Hacker News new | past | comments | ask | show | jobs | submit login
Microsoft to Deliver Microsoft Cloud from Datacenters in Africa (microsoft.com)
208 points by el_duderino on May 18, 2017 | hide | past | web | favorite | 109 comments

I am hosting my SaaS site using Google Cloud. All my clients are based in SA. I chose Google Cloud after experiencing very bad service from Azure. I won't go back to Azure, even if they build the datacenter behind my yard. I love the pricing and simplicity of Google Cloud. Ironically, my site is on Microsoft .NET stack, but I still don't trust Microsoft with hosting my site.

I hope Google and AWS will follow suite and build datacenter in SA.

Cool! Luckily, we have a POP in Johannesburg and it's "only" 160ms from there to London/Paris along our backbone (then another handful to europe-west1 in Belgium). Do you have good telemetry on the kind of latency your customers on seeing?

Disclosure: I work on Google Cloud.

I haven't checked. All I know is that they are happy. I will check and report back.

Just mentioning, i didn't knew Google Cloud does the .Net stack also. But i haven't had any bad experience with Azure ( small webapps, not huge ones). ( i chose Azure over Amazon)

I just publish and run. Some web apps have been running for over 4 years, without me touching it and paying customers ( 1-10). I only had one small "downtime" moment, but it was over after 15 minutes.

I wouldn't expect anything more from where i host it. The biggest annoyance i had was the redesign of the dashboard, which had me confused for a while ( it changed a lot, so that's normal), then again, i don't hate change. I switched to the new interface asap.

It was also a surprise for me when I found out that Google Cloud does Microsoft .NET. Within minutes of setup, I was already connecting to my newly created server via remote desktop. The rest I do manually.

My first choice was AWS. Their telephonic verification system failed when I was trying to signup and I wasn't allowed to finalise signup. Tried 3-5 times more and gave up. Just when I was about to rewrite my system using open source technologies, I discovered that Google Cloud does have .NET support. It was a big relief.

Service from Azure support has always been superb for me. What type of trial were you looking at?

I only needed a month trial.

they don't want your $59/month.

That's what I suspected too. Let me give it to Google who wants it. Even when the company grows and I require more power and more storage, I will still give it to Google.

eventually you'll find out google doesn't really want your $59/month either.

Care to give some specifics?

First I was told that I am not eligible for trial. Google offered me $300 credit which lasted for 5 months because the total monthly cost is $59.00. Azure was expensive compared to Google Cloud. Their pricing was also complicated and full of jargons. Microsoft offering was also not appealing. They were offering less memory and less storage space.

Up to this day, I still don't understand why Google can offer services using Microsoft software cheaper than Microsoft itself. Maybe I am just not their target market.

You don't have to be eligible for a trial. Everyone can open a trial account with 200$ for 3 months. Additionally Azure App Services has a nice free tier which, if your needs are basic, might even let you run free forever. Disclaimer - I work for Microsoft.

"Free Tier" is "60 CPU minutes / day" [1]. That's barely enough time to create the default project in Visual Studio and build it, let alone do some trial development off of it.

[1]: https://azure.microsoft.com/en-us/pricing/details/app-servic...

You do the development on your computer, then deploy it to Azure. See this example:


The CPU minutes/day come from serving your site's HTTP requests and the like.

So if you put up a site on a random domain and nobody ever touches it, it won't accumulate minutes?

Not without scheduled jobs (crons or the equivalent) or always-live processes, no. If you're leaning on a runtime the server owns and your own code is never invoked, you are (or your account is) not using any CPU time.

Thank you for the explanation. I learned something today. (To be fair the site is not clear.)

You said, "after experiencing very bad service from Azure".

Now that you give details it doesn't seem to fit what you said. Sounds like your issues were pricing (no trial, higher cost, fewer resources per $) and your confusion with Microsoft "jargon". That doesn't equal very bad service to me.

Your use of quotes around jargon qualifies as "scare quotes" clearly indicating your disdain for OP's concern about Microsoft jargon. Microsoft does indeed use jargon, and it's an entirely reasonable principle of purchasing to avoid products the purchasor don't understand. Jargon is a well-known source of misunderstanding.

That's just a quarter of the story. I wasn't happy with their customer service and setup tools too. That's part of the customer experience.

As for Google, I never needed to contact them. Everything just worked. Right information was provided every step of the way, which eliminated the need to contact them. The whole setup took me less than an hour and my site was up and running.

I had mostly positive experience with Azure. Sure, it's definitely not the cheapest cloud - but on the other hand it compares well with AWS. I credit Azure for focussing on simplicity and having a polite customer service.

Regarding Google, I had really bad experience with that at some point. When App Engine was still young, it was mega unreliable, it seemed not even the Google people could predict when it failed. Not only that, at some point, when it wasn't so beta anymore, they changed pricing model and raised the prices unproportionally high. Eventually I painfully migrated my app onto some Virtual Private Server.

> First I was told that I am not eligible for trial.

I had 150 € for free a month, for half a year or so and afterwards access to a 60k € program. You can try to apply for that.

Anyways, I don't want to make any advertisment. My current project is hosted on Scaleway using Docker, so I can do what I want and I don't depend on any Tech Giant's benevolence.

I'd suggest that's easy, Google are running a loss-leader to get you onto their platform.

In terms of market share on cloud computing platforms, Google is currently running third to AWS and Azure, so it makes sense for them to take some short term losses to help build their position.

Whether they will maintain that kind of pricing in the long term, remains to be seen.

Google has been running the most advanced cloud far longer than the others, and has the efficiency and costs to show for it, which they pass on to their customers.

They're definitely being aggressive with pricing but they're still making money. Their pricing is definitely way better to deal with than AWS or Azure.

Advanced I'd suggest is subjective, however Google cloud has not been running longer than AWS (GCP started in 2008 (https://en.wikipedia.org/wiki/Google_Cloud_Platform), AWS started in 2006 (https://en.wikipedia.org/wiki/Amazon_Web_Services), with Azure coming along a bit after those two in 2010 (https://en.wikipedia.org/wiki/Microsoft_Azure)

I'm giving OP the benefit of the doubt, assuming they're including the time Google itself was the only user of their cloud. Still not sure how that timeline compares to the Amazon-only pre-AWS.

I'm not talking about their public cloud offering but rather their entire operational experience as one of the first companies to truly use next-gen architecture to build their software and data centers.

They're just now offering the same tech to the public with GCP products like Compute Engine, Spanner, BigQuery, etc. which shows in the level of sophistication and performance compared to Azure and AWS.

I'm afraid that all feels very subjective to me, Microsoft were providing online services before Google even existed as a company, so I'm not sure how Google could have more operational experience of online services.

You obviously feel that Google's services are more advanced than Microsoft's and Amazon's but from a market share perspective they're clearly running third to those two.

What does marketshare have to do with how advanced the tech is or the lineage of their architecture? Especially marketshare of just their public clouds?

All 3 majors have great tech but Google has been doing much more advanced stuff which can be seen directly in their products. Everything in GCP is seamless from scaling without warm-ups, per-minute billing, automatic discounts, custom vm shapes, live migration, truly global load balancing, etc... and that's just in their compute engine product.

I doubt it is a loss leader. That's a fair price to pay for the kind of hosting I require. Local web hosting companies who are profitable charge less than what Google is charging me.

if they provide you with $300 of free credit how can that be anything other than a loss leader?

They're providing you something for free, which costs them money to provide...

> They're providing you something for free, which costs them money to provide...

At Google's scale, unless you mine bitcoins and actually use CPU/GPU time, the money Google actually spends on your tiny $300 free trial is maybe worth a dollar or two (and if they run free trials on machines that are paid off in terms of taxes, the only "expenses" are a tiny tiny sliver of maintenance - at Google scale network connectivity is basically free).

Cloud providers, especially for 24/7 running services, are a total rip-off even compared to managed servers, and deploying your own bare metal will blow any cloud calculation away when you're getting serious (> 200GB RAM, 50+ cores). The real strength of cloud is when you need extremely high computing power for very short time (e.g. heavy software builds, occasional transcoding, ML training).

that's still a loss leader...

And it's opportunity cost we're looking at here. It may only cost them a small amount to provide the resources (although that ignores the cost of developing/maintaining the services) but if they provide $300 of credit to one party that's resources they're not selling for $300 to someone else.

Funny, I don't have personal experience with either, but I'm always reading that Google's customer service is abysmal (i.e. non-existent). But then, maybe you just haven't needed them yet.

A month ago on the thread about AWS opening up a region in Sweden, a discussion ensued [1] about the lack of major cloud provider datacenters in Africa.

I wrote [2] that a good site in Africa would be challenging, because one would have to "pick a spot touched by more than one thick pipe, in jurisdiction known for political and civic stability and a regime with rough compatibility to the ideology and national security apparatus of western allies -- their home base and primary source of customers; receptive and promoting of foreign investment, and having access to multiple reliable, redundant power sources from which to draw energy."

While South Africa ranks high on stability and ideological compatibility, and reasonably on fiber [3], my understanding was that reliable and redundant utility power supply is a significant issue [4][5]. I'd be curious to see whether co-generation will be used to overcome these limitations.

[1] https://news.ycombinator.com/item?id=14031565 [2] https://news.ycombinator.com/item?id=14035251 [3] https://upload.wikimedia.org/wikipedia/commons/d/d4/Cable_ma... [4] http://www.fin24.com/Economy/SA-fears-dark-days-ahead-as-pow... [4] http://www.aljazeera.com/indepth/features/2015/03/crippling-... [5] http://www.aljazeera.com/indepth/features/2015/03/crippling-...

I am seeing a lot of comments about the need for stable power. It seems logical, but it’s not an absolute need. I managed a data center win San Diego at the time of the California electrical power crisis [1] near the end of the Dot Com days. I have operated more than one at a time since. During those times, which produced rolling blackouts, we just ended up running our generators more often. Power outages in Data Center happen during maintenance and upgrades, not so much during power grid outages. I have still never never had a data center go down during a power grid outage… but have seen many during upgrades, and maintenance.

In the years since the rolling blackouts, I have been colocated in data centers that have made the decision to use special pricing available from the power utility, available when you can have a preemptable load… during peak hours of peak season the power company can tell you to get off the grid with short notice. The data center operators did OK when they did this for most years. They got a lower price on power all year, in exchange for running their generators more. I’d say that the typical year, they ran the generators for several hours per day for 7 to 10 days in the Summer months. The last year they dd this, they ran a much higher number of days than expected… maybe 15 - 18 days from noon to 7pm. I understand they had to stop doing this because of the pollution of the generators or the permits required for such heavy usage... but that info is not first hand. It could have been many other reasons such as neighbors complaining about the noise in the business parks (2MW diesel generators are deafeningly loud), costs related to running the generators for so many hours, etc.

To operate like this, you have to be on your game for maintenance of the generators and checklists and training. You also have to have multiple contracts with fuel delivery trucks, just incase your outage lasts for a while. These data centers were all under 5MW in size each. We never lost critical load during a power supplier outage. I hope I have illustrated that a reliable power supply is not strictly required to run a reliable data center or service that is dependent upon a reliable data center.

[1] https://en.wikipedia.org/wiki/California_electricity_crisis

SA experienced power challenges few years back, but those issues are long gone. I haven't experienced any power cuts in over a year now.

Is it just me or is one year a pretty low bar to set for granting a problem the title of "long gone"?

Given the context of what caused those power issues: - lack of maintenance (not non-existence of maintenance) for almost 2 decades - the electrification of more homes without bringing on additional capacity - poor/subpar management of the sole power producer

A year or so is fine because those issues have been resolved by: - a slower economy (I'm putting this here because we shot ourselves on the foot) - more generational power coming in - the rise of independent power-producers.

So there should be stability from a power generation perspective.

It's "long gone" because depending on which part of society you come from, you might have gone a few years without experiencing load-based power issues.

In general the power issues still exist once you step out of the suburbs or metros, but then again the DCs will likely be hosted there, or MS will enter into a power agreement that gives them preference over the rest of the average consumer.

I'm just glad that we'll have Azure DCs as a start, and hope Google and Amazon follow suit if MS succeeds.

I consider it long gone because a permanent solution was put in place. Any problem disappearing for over a year is considered long gone in any country. Not about setting a low bar. In fact, I have heard politicians claiming to have fixed problems less than 100 days after they were put into power.

Google has already started working on this to a degree.


Cool. From what I know Africa is like the only continent so far without any datacenters from the major players. Correct me if I am wrong.

Glad to see things change and a small step into a more connected and peaceful world.

One issue is that neighbouring African countries don't have good connectivity - the connection to the EU is typically better than to their neighbours. So if you want to serve e.g. users in Kenya, it would typically be better to use a datacenter in the EU than one in South Africa.

Based on my own experience:

...because they don't know that the internet is a network of networks. I.e. They should build their own strong internal network.

Instead, their networks are designed only as dumb pipes to deliver the internet.

Yep. We've had to do some work for clients in Africa in my job. One we did in Morocco was fine from EU data centers on AWS, but South Africa has been very difficult to serve with acceptable latency from any data centre we've tried.

We're 90% AWS here, but would probably use Azure for those clients if they had a data centre in that region.

Some industries flat-out give you issues with moving SA data out, and others have players that are very risk-averse and aren't moving to the cloud.

My day-job is as a consultant in the financial sector. It's been always burdensome getting a project approved where the data will reside outside of the country. We have smaller players , but their rates are exorbitant in comparison to MS, GOOG and Amazon.

I'm glad and optimistic about this move from MS, because now we can get more work done with less red-tape.

UIh, any specific issues? My company will be rolling out SAP from Germany to ZA soonish, and so far we have not yet had a chance to actually really test the connection. Self-hosted data center here, so probably not that comparable. Just overall crappy connection?

My company handles live video streaming, so slow speeds and high latency tends to result in a lot of buffering. You're looking at about 200ms latency minimum from Europe to South Africa, and when you need to download a ~2MB video chunk in 2000ms, it's pretty hard to get a reliable stream, especially when you add in generally slower/less reliable connections along the route.

It can be mitigated somewhat by reducing the quality of the video and using CDNs, but even edge nodes are scarce in that are of the world.

I imagine it would be fine for normal applications, aside from being a little slower, but video is a challenge.

Thanks. We do not serve video, so it is indeed not that comparable. But we do have hosted web applications (SAP CRM), and syncing of offline applications for field staff (technicians) (and that can be quite a lot of data). It seems that our use cases mean either low latency and low bandwidth (non-technical users and their web apps, where "stuff is broken" if it is slow, but they might be already used to that from other areas of their online lifes) OR any latency and high bandwidth (sync doesn't really care if the packages come in with ultra low latency, as long as it syncs). Might work out indeed.

Hetzner South Africa have datacenters in Cape Town and Johannesburg. Their reliability and support is excellent. I am assuming that your services are not limited to cloud VMs and can run on bare metal machines.

A few years ago I recall that IBM announced a plan to set up something in Kenya.

But don't blame the industry for not having a major datacenter anywhere on that continent. You need a stable power supply for that to make sense, which very few countries there have. Even Ghana that used to brag about that is now a basket case, last I checked.

> You need a stable power supply for that to make sense

You also need a stable and at least substantially non-corrupt political situation, I would think? That rules out a lot of countries in Africa.

> You also need a stable and at least substantially non-corrupt political situation...

That rules out a lot of countries in the world including Russia, USA, and Brazil :/

All facetiousness aside, I am really glad to learn of this development. It goes to show the dynamic potential locked up and waiting to be exploited in Africa. The infrastructure problems are not insurmountable and the high returns accruable to investors will more than compensate any willing entrepreneur.

Africa is not an investment destination for the future. It is an investment destination for today.

(On a related note, a company I am inolved with is starting a training school for the next generation of elite developers from Nigeria and we are searching for international technical and financial partners. I'd love to hear from any interested HN readers.

Email in profile.)

> Email in profile.

I don't see any email in your profile.

My mistake. It's fixed now.


You also need a stable and at least substantially non-corrupt political situation

Actually that plays in the favor of foreign corporations, who are more likely to pony up bribes in your favorite foreign currency.

IBM launched their datacenter last year


Not a lot of datacenters in Antarctica. :(

Think about how much you would save in cooling...

Not much really, you'd be surprised

Well, you don't need all that much capacity. The penguins are into text-based adventure games running at dial-up speeds and the seals are table-top RPG players. Neither population has discovered online shopping.

Actually, they need bigger fans there because of the thinner air (high elevation).

Fully agree.

It is great that those countries get the infrastructure to enjoy what is already possible in the majority of the world.

Sure there are many roadblocks for that to be an easy goal, but not doing anything is worse I think.

Africa's a big place, but I think you're right. If you take out South Africa the rest of the continent is completely "out of the loop" concerning cloud services from the major players, it seems.

That's a lie. Kenya, Nigeria, Morocco, Egypt are just as adept to cloud as South Africa is.

EDIT:Screw the 'groupthink' on HN. Why would you downvote a simple non-offensive comment like this? Africa and its technology are my specific area for research. I know what I am talking about.

I think people are responding to your first sentence, "That's a lie." Which is a rather aggressive lead, particularly when I don't think your parent was trying to pull the wool over anyone's eyes. Maybe edit it to say "I can understand you position, however, Kenya...."

Or even just "That's incorrect".

I don't think Oblio was talking shit about Africa's tech scene just specifically that none of the big Cloud companies have any data centers outside of SA.

Though yes, if Oblio was talking shit about Africa they could use a bit of Three Stooges-esque slapping about.

I wasn't "talking shit about Africa", I have no hidden agenda about this topic (or any other kind of agenda).

My phrasing was a bit awkward but my point was exactly what you said.

Every time there's a discussion about major cloud providers (Amazon, Google, etc.) in Africa someone says: "oh, there are DCs in Africa". Then you look them up and they're all in South Africa and the rest of Africa is blank.

Im sorry but Kenya isn't as far ahead with the Cloud as South Africa is. The market has far more depth in South Africa. There are power problems around Mombasa (where the Seacom cable gets in), cost is just miles out on the electricity end too.

There's a reason AfriNIC is better suited to being in Kenya, but is in Mauritius. Even on raw metric terms you wouldn't find Kenya higher on a single metric in terms of cost, or bandwidth. The only one would be latency to Europe and that too due to physical constraints - and even so just on backbone lines.

"That's a lie."


I'm afraid these two things are incompatible with each other.

Really? I don't mind learning, so please point to me the data centers from major cloud providers in these countries.

If there are any, they're super recent.

I might believe Morocco and Egypt but Nigeria doesn't belong on that list.

Can you substantiate your disapproval? Don't just disapprove because you 'think so'. Better you 'know so'.

My Nigerian friends will read his statement and say, "He's right."

From what they tell me, some parts of Nigeria have never had electricity, especially in the oil-rich Niger Delta. Hard to believe, but that's the way it is.

The thing is though, North African countries like Morocco and Egypt are much closer to existing EU data centres than Johannesburg is. Even Nigeria seems to be about half as far as Joburg.

It seems to me that staking out the unclaimed position at the furthest southern economic hub makes sense as a first step, whether that hub has the most customers or not.

What's hilarious is that Amazon's S3 service is literally developed in Cape Town. Amazing that they weren't first with this.

As far as I know it's EC2 that's developed in Cape Town, not S3. Your point is still valid though.

I was just in a comment thread here where someone describe Africa as "a shithole" (their exact words ) when it came to hosting data centers. Glad to see that the people actual making decisions in tech have some rational ability.

I've never seriously considered using Azure, I'm quite happy with AWS. Problem is the closest zone to South Africa is EU. This might make me consider using Azure. I hope Amazon has similar plans in the works.

The thing that Azure has over AWS, imo, is the integration it has with Visual Studio.

South Africa is pretty much the only country where this was ever going to be possible.

Perhaps Rwanda as well.

Uninterrupted power is a major challenge here on the continent.

>Uninterrupted power is a major challenge here on the continent.

Power is no biggie - network providers have learned to live on power generators.

The main issue is that Cable, Fiber penetration is very low.

> Power is no biggie - network providers have learned to live on power generators.

Equinix datacenters in SV are actually cheaper to run on a generator, but it's not sustainable and against regulations (except obviously in a power outage).

Why is it not sustainable?

> Why is it not sustainable?

Because oil prices won't stay that low forever, in fact Saudi-Arabia and other OPEC members lowered the production quotas to get higher oil prices. The budgets of many countries with natural resources expect a certain price floor for oil - this especially hits Saudi-Arabia as they don't have much economy except oil and a tiny bit of Mecca tourism, but also Russia which has been hit hard by the drop in oil price and which doesn't have any kind of industry that's comparable to oil in terms of exports.

Fair enough, I thought it had something to do with the portable aspects of the generator in particular.

I am of the opinion that oil should really not be wasted for things that don’t require very high energy density, unlike transportation etc. But setting up a coal plant seems like a project of entirely different magnitude.

The margin in the telecom business is crazy high. As far as I know, no network in West Africa provides unlimited phone calls. - mostly prepaid and per second billing.

In other words, input (diesel in this case) is directly related to output (profit.)

I think you could do Morocco as well, but isn't uninterrupted power also a major issue in South Africa. You only have Eskom and they don't seem to be that well run or have enough well maintained power plants.

Having data centers in Morocco probably doesn't give that much of an advantage, seeing as it's so close to Europe.

Out of all the African countries, why Rwanda?

It's criss-crossed with fiber, the ICT policy is very open and the leadership friendly. Also it's a small country.

And landlocked, no natural resources, just a lot of people, so they need to export something to get foreign currency. IT is ideal so they are educating for that.

Also safe and not corrupt. Can't bribe yourself out of a speeding ticket there...

When I hear Rwanda I can't thinking the genocide that happened there only 23 years ago. Has it changed so much since those days?

It's completely unrecognizable.

They've made incredible strides.

If you're going to build a huge data center, how much harder is it to throw in a little natural gas or oil power plant on the edge of the campus?

That may be the solution they end up implementing.

Maybe Ghana as well or even Egypt.

I think the parent is too pessimistic. You are right, and Nigeria and Kenya would be acceptable choices too.


Are these cloud services going to be served from newly built 100% Microsoft datacenters or servers colocated with Hetzner South Africa (Cape Town and Joburg)?

I don't know anything about the technicals, politics, and economics of this kind of thing. What's the difference between the two setups?

I'm curious what the median salary is in these data centers.

MSFT is starting to get it. Many startups are being funded for stupid projects that target the US and are based in SF. The future belongs to markets like Africa and S. America. They currently take about 6% each of world's total software market.

This must increase the cooling demands a lot.

bad idea ....

No developer goes back to higher latency once you have tasted the smaller one !

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact