Weekly Edition
Return to the Distributions page
| |||||||||||||
More on Statistics (SUSE openSUSE blog)
The SUSE openSUSE blog has an article with some in-depth statistics on the reach of the distribution. It includes various numbers and graphs on downloads, installations, the use of the Open Build Service, as well as a comparison with Fedora. "As you can see, Fedora has more downloads than openSUSE. Looking at the users, the situation is reverse: openSUSE has quite a bit more users than Fedora according to this measurement. How is this possible? The explanation is most likely that most openSUSE users upgrade with a 'zypper dup' command to the new releases, while Fedora users tend to do a fresh installation. Note that, like everybody else, we're very much aware of the deceptive nature of statistics: there is always room for mistakes in the analysis of data. To at least provide a way to detect errors and follow the commendable example set by Fedora, here are our data analysis scripts in github."
(Log in to post comments)
More on Statistics (openSUSE blog) Posted Aug 23, 2013 19:28 UTC (Fri) by jspaleta (subscriber, #50639) [Link]
Just one question, which the article sort of brushes over.
How did they make the comparison across user counting methodologies? They note the methodology is different, but I would have like to see a brief explanation of how they account for it..before I go looking at the scripts to validate their described methodology.
My understanding from the article is that Fedora counts unique IPs and they are counting specific update interactions. So either to make the comparison.. they converted their update interactions into unique IPs to compare with Fedora's numbers. Or they converted Fedora's unique IP count into the equivalent of an update interaction. I know which procedure I would use to get a clean comparison with the least number of assumptions necessary in the mapping of one raw data metric into the other, but I'd like to see a description of what they did.
-jef
More on Statistics (openSUSE blog) Posted Aug 23, 2013 20:10 UTC (Fri) by jospoortvliet (subscriber, #33164) [Link]
For the comparison we simply used the same system as Fedora used to count. As in, count unique IPs. We used to do that and made those public, just like Fedora, for years - until Coolo decided he did not want to publish them anymore as they are not terribly reliable. He's a perfectionist ;-)
Of course, they're not very reliable, in absolute terms, they over-estimate user counts by about a factor 10. But you can of course make comparisons, provided the bias both openSUSE and Fedora have is similar. Which seems not a stretch to assume...
Check out http://en.opensuse.org/openSUSE:Statistics for the old numbers and compare them to those of Fedora from that time, if you like.
The comparisons in the blog are done on more recent numbers. We could update the statistics page, not sure if we will... In the end, the numbers are meant for internal (openSUSE, that is) tracking of how we're doing and to help us make decisions.
More on Statistics (openSUSE blog) Posted Aug 23, 2013 20:26 UTC (Fri) by jspaleta (subscriber, #50639) [Link]
I would agree that until Fedora starts generating and sending back a similar UUID as part of system install, then its a reasonable assumption to make that they both distro userbased operate under similar scaling laws between the metrics.
However, I'd like to bracket the scaling factor with a reasonable error bar based on the openSUSE dataset. Do you have any existing analysis done up that shows the metric scaling factor in the existing openSUSE dataset? Does that scaling factor have some seasonal oscillation? Or is a really tightly defined number day-in and day-out over any timescale you can look over? I'm not saying its not a real thing. But I'd like to put an error bar on it if I can. I love me some error bars.
-jef
More on Statistics (openSUSE blog) Posted Aug 27, 2013 14:25 UTC (Tue) by jospoortvliet (subscriber, #33164) [Link]
I have no idea about this, I'll share this comment with the dude doing the analysis (Alberto). He might look at it a next time.
Tnx for the feedback!
More on Statistics (openSUSE blog) Posted Aug 27, 2013 14:40 UTC (Tue) by aplanas (guest, #92569) [Link]
Hi : )
I want to note that there is not scaling factor. What we did is to replicate all the computation using the Fedora methodology with our datasets. In other words, we do not convert the UUID way of counting: for the comparison we discard this result and redo all using IPs and in the way describer by Fedora's scripts.
This is the reason that we do not have errors or confidence intervals here. But could make sense for the linear regression model that we used to extrapolate the evolution of the distribution. But this model is not related with Fedora comparison.
Alberto Planas.
More on Statistics (openSUSE blog) Posted Aug 27, 2013 15:07 UTC (Tue) by jspaleta (subscriber, #50639) [Link]
Thanks, that clears that up.
So just to make sure I'm understanding the comparison section of the post correctly, when reusing the Fedora unique IP technique with the OpenSUSE data, you are getting more unique IPs for the OpenSUSE dataset than the number of unique IPs being reported for Fedora?
-jef
More on Statistics (openSUSE blog) Posted Aug 27, 2013 16:47 UTC (Tue) by jospoortvliet (subscriber, #33164) [Link]
Yup. You can verify by yourself, if you like, comparing data from the time openSUSE still recorded unique IP's, see our statistics wiki and the Fedora one. Let's say we compare openSUSE 11.0, 11.1, 11.2 and 11.3 in week 23 (the common denominator between the Fedora's):
11.0 4,391,810 11.1 3,800,252 11.2 3,871,821 11.3 3,021,753 (btw that's a scary downward trend. if you go a little further in the weeks you see it's not that bad, our users seem a bit conservative in moving over to 11.3 especially)
Fedora had the following numbers in week 23 after the releases of 10-14:
In the graphs that we have in the blog you see a similar (or even bigger) gap. If we then assume everything is similar between Fedora and openSUSE we can probably say that openSUSE has something between 20% and 40% more users than Fedora but with the differences in release cycle and other biases I wouldn't be totally surprised if the distro's are even closer or if the difference approaches a factor 3. I would find it surprising if there was an order of magnitude difference in user base size or if Fedora would have a far larger user base as openSUSE - I don't think either of those statements would be realistic.
If you look lower on the page, you see that these numbers overstate the number of unique cookies for openSUSE by a factor ~10, hence my statement that IP addresses probably are about an order of magnitude off. I suspect that is how Mark Shuttleworth came up with such huge numbers of users - many, many millions, if I recall. Just unique IP's doesn't mean that much.
Of course, the nr of cookies that connect to a server in total is less interesting than the number of cookies which does so regularly. That last is the most reliable guesstimate we have of our user base: about 440.000 users, again, half of the cumulative nr of cookies in total. That might put Fedora on 280-380K users. But all this is dangerously close to pure speculation :D
More on Statistics (openSUSE blog) Posted Aug 27, 2013 17:15 UTC (Tue) by jspaleta (subscriber, #50639) [Link]
Thanks for the details.
This comparative methodology looks sound to me and I've no reason to think the guarded conclusions you are drawing are unreasonable given the available datasets.
More on Statistics (openSUSE blog) Posted Aug 23, 2013 20:14 UTC (Fri) by jospoortvliet (subscriber, #33164) [Link]
Just to be pedantic - this is the blog of the openSUSE team at SUSE, the former 'boosters'. It's not from openSUSE, the community, but from some of SUSE's people active in it ;-)
More on Statistics (openSUSE blog) Posted Aug 23, 2013 20:19 UTC (Fri) by jake (editor, #205) [Link]
> this is the blog of the openSUSE team at SUSE, the former 'boosters'.
> It's not from openSUSE, the community, but from some of SUSE's people > active in it
well, that's kind of a mouthful ... it might not really fit in the title :)
more seriously, what should we call it? SUSE openSUSE blog? is that better?
jake
More on Statistics (openSUSE blog) Posted Aug 27, 2013 14:26 UTC (Tue) by jospoortvliet (subscriber, #33164) [Link]
Sry for the late reply. The SUSE openSUSE blog would be fine, yes ;-)
More on Statistics (openSUSE blog) Posted Aug 27, 2013 15:18 UTC (Tue) by jake (editor, #205) [Link]
> The SUSE openSUSE blog would be fine
okey, fixed ...
jake
Relative Package Availabilities Posted Aug 27, 2013 8:29 UTC (Tue) by ldo (subscriber, #40946) [Link]
I’m a fan of Blender. I managed to figure out how to build it with most of its capabilities enabled (including the Cycles renderer and OpenCollada import/export) on a Debian Unstable system. I have a wrapper script which sets up all the right CMake options; I could do a fresh build every day, but I keep it down to once a week. :)
I tried to do a similar thing on OpenSuSE 12.3, but kept hitting missing packages that I also had to download and build from source. I tried installing the provided RPM, but that is only for Blender version 2.64, and doesn’t include Cycles!
I’m wondering now, whether to switch my backup machine from OpenSuSE to Debian (perhaps Debian Stable—it is for backups, after all). Cycles can produce breathtaking results, but can take a-g-e-s to render, which makes it hard to do animations. I figure if I put together my two quad-core machines, I would have, not exactly a render farm, but at least a render windowbox. :)
Relative Package Availabilities Posted Aug 27, 2013 14:30 UTC (Tue) by jospoortvliet (subscriber, #33164) [Link]
A usecase like this is what the open build server is great for - you could get it to make nightlies for you quite easily. Just go and fork the existing Blender package (and perhaps its dependencies) and add the cycles stuff. When you're happy with the results you could even do a merge request, see if the maintainer takes the changes and then you'll have others maintaining it in the future ;-)
You can get started here:
Let the Open Build Service build the Blender packages so your system can focus on what it's supposed to do - rendering your animations, not building your packages :D
Relative Package Availabilities Posted Aug 27, 2013 19:50 UTC (Tue) by Wol (guest, #4433) [Link]
If you're going to change distros (and I'm a SUSE/Gentoo user), I'd recommend Gentoo. You can be as bleeding-edge as you like. I switched to Gentoo for the same reason you did, except the package that did it for me was lilypond.
Gentoo can be some hassle, but it's a very good teaching tool :-)
Cheers,
|
|||||||||||||