This is a very interesting point and also an opportunity! There aren't that many underseas cable to and from Africa either am I right? So at least in southern parts I'd say that congestion over the cables will become an issue. Something that wouldn't be the case between say Europe and the US.
but clearly the nortern parts of the continent will need more coverage.
I wonder what the continental fibre infrastructure looks like. Does most of the traffic go via a handful of choke points on the way out of Africa, and would traffic from Kenya to Nigeria (for example) go via somewhere like South Africa, despite the extra length of distance involved?
Uganda is home to a large regional hub for UN logistics, a state-of-the-art CDC infectious disease laboratory, and similar outposts--presumably due to Uganda's relative stability over the past few decades and strategic proximity to several less-stable neighbors.
There's distinct hubs in West Africa, Southern Africa, and East Africa, but a lack of intercontinental terrestrial connections (there's almost a link across the Sahara, nothing live directly connecting East and West Africa).
From what I can tell, Chad, DRC, and CAR are all not really politically stable, so I don't know how realistic cables running across those countries is.
But isn't the issue wider than just not having datacenters? As I understand it the issue is multi-faced and composed of:
1. Reliable power
2. Reliable connectivity.
3. Lack of trust from users (prefer outside providers)
4. Expensive (no economy of scale)
5. Lack of inter-country connectivity.
All of this makes it hard to actually build a profitable data center provider.
I am sure I'm wrong though, I'd be interested in your opinion on this.
The article states that "collocation in African Data Centers costing more than double the cost in the US and Europe."
So the issue is not that "more data centers are needed urgently" but rather than the question of why the price situation is the way it is.
The main expenses of a data center (just colocation, not renting VMs) are the capital costs of building/renting the facility (including digging some land to pull fiber etc) and paying staff. It's not obvious to me why in Africa these costs would be higher than in Europe, but apparently there's some reason.
I don't see reliable power as a big differentiator since even in the first world your data center anyway needs backup generators; it's just that they would get used much more often.
I don't see expensive bandwidth as a big differentiator since (as far as I understand) the big bottleneck is the undersea cables, and hosting your data on the "wrong side" of these cables should be more expensive - ISPs (at least elsewhere) prefer to cache/re-host/CDN popular data locally as some extra servers in ISPs datacenter turn out to be cheaper than hardware needed for all the extra traffic.
Power. Reliable power is hard to come by and is more expensive. Power availability and cost is a significant factor in data center location in the states.
How hot do you think DTLA gets with regularity? People don't consider LA to have among the best weather in the world because it's hot, they do so because it's temperate. The record temperatures for the hottest months are well up there, but the averages are pretty comfortable. Obviously data centers have different senses of "comfortable temperatures" than humans do, but the challenge is proportionally larger when temperatures are much higher.
That being said, I'm not sure the GP complaint is that significant. Africa is a BIG continent, and data centers locations would obviously be chosen in areas that are relatively more temperate.
Humidity is a bigger factor than temperature. 90 degrees on the west coast is nice because it's not humid, 90 degrees in Southeast USA is unbearable because it's 90% humidity.
>However, temperatures across the region often exceed 90 °F (32 °C) during the summer. This happens when an atmospheric high-pressure area becomes dominant over the Great Basin of Nevada and Utah (a frequent occurrence), and the resulting offshore flow of the atmospheric air mass shuts off the normal coastal sea breeze in Los Angeles.
>It's not obvious to me why in Africa these costs would be higher than in Europe, but apparently there's some reason.
Physical security and security of investment (via security of government, infrastructure, judicial systems, pipeline of educated and skilled workers) is worth a lot.
Logistics infrastructure is one major cost factor. To get a data center up and running requires moving large quantities of heavy, high-value, and relatively fragile equipment through ocean ports and over roads and/or railroads. In regions of the world with less quality infrastructure, it's significantly more expensive to transport that stuff.
Container ships have internet via satellite today but they use it only when absolutely necessary because it's very slow and expensive. If and when the cost and bandwidth of satellite internet fall to within parameters that make it competitive with fiber optic cable, perhaps.
Any extra African data centers would be located in the major coastal metropolises, which are already connected with the major undersea cables. Sure, it's logistically difficult to ship stuff to inland rural Africa, but delivering equipment from Shenzhen to Lagos or Mombasa is not much more expensive as delivering it to London or LA.
Sure, but building a data center in a metropolis can be a lot more expensive than building it in a rural area because of other factors like higher land values, taxes, energy costs, etc.
The ocean freight for Shenzhen to Lagos probably costs 2-3x more than Shenzhen to Los Angeles, but ocean freight alone is still a tiny fraction of the total landed cost of the data center.
Poor countries have less capital. Thus the cost of money is high. For example in Uganda a bank loan costs 20% interest. At such rates only the most effective investments are viable.
Many countries in Africa have such large room for improvement through investment. Data centers are not the biggest need. For example in Uganda they grow avocado's, but lack the cold chain distribution network to sell into the international market.
Thus you have avocado's selling for peanuts at local markets instead of earning big money selling to the west.
Building up a nation's capital stock takes time and long term stability. Thankfully Africa has the stability but time is slow.
in Uganda they grow avocado's, but lack the cold chain distribution network to sell into the international market.
Let's do 5 Whys with this? Basically, that means the industry can't reliably truck or otherwise ship refrigerated produce to the ports of Mombasa and Dar es Salaam. Why? I'm going to guess that the road infrastructure is too poor to support refrigerated trucks. Why?
I think it is worth noting that historically, land-locked countries tend to be less well-off than its counter parts. Examples such as Switzerland & Austria are an abnormality due to a variety of unique factors that have only occurred in Europe so far.
Overall, the rule of law, regulatory environment and open market of Uganda and its neighbors are not at the level needed to to maximize its potential. This, coupled with the cost of borrowing (think interest rate), asset risk (E.g, forfeiture, robbery, piracy) and a few key negative economic indicators (Major balance of trade, current account & gov budget deficit) can put a dent in a country's mid-long term success.
Even if we consider earlier history as irrelevant, it's worth looking at the differences of industrialization and wealth in coastal China vs inland China, as many of the industries driving Chinese economy heavily rely on low cost sea import and export.
- road network unreliable
- theft of property along road
- electricity unreliable
- workers may or may not show up
- bribes for transport, bribes for regulatory checks
- unreliable access to credit
- unreliable access to mechanics for truck, equipment
- unreliable commercial and regulatory structures
- unstable political climate
- unreliable export/import laws
- unreliable transport partners
- unreliable internet
- unreliable water source
- excessive weather risk
- unpredictable demand
- unreliable payments from customers
- payments delayed by snafus in credit/financial system
- lack of appropriate insurance products
etc. etc. etc.
'Lack of data centres' in Africa is more likely a symptom, not a cause of anything.
Having access to high quality, well run data centres is actually a positive thing for Africa, since that is just a little higher up on the value chain than maybe some places can handle.
Once other things are in place, Africa can make all the data centres they need, but I'd argue that this is just an optimization. I don't see a hugely material issue with not having data locally.
I don't see reliable power as a big differentiator since even in the first world your data center anyway needs backup generators; it's just that they would get used much more often.
#2 and #3 are lower for a generator you run 8 hours a year versus 800 hours a year. And you may want added redundancy for your generator capacity if you plan to run off generator power frequently, so #1 can be higher too.
(I broke out maintenance costs from operating costs since there's fixed maintenance costs even if you never run the generator, and there's variable maintenance costs based on the number of hours you run it)
It's been a long time since I've rented coloc space, but at the time, the cost of the power drops to the cage cost more than the space itself. Conditioned reliable power is expensive.
Lighterweight virtualization can make it affordable to have many more, smaller, data centers, allowing a presence in Africa. If you write a Cloudflare Worker [1] it can be executed in any of our 11 PoPs there [2].
Too difficult to run a business there. ZA Has become a basket case banana republic and the rand keeps falling. Full on GTFO by anyone who wants a future and can afford to leave.
If you host your data center in Africa, but use cloud services in USA or Europe, the latency is going to kill you and your app will even be slower. So you must be willing to give up using AWS, Azure, GCP, etc. If not, then the title of the talk should be AWS, Azure, GCP should build data centers in Africa, but then it doesn't make sense economically. The article mentioned how Nigerians spend about $60 million a year on outside data centers. VCs won't even invest in a software only startup that will make $60mil a year. Let alone a business like a datacenter that needs physical location, properties, personnel, etc
You are ignoring that there are services in Africa focused on the African market. These services could greatly benefit from hosting in a data center and serving their customers from the same continent without having to hop to the EU and back. Granted, depending on where you are (equatorial coast vs. southern) then you may end up hopping anyways because there isn't much (any?) direct usable fiber across the continent.
What's EU latency in Africa? I live in Kazakhstan and EU latency here 80-150 ms. It's pretty fast. I'm actually using Amsterdam VPN almost all the time, and I don't notice any slowdowns.
How much datacenter could one fit into a satellite weighting 30 tons? What if there was a Starlink style contellation of datacenters? Power would be a big issue, of course. If these were in low Earth orbit, latency would be low, and if they had high speed point-to-point communications, virtual servers could be paused, serialized, then sent to the next satellite to go overhead in a specific region.
Granted, it would be way too expensive at the present time, even with reusable Falcon rockets.
True, but an airliner full of fuel might also cause an explosion and fire. Those are flying over highly populated cities, and we trust the industry to retain control of them and set them down safely.
The kinetic energy of a 30 metric ton datacenter moving at 9.8 km/s is 1,440,600,000,000 joules. That's a bit more energy than 344 tons of TNT and a bit more than 12,000 gallons of gasoline equivalent. That's actually much less than the energy embodied in a 747 with full tanks. (If I've calculated correctly.)
While working briefly for a telecom I was involved in a project that was installing point to point microwave transmitters in Africa instead of fiber optic cable because the locals kept cutting the cables hoping to steal the non-existent copper.
Without the logistical infrastructure in place I don't know how you hope to drop a data center in Africa and hope to have costs be anywhere near what they are in the developed world.
61 comments
[ 2.5 ms ] story [ 123 ms ] threadhttps://aws.amazon.com/local/africa/
Azure also has pair of DCs in South Africa
https://azure.microsoft.com/en-gb/global-infrastructure/sout...
but clearly the nortern parts of the continent will need more coverage.
I wonder what the continental fibre infrastructure looks like. Does most of the traffic go via a handful of choke points on the way out of Africa, and would traffic from Kenya to Nigeria (for example) go via somewhere like South Africa, despite the extra length of distance involved?
There's distinct hubs in West Africa, Southern Africa, and East Africa, but a lack of intercontinental terrestrial connections (there's almost a link across the Sahara, nothing live directly connecting East and West Africa).
From what I can tell, Chad, DRC, and CAR are all not really politically stable, so I don't know how realistic cables running across those countries is.
1. Reliable power 2. Reliable connectivity. 3. Lack of trust from users (prefer outside providers) 4. Expensive (no economy of scale) 5. Lack of inter-country connectivity.
All of this makes it hard to actually build a profitable data center provider.
I am sure I'm wrong though, I'd be interested in your opinion on this.
So the issue is not that "more data centers are needed urgently" but rather than the question of why the price situation is the way it is.
The main expenses of a data center (just colocation, not renting VMs) are the capital costs of building/renting the facility (including digging some land to pull fiber etc) and paying staff. It's not obvious to me why in Africa these costs would be higher than in Europe, but apparently there's some reason.
I don't see reliable power as a big differentiator since even in the first world your data center anyway needs backup generators; it's just that they would get used much more often.
I don't see expensive bandwidth as a big differentiator since (as far as I understand) the big bottleneck is the undersea cables, and hosting your data on the "wrong side" of these cables should be more expensive - ISPs (at least elsewhere) prefer to cache/re-host/CDN popular data locally as some extra servers in ISPs datacenter turn out to be cheaper than hardware needed for all the extra traffic.
One reason cell phones are so big in the poor world is that you can't leave cables lying around there.
That being said, I'm not sure the GP complaint is that significant. Africa is a BIG continent, and data centers locations would obviously be chosen in areas that are relatively more temperate.
https://www.weather.gov/safety/heat-index
>However, temperatures across the region often exceed 90 °F (32 °C) during the summer. This happens when an atmospheric high-pressure area becomes dominant over the Great Basin of Nevada and Utah (a frequent occurrence), and the resulting offshore flow of the atmospheric air mass shuts off the normal coastal sea breeze in Los Angeles.
Physical security and security of investment (via security of government, infrastructure, judicial systems, pipeline of educated and skilled workers) is worth a lot.
EDIT: PDF map of ports in Africa
https://www.pwc.co.za/en/assets/pdf/large-map.pdf
Microsoft and others are also testing miniature undersea data centers, and using seawater for cooling could help address air conditioning costs: https://news.microsoft.com/features/under-the-sea-microsoft-...
So maybe not with Starlink, but a bit afterwards?
The ocean freight for Shenzhen to Lagos probably costs 2-3x more than Shenzhen to Los Angeles, but ocean freight alone is still a tiny fraction of the total landed cost of the data center.
Poor countries have less capital. Thus the cost of money is high. For example in Uganda a bank loan costs 20% interest. At such rates only the most effective investments are viable.
Many countries in Africa have such large room for improvement through investment. Data centers are not the biggest need. For example in Uganda they grow avocado's, but lack the cold chain distribution network to sell into the international market.
Thus you have avocado's selling for peanuts at local markets instead of earning big money selling to the west.
Building up a nation's capital stock takes time and long term stability. Thankfully Africa has the stability but time is slow.
Let's do 5 Whys with this? Basically, that means the industry can't reliably truck or otherwise ship refrigerated produce to the ports of Mombasa and Dar es Salaam. Why? I'm going to guess that the road infrastructure is too poor to support refrigerated trucks. Why?
Overall, the rule of law, regulatory environment and open market of Uganda and its neighbors are not at the level needed to to maximize its potential. This, coupled with the cost of borrowing (think interest rate), asset risk (E.g, forfeiture, robbery, piracy) and a few key negative economic indicators (Major balance of trade, current account & gov budget deficit) can put a dent in a country's mid-long term success.
- road network unreliable - theft of property along road - electricity unreliable - workers may or may not show up - bribes for transport, bribes for regulatory checks - unreliable access to credit - unreliable access to mechanics for truck, equipment - unreliable commercial and regulatory structures - unstable political climate - unreliable export/import laws - unreliable transport partners - unreliable internet - unreliable water source - excessive weather risk - unpredictable demand - unreliable payments from customers - payments delayed by snafus in credit/financial system - lack of appropriate insurance products
etc. etc. etc.
'Lack of data centres' in Africa is more likely a symptom, not a cause of anything.
Having access to high quality, well run data centres is actually a positive thing for Africa, since that is just a little higher up on the value chain than maybe some places can handle.
Once other things are in place, Africa can make all the data centres they need, but I'd argue that this is just an optimization. I don't see a hugely material issue with not having data locally.
Generators have several costs.
1. Fixed costs 2. Maintenance costs 3. Operating costs
#2 and #3 are lower for a generator you run 8 hours a year versus 800 hours a year. And you may want added redundancy for your generator capacity if you plan to run off generator power frequently, so #1 can be higher too.
(I broke out maintenance costs from operating costs since there's fixed maintenance costs even if you never run the generator, and there's variable maintenance costs based on the number of hours you run it)
It's been a long time since I've rented coloc space, but at the time, the cost of the power drops to the cage cost more than the space itself. Conditioned reliable power is expensive.
I'm sure China telecom will step up as part of their widening reach and influence in Africa.
1- https://workers.cloudflare.com/
2- https://www.cloudflare.com/network/
So you're saying capitalism has limitations on what it focuses on?
Granted, it would be way too expensive at the present time, even with reusable Falcon rockets.
The kinetic energy of a 30 metric ton datacenter moving at 9.8 km/s is 1,440,600,000,000 joules. That's a bit more energy than 344 tons of TNT and a bit more than 12,000 gallons of gasoline equivalent. That's actually much less than the energy embodied in a 747 with full tanks. (If I've calculated correctly.)
The other has to be pretty hands off to be profitable. and would be a literal building in space.
Both would be dangerous to have crash, it's easier to prevent the plane from coming down though.
Without the logistical infrastructure in place I don't know how you hope to drop a data center in Africa and hope to have costs be anywhere near what they are in the developed world.