Would it seriously have gone to anywhere else? Not saying there aren't other big economies (such as Nigeria), but is there any real reason this isn't a no brainer?
- Inconsistent electricity supply (brown outs are somewhat common)
- Poor physical infrastructure (at least a decade ago, copper cable theft was common)
- Business-unfriendly practices - I don't have the exact details, but you often have to run things through a local company with sufficient black representation. Sounds innocent enough, but does create a monopoly you have to deal with. Effectively a tax or bribe depending on perspective.
- Poor connectivity to the rest of the world (slowly getting there with more fibre cables arriving)
- (Edit) Severe water shortages, at least a couple years ago in Cape Town.
I would have thought somewhere North Africa would have been a real contender to make the decision require more thought. E.g. Morocco or Egypt, each with their own pros and cons.
Spitballing, what is internet connectivity like between North Africa and Sub-Saharan Africa?
If the point of putting a datacenter in Africa is to serve users in Africa better, North Africa may already get pretty good connectivity from European datacenters, so putting the datacenter on the other end may help more overall?
> Spitballing, what is internet connectivity like between North Africa and Sub-Saharan Africa?
Almost entirely via undersea cables. Cape Town, on the coast at the SW corner of Africa, is actually one of the closest places in Southern Africa to Europe and US, measuring by ping time.
North Africa is already "close enough" to data centres in Europe and/or Middle-east.
Yep. I work on latency-sensitive stuff (realtime A/V infra) and our customers have users throughout Africa. North Africa is pretty well connected to Europe, and the Horn of Africa is well connected to the Gulf region and to Mumbai. Southern Africa is the obvious location for a first datacentre, and South Africa is probably the only decent choice in southern Africa.
Someone like Google should have very good information on where their current customers (and their customers) are, and that alone could be decisive.
If many of the customers in Africa are already well served by Europe (think Egypt) or the Middle East, and of the remainder a large percentage is in South Africa itself, then South Africa makes a good choice for the first.
After all, starting in SA doesn't prevent them opening further datacenter in other countries as they grow.
I didn't downvote, but honestly just don't understand your comment. I thought it was obvious but maybe I shouldn't have left it unsaid: South Africa has significantly better infrastructure than any other country in southern Africa. We're not talking a small difference. If you want to open a cloud region in southern Africa, I don't think you have any other choices.
Wouldn't Botswana be as good if not better? It has significantly lower corruption issues, and higher GDP per capita and HDI when compared to SA. It also is an incredibly fast growing economy and it has lower crime. Does Botswana just currently not have the talent or infrastructure?
I think a significant factor, apart from generally worse infra and a smaller talent pool, would be the lack of a coastline. Basically all the long-distance capacity in Africa is in the form of submarine cables going around the coast. A landlocked country means you have to rely on terrestrial systems through at least one third country, introducing a whole range of new costs and risks, and probably makes it hard to get the desired level of control/ownership for the inter-region connectivity (cloud providers are often investing in submarine cable systems directly now).
> - Inconsistent electricity supply (brown outs are somewhat common)
No serious data center in South Africa is affected by this. All have batteries and diesel generators.
> - Poor physical infrastructure (at least a decade ago, copper cable theft was common)
Luckily copper theft affects DC's and fibre less. With redundancies it is manageable.
> - Business-unfriendly practices - I don't have the exact details, but you often have to run things through a local company with sufficient black representation. Sounds innocent enough, but does create a monopoly you have to deal with. Effectively a tax or bribe depending on perspective.
Yes this is a long term problem with South Africa. It stems from the ruling party (ANC) being super corrupt and embracing victim mentality to shift the blame. But there are often ways around Black Economic Empowerment not involving bribes.
> - Poor connectivity to the rest of the world (slowly getting there with more fibre cables arriving)
There are tons of cables now. We rarely see international outages. Since COVID also a couple of additional redundancies.
> - (Edit) Severe water shortages, at least a couple years ago in Cape Town.
I don't think this affected DC's much. And it's not like we had empty taps. I remember just using sprinklers less and topping up my swimming pool from roof-collected water instead of drinking water. Water shortage is hopefully being addressed long term and reservoirs are mostly full.
There's a pretty big one in Cape Town. It's beautiful here, the food is good, internet is stable, and the coffee is cheap, so naturally we have hoards of Europeans (mostly Germans) here in the summer.
Active tech scene particularly around Cape Town, the waterfront & Stellenbosch. Plenty makers, hackers & digital nomads. If you're looking for the outdoor life, MTB & hiking checkout Greyton, 2hrs from Cape Town. Just connected to the fibre network with speeds of 500mbps available. Beautiful mountains, lots of well priced restaurants&pubs. There's a Google engineer, an AWS product manager and plenty designers, coders and nomads in the village. Just won South Africa's Dorp (town) of the year. Ye olde England charm with warm sunny weather and super chilled locals.
> No serious data center in South Africa is affected by this. All have batteries and diesel generators.
It does mean higher operational costs since you're going to be paying for the unreliable electrical service in addition to regular usage of on-site electrical generation. So I wouldn't say no one is affected by this, but I will acknowledge the effects aren't measured in visible downtime.
I would be amazed if the electricity situation in South Africa is anywhere near as bad as in Nigeria, but I've never been to South Africa so it's just an assumption based on how incredibly bad Nigeria is.
> I would have thought somewhere North Africa would have been a real contender to make the decision require more thought.
North Africa is "not that far" from existing Data Centres in Europe and Middle-east. It doesn't fill a gap like a data centre at the other end of Africa does.
> I would have thought somewhere North Africa would have been a real contender
Is there really a point in having your first African region in North Africa? North Africa, especially Morocco is pretty close to Europe, where Google already has a big presence. If you want to serve Africa, you might as well start with a location that is further away from where you already are. Another argument going for South Africa.
yeah, we/they do (I'm South African but don't live there).
There is a famous app is RSA, called EskomSePush (https://sepush.co.za/), that sends users push notifications on when there will be blackouts.
Eskom is the Energy supplier in RSA, and "se Push" is a word play on Push notifications and Afrikaans slang for vagina (poes)
Still going on. We were without power from 12h00 to 14h00 this afternoon and will be again from 20h00 to 22h00 tonight, and that pattern will persist for at least the next few days (along with another outage in the wee hours so most of us don't care much.)
I have a solar contractor coming tomorrow to do the planning for an off-grid installation...
From the article it says their cable had four landings- Togo, Namibia, Nigeria and South Africa. In my opinion the location isn't really the interesting part of the article, it points out how the other major cloud players are already set up in Africa. What took Google so long?
And if the oceanic cables land in population centers, the latency to other locations might not have been that bad - you wait until you have indication that enough load would shift to your new datacenter before building.
> And if the oceanic cables land in population centers,
The cable will land in Cape Town, a coastal population centre.
Johannesburg is the largest business and population centre, but it's far inland. Cape Town, though it is further south, will have a smaller ping time to US and Europe.
> is there any real reason this isn't a no brainer?
Yes, very much so. The South African energy crisis[1]. They've been having trouble keeping power on since 2007. They suffer virtually constant rolling blackouts. Running Google datacenters there there will probably come at great cost to others, so this probably took a lot of negotiating.
This was my first thought. But perhaps there is less utility in placing this datacenter in Egypt, where it would be relatively close to datacenters already in Europe? I think South Africa probably gets them more bang for their buck, since it's much further from others.
Egypt already has low latency and good connectivity to cloud regions in Italy (and further afield in Europe), Bahrain, Saudi Arabia, Dubai etc. Meanwhile latency from South Africa to European cloud regions is >150ms. IOW, putting the first cloud region in Africa in South Africa makes sense because southern Africa is the region that needs one the most.
Much larger population, more deeply entrenched data (phone) usage, and cable landings that are close, in hops, to European cable landings like Marseille.
During a lunch conversation years ago with init7 (European IP transit) engineers we discussed a fictional CDN and I, naively, threw out Dubai as a potential regional hub and they scoffed. Egypt is where the peering is.
Egypt is also larger in GDP than South Africa, 2nd on the continent behind Nigeria[1].
The interesting piece in decision process from my view is can North African customers support their operations with European regions due to the better peering/additional cabling while further south they really need something local in order to drive cloud adoption?
Having had to find a location to put a POP some years ago: Morocco, Egypt, Kenya, Nigeria are also OK, depending on your needs. WACS strings along the entire Western coast of Africa, so anywhere along there that you can get reliable power, you can probably get a drop from WACS to the wider Internet and run your services.
To your point, tho, I ended up putting the POP in Cape Town, because it's a mega cable landing site and in a developed country.
A downside: there are (or at least were) some annoying "data export charges" levied per-Mb by the ZA government on traffic to services hosted outside of ZA. That's the whole reason I needed to put a POP there to begin with. Our backups over the wire were a little expensive, though.
The list of countries that I see as plausible for consideration in Sub-Saharan Africa are: Nigeria, Ivory Coast, Ghana, Botswana, South Africa, Tanzania, Kenya.
Most of the fiber cables in Africa run along the coasts, so inland countries are likely to have poor connectivity. Botswana is the only inland African country I'd contemplate, and that's because it has the massive benefit of being generally considered the most well-run country in Africa (outside of maybe the island nations). But I don't know what its internet connectivity looks like.
I hope they hording diesel, the power provider is continuously forcing 4000MW+ load shedding/reduction, with 4+ hours of rolling blackout everyday over the entire country.
I'd also bet some money on that. Even at a residential level, the fiber rollout in many suburbs and few townships allows us to remain connected during the 2 hour planned outages. A lot more people are buying small UPS devices to power our routers and modems.
A lot of businesses continue to invest in solar, I don't think a cloud provider that's carbon-conscious would leave their operations to diesel generators and mostly non-renewable electricity (https://www.eskom.co.za/dataportal/supply-side/station-build...). The link is to the supply mix over the past 7 days.
I'm South African, working remotely for a US employer. We often joke that my colleagues' broadband is 3rd world while mine is 1st world, as it seems easier to get a good fiber line here than there.
Some parts of SA are pretty high crime. Id imagine they have high security in place. Many of my friends there have electric fences and walls around their homes and have still been robbed at gun point.
The biggest risk is the government. In that way, South Africa is the only choice. ...and even that one is not so great so I suspect they will limit their footprint.
64 comments
[ 3.9 ms ] story [ 125 ms ] thread- Inconsistent electricity supply (brown outs are somewhat common)
- Poor physical infrastructure (at least a decade ago, copper cable theft was common)
- Business-unfriendly practices - I don't have the exact details, but you often have to run things through a local company with sufficient black representation. Sounds innocent enough, but does create a monopoly you have to deal with. Effectively a tax or bribe depending on perspective.
- Poor connectivity to the rest of the world (slowly getting there with more fibre cables arriving)
- (Edit) Severe water shortages, at least a couple years ago in Cape Town.
I would have thought somewhere North Africa would have been a real contender to make the decision require more thought. E.g. Morocco or Egypt, each with their own pros and cons.
If the point of putting a datacenter in Africa is to serve users in Africa better, North Africa may already get pretty good connectivity from European datacenters, so putting the datacenter on the other end may help more overall?
Almost entirely via undersea cables. Cape Town, on the coast at the SW corner of Africa, is actually one of the closest places in Southern Africa to Europe and US, measuring by ping time.
North Africa is already "close enough" to data centres in Europe and/or Middle-east.
It’s not racism, it’s just cost-benefit analysis, am I reading it correctly?
If many of the customers in Africa are already well served by Europe (think Egypt) or the Middle East, and of the remainder a large percentage is in South Africa itself, then South Africa makes a good choice for the first.
After all, starting in SA doesn't prevent them opening further datacenter in other countries as they grow.
But yeah, the question is honest, and would appreciate an answer.
> - Inconsistent electricity supply (brown outs are somewhat common)
No serious data center in South Africa is affected by this. All have batteries and diesel generators.
> - Poor physical infrastructure (at least a decade ago, copper cable theft was common)
Luckily copper theft affects DC's and fibre less. With redundancies it is manageable.
> - Business-unfriendly practices - I don't have the exact details, but you often have to run things through a local company with sufficient black representation. Sounds innocent enough, but does create a monopoly you have to deal with. Effectively a tax or bribe depending on perspective.
Yes this is a long term problem with South Africa. It stems from the ruling party (ANC) being super corrupt and embracing victim mentality to shift the blame. But there are often ways around Black Economic Empowerment not involving bribes.
> - Poor connectivity to the rest of the world (slowly getting there with more fibre cables arriving)
There are tons of cables now. We rarely see international outages. Since COVID also a couple of additional redundancies.
> - (Edit) Severe water shortages, at least a couple years ago in Cape Town.
I don't think this affected DC's much. And it's not like we had empty taps. I remember just using sprinklers less and topping up my swimming pool from roof-collected water instead of drinking water. Water shortage is hopefully being addressed long term and reservoirs are mostly full.
Thanks in advance.
NB: summer in SA is November -> March, which is a huge win if Northern Europe's dreary winter is getting you down.
But it can be beastly hot in Jan/Feb, if you don't have reliable aircon.
https://youtu.be/9Gk_I_0eMDA
It does mean higher operational costs since you're going to be paying for the unreliable electrical service in addition to regular usage of on-site electrical generation. So I wouldn't say no one is affected by this, but I will acknowledge the effects aren't measured in visible downtime.
I’m not saying South Africa is wrong, just saying it shouldn’t be considered the only real option.
North Africa is "not that far" from existing Data Centres in Europe and Middle-east. It doesn't fill a gap like a data centre at the other end of Africa does.
Is there really a point in having your first African region in North Africa? North Africa, especially Morocco is pretty close to Europe, where Google already has a big presence. If you want to serve Africa, you might as well start with a location that is further away from where you already are. Another argument going for South Africa.
Eskom is the Energy supplier in RSA, and "se Push" is a word play on Push notifications and Afrikaans slang for vagina (poes)
I have a solar contractor coming tomorrow to do the planning for an off-grid installation...
The cable will land in Cape Town, a coastal population centre.
Johannesburg is the largest business and population centre, but it's far inland. Cape Town, though it is further south, will have a smaller ping time to US and Europe.
Yes, very much so. The South African energy crisis[1]. They've been having trouble keeping power on since 2007. They suffer virtually constant rolling blackouts. Running Google datacenters there there will probably come at great cost to others, so this probably took a lot of negotiating.
1. https://en.wikipedia.org/wiki/South_African_energy_crisis
Absolutely - Egypt.
Much larger population, more deeply entrenched data (phone) usage, and cable landings that are close, in hops, to European cable landings like Marseille.
During a lunch conversation years ago with init7 (European IP transit) engineers we discussed a fictional CDN and I, naively, threw out Dubai as a potential regional hub and they scoffed. Egypt is where the peering is.
The interesting piece in decision process from my view is can North African customers support their operations with European regions due to the better peering/additional cabling while further south they really need something local in order to drive cloud adoption?
[1] https://en.wikipedia.org/wiki/List_of_African_countries_by_G...
South Africa is filling a gap.
To your point, tho, I ended up putting the POP in Cape Town, because it's a mega cable landing site and in a developed country.
A downside: there are (or at least were) some annoying "data export charges" levied per-Mb by the ZA government on traffic to services hosted outside of ZA. That's the whole reason I needed to put a POP there to begin with. Our backups over the wire were a little expensive, though.
Most of the fiber cables in Africa run along the coasts, so inland countries are likely to have poor connectivity. Botswana is the only inland African country I'd contemplate, and that's because it has the massive benefit of being generally considered the most well-run country in Africa (outside of maybe the island nations). But I don't know what its internet connectivity looks like.
That said, I'll bet solid Swiss money that Google is aware of the electricity supply situation here and has planned/is planning around it.
A lot of businesses continue to invest in solar, I don't think a cloud provider that's carbon-conscious would leave their operations to diesel generators and mostly non-renewable electricity (https://www.eskom.co.za/dataportal/supply-side/station-build...). The link is to the supply mix over the past 7 days.
I'm South African, working remotely for a US employer. We often joke that my colleagues' broadband is 3rd world while mine is 1st world, as it seems easier to get a good fiber line here than there.
https://ourworldindata.org/grapher/homicide-rate-1990-vs-201...
https://ourworldindata.org/grapher/homicide-rates-across-wes...
https://cloud.google.com/about/locations
- Google promised to invest $1B in Africa
- Google will build subsea cable will cut across South Africa, Namibia, Nigeria and St Helena, connecting Africa and Europe.
- Investments in Startups and small businesses
[1] https://techcrunch.com/2021/10/06/google-confirms-1b-investm...