, and we really are grateful for that. So without further ado, let me welcome Dawit for a few remarks, and then we can start our first session. We can give a round of applause to Dawit. You.
Thank You. Gila damard, Chu We apologize for the conveniences that the change of room has created. Unfortunately, we didn't know that there was a lively event next to us, so we didn't plan it so, but I think it's not a bad room, and I hope that we'll be able to have our meeting here without any issue from now on. So as Jean said, it's the first peering day that we are having here. We had workshop on internet, exchange, point best practices. But a long time ago, I think 2015 at the African Union, but at that time, there wasn't even liberalization, even the talk of liberalization. So I don't think it was that useful. It was for the whole east Africa region. And we just wanted to have to give some awareness. But it was not timely. But we had small discussion, a small group of people who discussed about internet exchange point in 2020 as the first internet development conference. And that initiated the idea of having an internet exchange point. And we are very happy that some of you took it to the next level, and now we have, even if it is beginning but an IX here in Ethiopia. So I would like to congratulate those who have taken that responsibility and created something, and we have to build on that. So the objective of the Appearing Day, as Gila indicated, is to create awareness. Because I don't think all of you have the knowledge of Arlington exchange point, because it's very new for Ethiopia, and also creates the environment so that IX P can strive. Because an IX piece like an airport, it's very nice to have an airport, but if there isn't any plane landing, then it's not an airport. It's just a building. So we want the internet exchange point to be lively, to have many traffic exchanges so that it benefits the country, because right now we are all losing I think someone checked and there is still some traffic that goes from Ethiopia to Marseille and to comes back to Addis, and it's not free. I think it's free. It's not we are paying it. We have to stop it. Nobody is winning that. We are all losers. So let's try to understand, first of all, what's the current situation, how we can improve it, and how we can all contribute. There won't be any loser. If we have an internet exchange point that is thriving, everybody will be winning. And the more we wait, in fact, the more we are losing. Because, you know, the whole world will not wait for us. Adi, Saba bah can be Ethiopia in general, can be a hub for the region, but currently we are not, and we can do better. So let's have a lively discussion today. There's no silly question. If you don't understand any something, please do ask. It's very nice. We didn't want it to be a big group. It's nice that it is a small group, so that you can have discussions. It is much, much informal than the last three days. You can, you know, you can feel free to ask questions, to have comments, to have discussions, whether it is here or during the break the lunch, so that our objective is that by the end of the day, we have an army of people who will work for development of internet exchange points here now we have a small group that's been working for the last few years, lots of challenges sometimes, but the more we are, more results we'll have. I would like to thank those who have contributed to making this event happened. First of all, wingu, or local host and sponsor for this event. Of course, my colleagues at Internet Society, led by Gila and whole team that is working on internet exchange points every single day around the world, and also the events team, everyone the chapter, everyone who has been, you know, supporting us, whether it is by being moderator, speaker, so many of You who have made this event happen, viewers in America. Uh, one, no, Alama, chin, Git, exchange points appearing through the no Mal, Baba, tomorrow, no Rosie. Government, would you know so
that any unofficial interpreter? Kill color level. Thank you very much. You
Thank you, David, I have to confess that
with many travels in Ethiopia, I haven't put I haven't been able to learn any Amharic language, so any word, but I promise that I will try as much as possible to to learn at least 111, or two words. So I will want now to welcome with the first speaker. He's robbel Gita BA and without further ado, I will, he will talk a little bit about himself. We don't want to do the long introductions. We want to redeem time. And I would run also to ask if it is possible that if you see that a speaker has covered an aspect in the interest of redeeming time, you can pass on, and then we can proceed. So can we have Robert's presentation on the screen please, as he's talking about himself, round
of applause. Thank you. You.
I have been asked to make my presentation quick, so I'll try to
try to go quickly. Good morning, guys. My name is Robel. I'm a torque engineer by trade, so I have around 17 years of experience here in Ethiopia, in Europe and in the US, focusing mostly on global backbone and internet age, particularly on the CDN network operation. I worked at Amazon and meta on the global backbone, implementing, designing and operating their global infrastructure, and also in Ethiopia. Before my relocation, I worked at ZTE, deploying the then new network called the engine, which I believe is still here. So the goals of this talk is to give you guys a little bit about the basic of what the internet is, depending on who you ask different people can give you what internet is for them, but we are here engineers, and as an engineers, we need to go a little bit technical, so I'll try to cover that, and then probably provide some insight on the current interconnections status here in Ethiopia, and provide also possible opportunities for future improvements. Okay, so the internet for most people looks like this. They would have a PC or mobile, and they will
try to connect to a website.
It's It's oblivious for them where that is, but some picture, in this case, like a funny cut picture. So this is what they see. This is what people would say internet. They might have some notion of that, that website is somewhere outside. In the Ethiopian case, it's outside the country. So when something happens, is like the internet is done, or the, you know, you would hear in the news that the fiber cable is cut. You know, there are some reasons right for us, for network engineers. This is what we see. So the PC would be usually called a client device, and then the picture hosted somewhere is called the server. And then in between this, their network devices, these are fancy name is router, but they are specialized computers to do only route traffic from point A to point B, in this case, from the client to the server, and a group of this devices are called a network. So in the internet, when we say network, just group of devices that move traffic
from one place to another,
and then a network as as often is the case, won't have the all the destinations, so it have to go to another network to provide you or
to deliver the necessary
website traffic. So then a group of this networks combined together to collaborate to form what we call internet, which is a, you know, a fancy name, of inter networking, which means one network would work with Another to deliver the traffic. In reality, this network interaction happens over a very long distance, usually 1000s
and 1000s of miles.
So usually networks would only care about their specific locations or this specific focus area. They would build the necessary infrastructure, and then they would hand off the traffic to the next network at a certain location. And usually it's very expensive to build this, this interconnections infrastructure, which, you know, which, which I would call a fiber. And there are only few handful of companies that have a global wide fiber infrastructure. So what usually happens is that there are few facilities that these networks can exist or want to exist, and. And then if they want to hand over to another network, they would go to these facilities, which we call inter interconnection hubs, or but what we call now IX, P, which is the internet exchange points. So these are focused on on kind of playing the role of a meeting hub or meeting location. So a network a would say, hey, where does network B exist? It's in this location. Let me go to this facility, and then I would interconnect with this guys. So this interconnection points are globally available. Even in at this there is recently, I heard one pilot as a test. There are about 1200 Internet exchange in the world, according to peering DB data, as recently as yesterday, most of this exchange or interconnection hubs are locally significant, which means they focus on a specific location in a city or in a country. They don't have that much significance out of that location or a country. Some exchange, however, our regional significance, which means that people from the region, or, sorry, network from the region, would go to this location to interconnect. A very good example in Africa is not Africa, which is, I believe, the first exchange points, or one of the first in in South Africa, and is the biggest, I believe, where the whole southern Africa networks would go in to peering or interconnect. However, globally, there are further handful, or a couple of dozens of interconnection hubs that are globally significant, which means if they go down a significant part of the internet, the global Internet, would be affected. They are very big. They have very big infrastructure available. There are only few of them as compared to the rest of the world. For for example, the London exchange, the Marseille exchange, the Marse exchange, actually serves Africa, Middle East and South Asia. So it's significant. The London exchange also serves the whole Europe and Africa to some extent. And then the Northern Virginia exchange probably first all of North America and Europe as well. So this exchange interconnect with each other as well. They are very well connected and well run and people network intentionally go there to find relatively good internet interconnection there. So when we look at the East Africa
or igadrizin,
is a little bit
not, not as it should
be. There are 12 exchange as I counted from peering dB, seven, I believe, are in Kenya, and then one in each of the rest of the countries. So there are some opportunities here to actually do more of this. And then when we talk about the interconnection types, there are two types as a as a larger umbrella, I'm not, I'm not going to go in depth of what the different types of the transit and the different type of the peering relationship is. There are subtle differences even within each but in general, to summarize, transit is where you pay the network and network to reach the rest of the internet or anywhere effectively. So in this case, network B will say, you know, you need to pay me some money, and then I'll give you access anywhere in the internet. If you want to go anywhere, give you access, but you have to pay this amount of money, and that is a relationship called transit, or provider, and a customer relationship, however, appearing is when two networks identify that there is mutual benefit to exchange traffic between them for free, but only to the destinations for for their own network. So in this case, network B will say, let's, let's provide access to each other. A, which means network a can only access network B, and network B can only access network a through this peering, and you don't have to pay anything. So that is a relationship called peering. There are print sub types of the peering, but in general, this is how it works. It doesn't give access to the rest of the internet, but to a specific network, and it's usually settlement free. And there are different scenarios where it makes sense to do peering versus transit. As you can imagine, transit relationship requires a lot of money being paid by the customer. So it might make sense to go and do peering. However, it's not always the case. Even if there are peering relationships available. You might want to do transit just because of the availability of the of the whole internet, right? So the transit providers usually would guarantee that you can reach anywhere in the internet. So that is a very compelling reason why, you know sometimes, sometimes it could be a performance, a peering might perform, sometimes slightly lower than the transit just because the transit providers would have a lot of global infrastructure in place, so it might make sense. So it's always a balance. The other aspect is that even if one network does, peering is not always cost free. There are always costs associated these facilities where you go and peer usually would have a membership fee or some kind of space and power fee. And of course, you know, you have to actually get to the interconnection facility, which means that your data center traffic might be, your data center facility might be somewhere and the interconnection facility might be somewhere else. So you have to actually, you know, buy a lease line or fiber to reach there. So there are some costs associated. But in general, relatively speaking, the cost is significantly lower than the transit, and we have to balance when to do that. So when we look at the
ET open interconnection status
versus ethos Telecom, which is the largest network in the country, it doesn't appear to do any peering according to peering DB in some of the BGP routing tables, ethos telecom network is solely feasible via transit networks, so which means they go somewhere and buy the transit stations. Usually, I suspect in Djibouti and Sudha tel in Sudan, some cable, I believe, in Somalia or in Djibouti and China Mobile and China Telecom in either Djibouti. So these are public, available data. I you know, I'm looking from the outside, maybe things might be different from the ethos telecom side of things. But from what we see, they don't do any peering, and they don't have any presence even in the local Internet Exchange, which I believe was recently formed.
So the next one is sapphire com, etopia,
the second largest network. So when, we look at the peering DB and some of the BGP routing tables, at first, they do look like that. They don't do any peering. They don't have any presence in any of the Internet exchange although I think from what I got some of the information this morning after I prepared that slide. They probably have a presence in the local internet exchange, but they probably might not have updated the appearing dB. So at first they might not look like that. However, Safaricom is a subsidiary of Safaricom. Ethiopia is a subsidiary of Safaricom limited, which, which in Kenya, has a versatile peering presence. So my suspicion is that, you know, because it's the same company, or relatively, they have a good relationship between the same sister companies. They do local peering between themselves. So is it's not, you know, it's not, it's not. It's not going to be fair to just blame them that they don't appear. I suspect that they do appearing. However, they also buy transit from Djibouti just because of the proximity to Ethiopia and. Kenya adds around 2000 kilometer to the international traffic paths, while Djibouti is around seven to 800 kilometers. So it makes sense to buy transit there. So so a case study which I tried yesterday, and which Dr Darius alluded earlier, is I wanted to see if I send a traffic between ethos, telecom and Safaricom, where would it actually land, and how the traffic pass might look like. So this is the trace route, which, which any, anyone of you can do from your computer, locally from from this Wi Fi hotspot, which uses CD telecom. And then the destination is one of the IP range from the Safaricom, Ethiopia, IP address allocation. So you can see here, the blue line is is low latency because it's local within the ETL telecom infrastructure. However, the next slide, which is the yellow one, there is around 18 millisecond jump, and that is the Djibouti Telecom, or the song cable, sorry, the SOM cable Pass, which is the SWAMI Alia cable, I believe so. The latency jump is, I assume, is from add this to wherever the Somme Alia cable is, possibly in Djibouti. And then, interestingly, the next line you see the red one is an international tier one provider, cogent, so Somalia cable hands off the traffic to cogent, possibly in East Africa, either in Djibouti or in in Hargeisa. But then cogen doesn't have any destinations or have info to reach the Safaricom Ethiopia traffic locally, so it would send it to Marseille, and then it would send it to another cogent router in Marseille, where that router have an info where to send it to reach Safaricom Telecom, and then it was send it to Kenya, cogent, and then cogent, I believe, either appears, or there is a transit Relationship to Safaricom limited in Kenya, and then Safaricom limited, which take the rest of the traffic back to Ethiopia. So it adds an unnecessary high latency of two milliseconds, where both the source and the destinations are in Addis, which, in theory, should be around three to four milliseconds. And this is the reason why peering is important, because sometimes you know, without an awareness, or with some reason, you do take traffic unnecessarily long paths when you have the option to actually send us locally, providing high quality, good internet service to your customers. So there is some opportunities here. One important point I want to mention is that maybe doing a peering in this case, might not make sense if you don't have any traffic to go between ethos, telecom and Safaricom. Maybe right now it might not. It might be very minimal traffic. So we don't want to, you know, have additional cost of getting into an Internet exchange and doing peering. But the principle is there that doing appearing locally makes sense for this kind of reasons. So
based on my personal view,
I can see a few challenge at the moment. One of the main challenges that there are relatively two big networks, Safaricom and ethos telecom. Maybe Itea telecom is the biggest one. And then there are very few networks that can actually economically make sense to do peering, either because they don't have traffic or they are exclusively home to either ethos, telecom or Safaricom. So this is the challenge, even if you have an internet exchange, if the only access to the internet is via Ito, telecom and Safaricom, it doesn't make sense to do peering unless you have out. Their network pass to provide alternatives. So this is a challenge I would see, although there is an IX be started here, and I hope that we can see more networks to start to recognize that they can do peering. For example, banks can do free peering between them, and then they can have lower cost and more quality. So this, these are the challenge I see from the network size perspective. The other main one, which I believe we mentioned a few items in the last session, yesterday or the day before, is the local content. So there are almost zero local content, I would say. So the whole Ethiopia internet traffic is heavily inbound. I would say, significantly inbound the traffic, which means that the the traffic load is coming from outside, and then the users here in the country are the consumer. So there is no source of the traffic here, and this, this ties into a lot of other issues, of, does peering make sense? If there is no local content, if there is no if there is no traffic locally, then you know, how can we create data center? Because there is no, you know, the commercial sense doesn't that is not there. The other slightly lower challenge, I would say, is the lack of engineering expertise and context to do peering. So I focused a little bit about local peering here, but there is the option of regional peering. Our region has at least few exchange where there are very good peering opportunities. So it might make more sense to actually and cheaper to actually buy the fiber infrastructure, to go to this exchange regionally and peer freely to this global internet companies. So that kind of option requires some engineering context, and I do recognize that there might be some lack there. And of course, when we talk about this, it's relatively high cost at this point of time, the fiber infrastructure, the lease lines might might make make less sense at the moment, but in the future, it might be good. And then the last one is the regulatory framework. Is a is at the beginning, so it's very early stage. So a lot of items needs to be encouraged and empowered, for the local content, or for the peering, or for this networks to have, like, a more fertile ground to improve. So I would say that probably that area is the most challenging. Because, you know, from my experience, technical stuff is easy to do. We can get a group of engineers into the room, and we can probably find a way, however, regulatory and policy frameworks tend to always, you know, creates unique challenge. So some opportunities
from my personal experience, I
would say Ethiopian network can improve, or there are some opportunities to do better. Is our region in East Africa, particularly, are seeing an IP transit price erosion, so which means that the price to do peering is significantly going higher or significantly larger than the price to buy bandwidth from the actual subsea cables and do regional and global clearing. For example, we will have a bunch of subsea cables globally. I'm sorry, in East Africa, in the coming two or three years, they are going to land with a massive amount of bandwidth capacity. So, you know, buying a bandwidth from one of this subsea providers, and do settlement free peering somewhere in Europe or in Africa will be cheaper than actually buying transit in Djibouti. So I hope that this is going to go down and down, and then, you know, the networks are going to. Recognize that there is this opportunity that they can make use of that the other one is the local content. And this one is a tricky one, but still, I put it as an opportunity, because there are local data centers and local interconnection facilities popping up, however, that also would encourage other companies to form a network and to create their local entities, and with this subsea cables available, I would hope that the global Internet companies can have the incentive to actually come to Ethiopia and do local content locally. So, as Mr. Doyt mentioned, we have the potential to actually become a hub, for example, relatively speaking, utility electric power is cheaper in Ethiopia than in the rest of Africa, where the data center main requirement is power, especially in the AI age. And of course, the fiber connectivity within the country is relatively better. And then once you have the international sub C cable, the final piece will be there to actually provide the incentive for others to do the content hosting and the content generation locally in Ethiopia. So I hope that this tool is, you know, is, will, will be, will be an incentive, and will be the, you know, the trigger to do more in the next few years. So, in conclusion, I want people to take into, into memory that the internet is not just a service, or the internet is not just a website, it's actually much more complex, and it involves a complex collaboration between not networks, and that collaboration is what makes the internet tick.