The most popular Linux server product ?
I was working on slides for a presentation I’ll be giving next month at UKUUG Spring conference in Manchester, and wanted to show the growth of Ubuntu Server usage, compared to other Linux enterprise-oriented server products.
Given that there is no registration or license in Ubuntu Server, we are obviously lacking hard numbers, so I used Google Trends to try to use “Google popularity” as a metric.
It’s quite easy to bend the results by exploiting the well-known “Ubuntu” search term popularity, so I used pure product names, i.e. “Ubuntu Server” searches rather than searches with “Ubuntu” and “server”. I tried to include any reasonable way of naming the two other Linux server enterprise distributions (usually considered as #1 and #2 in market studies), and here is the result:
I was pleasantly surprised by the result… If you think I missed a way to name the competitors, or otherwise made an unfair comparison, please let me know. Google popularity is clearly not a perfect metric, and does not accurately reflect enterprise adoption levels or market share, but that still tends to prove the “Ubuntu Server” product has quickly become a first-class citizen… and is still growing.
See for yourself here.
Leave a reply to Thierry Carrez Cancel reply
Ubuntu Server blog
- An error has occurred; the feed is probably down. Try again later.
Seeing the fnords 
If you add centos to the redhat line then the results change; and centos basicly is a free (as in you do not have to pay) version of rhel. I don’t think many people use centos on desktops so it’s IMHO valuable to add it since ubuntu server is also free.
and anyway a centos vs ubuntu server graph is also interesting:
http://www.google.com/trends?q=%22ubuntu+server%22+%2C+centos&ctab=0&geo=all&date=all&sort=0
If you include centOS to the redhat query, the picture changes dramatically:
http://www.google.com/trends?q=%22ubuntu+server%22%2C+%22redhat+enterprise+linux%22+|+rhel+|+%22redhat+server%22+|+%22centos%22%2C+%22suse+linux+enterprise+server%22+|+sles+|+%22suse+server%22+|+%22novell+server%22&ctab=0&geo=all&date=all&sort=0
I was coming here to make the same comment! CentOS is a huge part of Red Hat’s prominence.
Hi,
Red Hat is named in two words by some (among others Red Hat Inc. themselves) so your results are not accurate, but the trend they show is definitely right.
I re-run the analysis with the addition of Red Hat instaeda of redhat. You can see the result here:
http://www.google.com/trends?q=“ubuntu+server”,+”redhat+enterprise+linux”+|+”red+hat+enterprise+linux”+|+rhel+|+”redhat+server”+|+”red+hat+server”,+”suse+linux+enterprise+server”+|+sles+|+”suse+server”+|+”novell+server”&ctab=0&geo=all&date=all&sort=0
“Red Hat” is two words, but people write it sometimes as one word.
Also, people sometimes still use “RedHat Linux” or “Red Hat Linux” to describe RHEL.
I don’t think it’s fair to add a one-word “brand” (like “RedHat”, or “CentOS” or “Ubuntu”) to the mix, because it skews the results of my product-oriented search dramatically… And adding “CentOS server” to the RedHat results doesn’t change the results at all.
It’s true, however, that I should have taken the “Red Hat” case into account, and that the results are much closer in that case.
What about Debian?
For Debian, you have the same effect as with CentOS: If you add “Debian server”, its results are really low. If you add “Debian”, it scores very high, as all one-word brand searches. I don’t think either is a fair representation…
Do you really think people are using CentOS on desktops ? I would bet most desktop users care for fresh stuff so they would rather use current Fedora release not CentOS. CentOS is RHEL nothing changes there beside the lack of logo so if you google up “rhel” there is nothing wrong in including centos in the results since it is IMHO almost the same thing.
I don’t think you got my point : given how Google trends works, single-word searches should only be compared to other single-word searches, because they will be more prominent by an order of magnitude. That’s why I compared PRODUCT NAMES and not the underlying packages or technologies or “distribution family”. You can do a RedHat+CentOS+Fedora vs. Ubuntu comparison if you really want to use single words (http://tinyurl.com/yk7th5u).
Given how google trend works… it shouldn’t be used to make these sort of comparisons..period. These sorts of graphs from google trends are complete fantasy. Noone knows how to intepret what Google trends is saying… even Google.
To get as semantically picky as to go into detail about the problems associated with scoring one word versus multiword terms sort of misses the bigger problem with interpreting any comparative result from Google trends.
-jef
I’m going to point out the logical fallacy of using Google trends the way you are trying to do without getting into the complication of products across brands. Let’s just look at the Ubuntu branded products shall we.
Which would you intuitively expect to see more hits from in Google trends? The phrase: “Ubuntu Desktop” or “Ubuntu Server?”
http://tinyurl.com/yjxk9az
When Google Trends gives you the ridiculous result that “Ubuntu server” has always trended much much higher than “Ubuntu Desktop” do you interpret that as meaning that “Ubuntu Server” is wildly more popular as a product than “Ubuntu desktop?” in the same way that you have done for server products across brands?
Here’s the problem… you’ve made several implicit assumptions about how different groups search for individual “products” in how you have chosen to interpret any result from Google trends. Assumptions you have not tested for validity. Just because Google trends makes pretty graphs doesn’t mean your interpretation holds up.
Are you really prepared to stand up in front of everyone and publicly say that “Ubuntu Server” is and has always been that much more popular than “Ubuntu Desktop?” based on what Google trends is actually measuring? And if you aren’t prepared to say that, then you can’t say anything about how “ubuntu Server” compares to any other server product either.
It’s very simple. Google trends is absolutely not reliable for comparative analysis for any brand or product because we simply do not have a testable model of search behavior which correlates well with usage or buying habits.
-jef
http://www.google.com/trends?q=“Ubuntu+Desktop”+%2C+”Ubuntu+Server”&ctab=0&geo=all&date=all&sort=0
Not trying to compare hard numbers, would you agree that it shows that Ubuntu Server is getting stronger, while the others stagnate ?
Stronger in what regard? Stagnating in what regard? Google search stats do not correlate to usage. If you continue to try to make that leap of faith you are going to fall down flat in public.
If you look at “Ubuntu Desktop” versus “Ubuntu Server” would you say that Ubuntu Desktop is stagnating while Ubuntu Server is growing. That’s what Google trends says, are you prepared to say that with as much force as as what you are attempting to do with Google trends here?
You need to stop trying to make comparisons out of Google trends as selectively as you are doing. It is NOT a reliable comparative metric for anything.
If you want to count usage…start counting usage directly..stop trying to make more out of Google trends then is credible.
-jef
I think what it proves is that Google searches for “Ubuntu Server” are becoming more popular, while the other search combinations mentioned tend to stagnate. Doesn’t prove anything else (as I wrote in the original post), and certainly not usage metrics or proof of product quality.
In the same vein, Google searches for the exact “Ubuntu Desktop” words (better known as “Ubuntu”, in the same way “CentOS server” is more widely called “CentOS”) are also stagnating.
http://www.moolux.org/2010/02/facts-about-linux-kernel.html
Interesting how the trends reliably drop for christmas 🙂
Our customers use RHEL, CentOS, SuSE/SLES, Debian and Ubuntu in our datacenters, and we support them all. We preferred Debian in the past, and now recommend Ubuntu for most workloads. We want to migrate our customers away from CentOS to either RHEL or Ubuntu, as there really is no good reason to run CentOS except for saving some dollars if you *need* RHEL for specific applications. CentOS as a project is not backed by a commerical entity, and you can not buy support nor get certified configurations.
That’s wrong.. there are commercial entities which are willing to sell support contracts for CentOS.
http://www.owlriver.com/support/centos/
You are missing the point. Google trends is a complete fantasy metric to gauge anything meaningful because there isn’t an understanding of the physiology of search. You should never ever stand up Google trend as a comparative term analysis like what you’ve attempt to do here.
Every single Google trend graph that someone throws up to support is psuedo-science..its snake-oil. Yours is no exception to that trend.
Okay..think about it. Google trends is saying that search for “Ubuntu” and “Ubuntu desktop” is stagnating..or if you look really closely decreasing over the last 12 months. Is that at all consistent with hard numbers that you have access to internally at Canonical about overall Ubuntu Desktop popularity for whatever definition of “popularity” you choose to apply? Should that be taken as an indication that “Ubuntu” and “Ubuntu Desktop” are starting to stagnate or decrease in popularity?
I wouldn’t even attempt to make that argument, because I understand there is no self-consistent consensus interpretation of Google trends data that I can reasonable support when questioned that applies widely across different search terms in a way that is testable. Just because Google Trends appears to support an idea that I’d love to spread as a meme, doesn’t make it credible and worth repeating publicly.
-jef
Thierry, your approach is just a ridiculous try to give more importance to the so called Ubuntu server edition which is not worth it’s name!
Go an ask Canonical how many paid service contracts they sold and compare that number with established proven companies like Redhat and Suse 😛
How much I would like to compare, RedHat and SuSE do not disclose such numbers. Also remember that with RedHat and SuSE you are forced to get a support contract to get security updates, while with Ubuntu Server they are freely accessible, which certainly reduces the number of “support contracts” you end up selling.
Another data point that shows that for certain workloads Ubuntu Server seems relevant: http://nicolas.barcet.com/drupal/eclipse-survey-ubuntu-2nd
centos is powerfull for server attitude