Trends in Cloud Computing: Who Uses AWS, Who Uses Azure?
Microsoft’s earnings were released yesterday, so we thought we’d take a look at how Microsoft technologies including Azure are used in comparison to Amazon Web Services across time, technologies, and industries, according to Stack Overflow data.
Change Over Time
How do the two cloud platforms compare? The Stack Overflow Trends tool shows the percentage of questions asked that have a particular tag (amazon-web-services or azure), and can give us a sense of each technology’s popularity since 2008.
It looks like they’ve both been growing consistently in their share of Stack Overflow questions. In the first few years of the site, each tag made up less than .1% of questions asked, but now they each make up more than .5% of new questions.
We can examine this in more detail by considering Stack Overflow visits over time, rather than questions asked. Our current traffic logs go back only to late 2011, but this does include much of the growth of the two cloud technologies. What percentage of question visits went to each tag within each month?
Both platforms have shown steady growth in the last six years, as a percentage of overall Stack Overflow question views. But it’s apparent that while the two platforms started at a similar level of traffic in 2012, AWS has grown faster.
This shows that while Stack Overflow Trends is a useful starting place to estimate a technology’s adoption, it doesn’t always tell the full story. In this case, the results are different because the average AWS question tends to get more visits than the average Azure question.
What might cause developers to choose AWS over Azure, or vice versa? One likely possibility is the language and technology stack they use: some languages and frameworks are better suited to a particular cloud platform.
We can get a sense of this by dividing developers into categories based on the tag they most visited, and considering the percentage of their traffic that goes to Amazon and Azure questions.
There’s a clear distinction: developers that use C# overwhelmingly choose Azure, while other developers use the platform to a much lesser extent. This makes sense, since Azure is a Microsoft product and is a natural choice for a web application in the Windows stack.
In the other direction, we can see that Node.js developers are by far the most likely to visit AWS questions. They’re also the second most likely to visit Azure questions, which suggests that Node.js developers are disproportionately likely to work with cloud hosting. We can see that Python and Ruby on Rails developers visit a substantial amount of AWS as well, but very few Azure questions.
Developers who work with C and C++ were particularly unlikely to use either platform, which makes sense since those languages are rarely deployed on cloud applications. We can also see that developers who heavily visited the HTML tag tended not to visit cloud questions either; these likely represent front-end developers and designers who aren’t responsible for configuring cloud platforms.
Another factor that can affect a developer’s choice of platform is the industry they work in. What industries tend to choose one cloud platform over the other?
We can consider this question within the United States by matching IP addresses to companies, just as we recently did to analyze trends in government software developers.
AWS questions may be visited more overall, but it looks like Azure is the platform of choice in several industries, particularly consulting and energy. Further analysis shows that these are the industries where the Microsoft stack is most popular. In contrast, AWS is particularly popular in the technology industry (such as software and web companies), and especially in media companies (including publishing and entertainment). You can explore the companies hiring AWS developers on our Jobs listings.
It’s also worth noticing that the “Academic” sector visits both cloud platforms the least out of all their question views. This category is generally made up mostly of students and researchers, and we showed in a previous post that these users visit a very different set of languages and technologies than the rest of the United States. (In fact, that post found that Azure was the single most underrepresented tag in questions asked from universities). This suggests cloud platforms aren’t widely taught in universities or used extensively in research, at least compared to technologies like Python and R.
Developers in different regions also tend to choose different technologies. Much as we compared developers in major cities, we can consider how questions about these two platforms are visited within each country. Here we’re examining one year of traffic (July 2016-June 2017, about 6.2 billion visits overall).
Countries differ substantially in how they visit these two cloud products, relative to the rest of their Stack Overflow visits. Countries like Australia, Japan, the UK and the US often visited questions about these these platforms, while countries like Indonesia, China and Russia visited them more rarely compared to other tags.
Most countries visited more AWS questions than Azure, though to different extents. One notable exception is the Netherlands, which visited about twice as many Azure questions as AWS questions. On plausible factor is that Microsoft has built a major data center in the Netherlands, which along with other investments in the region likely means there are disproportionately many Azure cloud developers in the country.
All the numbers are below 1%? That’s not a lot.
1% of the total represents over 100,000 questions. There are over 50,000 tags. Just two of them each taking >0.5% of new questions is a lot.
Indeed, looked like that it’s quite huge
As the comment gets deleted.
Hi everyone, my name is Festus.
Hello Festyus, my name is Silas
Hello Festus, my name is Habeeb.
Hi Festus, my name is Google Cloud. People don’t seem to like my name much, so I go by Tyler.
TL;DR you’re a weirdo if you write C# and use AWS :'(
TL;DR . You’re an idiot if you use Azure.
I expect university researchers are less likely to use Azure or AWS because they have access to academic computing resources like high performance computing – effectively they have private clouds.
I was wondering the same. I’m thinking perhaps (just a guess, nothing data specific) that GCP would be around the same languages as AWS, except perhaps a higher amount using Go. It could also be that it was easier just to compare two instead of three.
The x and y axis have different scales in a lot of the pictures. I understand that you want to show all data and it makes it hard to visualise but it paints an unfair picture.
That’s right! Always look at the numbers along both sides of the graph to see if you’re being deceived. Thanks Head First Statistics!
In those graphs where axis stepping is different on each side, they include a dashed line that represents y=x, as it says above the graphs in small print. In comparing traffic by country, you can see it’s roughly 2 aws to 1 azure, and that’s reflected in that line, since if the graph were equal, that line would be roughly y=2x. The exception to this is the technology stack graph, which doesn’t have a y=x line, unfortunately.
More questions could also mean that one product is more problematic to use than the other, not necessarily that it is more popular.
Or maybe the learning curve it’s more complicated, or just not well documented.
“We can examine this in more detail by considering Stack Overflow visits over time, rather than questions asked.”
I didn’t read the article thoroughly but I think much of it was centered around visits rather than questions.
I’d expect visits and questions to be equally correlated with ease of use.
The AWS stack is considerably less user friendly, possibly because it matured differently (ie., with hadly much competition initially to compare its feature consumption) than Azure (which being a Microsoft product is more user friendly because its users are used to Microsoft features, how they behave, how to use etc – the UX is pretty familiar).
Consider the payment page, in aws, where the “Pay now” button is placed above a text box for CVV, looking like a heading of the small frame for taking CVV input. If you need to stop propagating a bill for stopped or deleted ECS instances (Yes, the billing happens if the user does not delete or set some propagation value in the instance’s resources!) so yes, the problematic part is well noted by @@200_success:disqus
I would have expected you not to ignore Bluemix, which is in the space with Amazon Web Services and Azure. Those two are not the only games in town!
Before that, don’t you think they should consider Google Cloud first? It’s just tough to display these graphs in 3D
I’m no Gartner shill, but take a look at the magic quadrant – it does tell the story of why the article deals only with (and I hate M$) the two outstanding leaders in the area. Why did they not cover Bluemix? Because it’s rubbish!
Really awesome blog!!
Can we have such analysis bin the context of Indian market
Thanks for the post.
I’m also an AWS user and it’s good for my business.
what type of business?
This is a timely discussion for me. I have just come back from a Cloud Computing workshop with sessions from AWS and Google Cloud on scientific computing. We scientists haven’t been spruiked at much by Azure yet. It would be interesting to see those stats broken down into business vs scientific or is that covered as “Academic”?
(And I work at an Australian Supecomputer Centre and academics are increasingly turning to cloud as we just don’t have enough compute cycles to go round.)
Hi rxk900, have you come across this paper comparing cloud platforms for HPC workloads: https://arxiv.org/pdf/1702.02968.pdf
Hi Igor, Thank you. I haven’t seen that particular one though in return I give you: https://www.researchgate.net/publication/306067884_Breaking_HPC_Barriers_with_the_56GbE_Cloud and references therein. I haven’t looked at this for a while but also have some related unfinished drafts focusing on chemistry. Feel free to contact me if you want to discuss further (you can guess which author I am).
thanks for sharing, interesting read!
Azure is very well documented so questions graph may be lower as Azure is becoming better understood by the developer community. If you want a true depictions I guess comparing consumption from each would be a more accurate measure.
+1 and this was my thought too – just because questions are asked and visited on SO, does it mean there is a commensurate relationship wih use of the queried technology? My reasoning is no; MSDN forums cater for azure queries too, and people may hit them up first/only as it’s all “Microsoft ecosystem” that way or, maybe Microsoft got their act together with azure documentation and blogging HOWTOs so well that nothing ever goes wrong and everyone can do what they set out to, without asking questions about it..
“In the first few years of the site, each tag made up less than .1% of questions asked, but now they each make up more than .5% of new questions.”
Either you are wrongh, or your graph is wrong.The maximum reached for AWS, according to the graph, is below 0.70% ..
.70% is more than .5%.
Yeah it’s like 65% more ;);)
Decimated that argument
Since when popularity is being measured by # of questions asked in SO? The article itself starts with a huge fallacy. Aside that nice analysis btw.
Amount of questions asked can also mean a service is hard to understand. Lack of documentation, Lack of support..
I agree with this Gentleman. Azure questions dropped over time because the MS team really made a massive effort in making the UI easier. Not just the UI but the process from start to end is very streamlined for many different aspects. StackOverflow is still the best place to find the correct answers for highly technical questions but allot of MS Evangelists provide extremely comprehensive tutorials now (maybe that is why there was a decline to SO) That includes me, actually right now, I am reading the VisualStudio blog on how to setup CI with Selenium in Azure. The company I work at currently (C# team) really wishes they chose Azure over AWS due to the far easier integration provided now. I am guilty of of trolling many many pages on SO about AWS- Just because I have no idea how that spaghetti junction is glued together.. It is a bit mental.
Frankly, if you’re using the UI, you’re doing it wrong. MS is learning that with it’s OS, seems like it’s not with it’s cloud platform.
Actually, the az cli is extremely feature rich. I can spin up a production Kubernetes cluster with 1 command from an authenticated shell session. A second command will fetch the credentials. By any measure, that’s pretty impressive and zero need for a UI.
You mean Amazon UI? I do not use Amazon so I gotta start somewhere yea?
Any reason why GCP is missing?
I had the same question as soon as I read the title.
I can see GCP really taking off. I’m a .Net and Js guy and I actually tend to favour the platform with the docs and management console (personal preference) which to me is Azure. But GCP really is a great experience overall in this regard. Not used it as much as the other two but have been very impressed recently.
I think the answer is obvious. Because GCP is way smaller compared with the other two. Probably, in terms of percentage, it would always fall somewhere near the zero point on the chart…
Hello all my name is Rakesh i am new here. nice post very informative.
Any update on the industry data used here given it is not fairly aged?
Shame on you for playing around with the graph scales to make Azure look much better. You completely skewed the look of the graphs doing this.
If you adjust scale back to equal, you would get the OPPOSITE interpretation of the graphs.
Lies, Damn Lies, and Statistics…
Amazon Web Services and Microsoft Azure may be fierce cloud rivals, but enterprises are pairing both giants up in multi-cloud deployments, according to a survey by network analytics company Kentik.