Scientific Linux 6.0 released [LWN.net]

Scientific Linux 6.0 released

Posted Mar 4, 2011 9:48 UTC (Fri) by lkundrak (subscriber, #43452)
In reply to: Scientific Linux 6.0 released by Np237
Parent article: Scientific Linux 6.0 released

It is actually compatible with Red Hat in any aspect we can imagine.

Also, I'd say quality plays important role here. (Disclaimer: I participate in Fedora, which happens to be upstream for Red Hat's distro, thus may be biased). Apart from usual (and for some questionable) argument about superior sanity of RPM and readily available packages from Fedora project, Red Hat devotes great amount of resources to polishing the distribution. My experience shows that however talented and capable Debian contributors are, Red Hat enjoys stronger involvement in development of most of key components of their distributions and thus its developers maintain superior technical insights.

to post comments

Scientific Linux 6.0 released

Posted Mar 4, 2011 10:07 UTC (Fri) by Np237 (guest, #69585) [Link] (11 responses)

For scientific applications? You must be joking.

I might be biased too since Im a Debian developer too, but as the technical manager for 1000 scientific workstations, I can assure you that Red Hat is not a viable option for scientific usage. Its just missing too many components, and it is too hard to derive things from it.

Scientific Linux 6.0 released

Posted Mar 4, 2011 10:40 UTC (Fri) by rahulsundaram (subscriber, #21946) [Link] (1 responses)

"I can assure you that Red Hat is not a viable option for scientific usage. Its just missing too many components, and it is too hard to derive things from it."

Such assurance from a Debian developer isn't as useful as having customers who deploy RHEL for such purposes as well as the presence of Scientific Linux itself. Perhaps "scientific usage" is a broad category to be assuring anything about.

Scientific Linux 6.0 released

Posted Mar 4, 2011 11:58 UTC (Fri) by ewan (guest, #5533) [Link]

Scientific Linux could moderately well be called Particle Physics Linux, since that's it's main application. In PP most application code is custom written, not off the shelf, and an awful lot of it is badly custom written at that, so porting things across operating systems is a major pain. What's required is a common base, freely available to everyone, with long, predictable, and overlapping support lifetimes. A RHEL rebuild is really the only way to go.

Scientific Linux 6.0 released

Posted Mar 4, 2011 12:51 UTC (Fri) by lkundrak (subscriber, #43452) [Link]

Well, I sort of suspected that I did not pick the right term. Substitute tat with "for various use by organizations which are among other things concerned with scientific research". Apart from platform for scientific applications which are not present in distro it includes common activities, such as word processing, web browsing or sharing files.

Scientific Linux 6.0 released

Posted Mar 4, 2011 16:28 UTC (Fri) by southey (guest, #9466) [Link] (7 responses)

You are welcome to have your definition of scientific usage but don't force others to have the same definition!

Yet many people like myself are successfully using Red Hat, Fedora and related distros on scientific workstations. Personally I prefer missing components because distros usually don't have the latest version (Debian experimental is not an option) or that I want to know what I have installed rather than a huge bunch of unused dependencies just waiting to cause havoc. Also some scientific packages (blas/lapack) have a history of being broken or incomplete so it is hard to automatically trust some distros.

Scientific Linux 6.0 released

Posted Mar 4, 2011 16:31 UTC (Fri) by Np237 (guest, #69585) [Link]

If its broken, just fix it. You will avoid doing the same work over and over again, each time you want to upgrade this or that component.

Scientific Linux 6.0 released

Posted Mar 4, 2011 17:11 UTC (Fri) by donbarry (guest, #10485) [Link] (5 responses)

Odd to hear how the *lack* of components is spun as a benefit.

I am an astrophysicist, and also the computing architect for
a NASA spacecraft instrument team. I migrated a team from
Sun workstations to Redhat 7.3 eleven years ago, and stayed
on the Redhat treadmill to RH9, while slowly learning the many
good things about the Debian system. With the mischief and
misdirection that followed RH9, I professionally migrated to Debian.
I've been there ever since.

Debian has a release cadence that is reasonable for scientific
teams, and it packages *far* more software than Redhat. Although
both systems make it easy to maintain customized repositories
of super-versioned software, the fact that much of it is already
in Debian means that one doesn't need to reinvent most of the
packaging decisions.

Furthermore, one can trust, with full integrity, that the underlying
distribution will not jerk you around in pursuit of enhancing their
revenue stream. Redhat has done that several times in the last dozen
years. With a new management team and an overvalued stock, they will
do it again in the next few years (see the recent kernel patch
distribution discussion). You have been warned.

Debian's decision system comes from -- the community. A *real* community, not just one assembled by corporate fiat, or temporarily bribed into existence (a la Shuttleworth). I trust that Debian's integrity, the heft of its package choice, and its direction will remain constant for another dozen years.

The "Scientific Linux" system is an impressive effort from Fermilab, but the decisions as to its architecture base, inevitably, were made by bureaucrats, not scientists. That's simply the way the large national laboratories work, when you seek internal funding to support multiple personnel for a multi-year effort. Of course, scientific guidance would have been sought, and given, but inevitably compromises between scientific interests and bureaucratic interests emerge. It is not at all surprising that RHEL was chosen as a base, given the natural tensions. A scientist
would advise: "RHEL is acceptable as a platform, if possibly not ideal, given the limited packaging of scientific software. Debian is better."
A bureaucrat would rationalize: "I can defend the RHEL choice based on
the substantial and growing revenues of Redhat and its contractual
adoption by major industrial figures. If SL fails and we had chosen
Debian, others may characterize it as putting the prestige of a
national laboratory atop a volunteer, hobbyist effort, and I will lose
face. My interests are served by the commercial effort, which
partially absolves me of risk."

Scientific Linux 6.0 released

Posted Mar 4, 2011 18:29 UTC (Fri) by kmccarty (subscriber, #12085) [Link] (2 responses)

It was actually suggested at one point (by a Debian user) on the SL devel mailing list that Scientific Linux rebase themselves on Debian:

http://listserv.fnal.gov/scripts/wa.exe?A2=ind0703&L=...

Not too surprisingly, there weren't any replies.

Scientific Linux 6.0 released

Posted Mar 4, 2011 19:08 UTC (Fri) by donbarry (guest, #10485) [Link] (1 responses)

I suspect that comes from two factors:

1) Not many scientists outside of subdisciplines of physics use
"Scientific Linux". I do not know a single astrophysicist who
does. I'm sure there are many users at a few particle physics
labs, probably dictated by management.

2) Those who know the inertia of bureaucracies figure they have better
fish to fry, and as long as someone else is packaging the software
they need, would prefer to tilt their lances elsewhere.

That said, I think that the "Scientific Linux" packagers would be
very well served by rebasing their efforts onto Debian.

Scientific Linux 6.0 released

Posted Mar 7, 2011 17:36 UTC (Mon) by kmccarty (subscriber, #12085) [Link]

Coincidentally, mention of Scientific Linux just came up on the debian-science mailing list. Some highlights:

http://lists.debian.org/debian-science/2011/03/msg00006.html
http://lists.debian.org/debian-science/2011/03/msg00007.html
http://lists.debian.org/debian-science/2011/03/msg00012.html

- Kevin

Scientific Linux 6.0 released

Posted Mar 5, 2011 19:19 UTC (Sat) by Np237 (guest, #69585) [Link]

We went precisely through this in my company. It was very hard to get Debian accepted while the company already had Red Hat as the primary OS for Unix servers. The first attempts were made with Red Hat, with mixed results. Those factors made Debian win:
- A number of motivated people explaining over the benefits to management over the years.
- Money money money; we can do with a team of 10 what a team of 20 would not be able to do with another OS.
- The universal operating system: we can use the *exact* same OS, only with differences in installed packages, for scientific workstations, computing servers, and HPC clusters. Binary compatibility between them saves a lot of money too.

Scientific Linux 6.0 released

Posted Mar 10, 2011 5:41 UTC (Thu) by dowdle (subscriber, #659) [Link]

Let's see... you haven't used RHEL at all... because you quit using RHL about 8 years ago and you are still complaining about what happened back then. You are positive that Red Hat sucks and will continue to suck... and offer a warning... that it will never be good enough because it has commercial interests... although it just happens to be the shining beacon of free software and the leading example of a financially successful commercial Linux provider... who has done a rather good job at being serious about and sticking to its FOSS principles... buying up several commercial companies/products and releasing them as open source... and being a significant contributor for years to the kernel, gnome, gcc, xorg, etc. As far as I can see, it is impossible for any commercial Linux to be worthy in your eyes... no matter what.

Notice how I did not denigrate Debian in any way as I do not feel the need to. I have a lot of respect for its developers and its users... and I certainly appreciate the competition it provides with its good work.

I have participated in many "distro wars" type conversations over the years but I'm beginning to grow weary of them. There are many use cases for Linux and no single distro that meets everyones needs.

Red Hat doesn't have a smaller package count than Debian by accident. As you may or may not recall, they intentionally reduced their package count going from RHL to RHEL in order to make it easier to support in a commercial environment. As you are surely aware, they support their releases for a long time... 7 years at least. If I remember correctly, Debian supports their releases for 1 year after the next release is made available. That is a significant difference that may or may not be important depending on your needs.

Please refrain from telling everyone that if they pick Red Hat over Debian that they are doing so because they are forced to by bureaucratic managers. While that might be the case in some instances, it certainly isn't in all.

Also be advised that your suggestion that Debian has the only true community... is offensive to me... and probably others. Fedora has a community that is sponsored by Red Hat. They do a lot of work. Like Debian's community, Fedora's is often consumed with lively debate and disagreements... an actual sign of a healthy and honest community. Red Hat has a community too... and they include users like myself... the CentOS community... and the Scientific Linux community... among others.

While I appreciate your commentary and the thought you put into your comment(s), I think you should be a little more open minded. Debian is a great distribution and great bunch of packagers, but you yourself benefit by a lot of the development contributions done by Red Hat even if you don't choose to use Red Hat's packaging of them.

Scientific Linux 6.0 released

Posted Mar 4, 2011 10:44 UTC (Fri) by malor (guest, #2973) [Link] (8 responses)

If you mean quality as 'stability', Debian is probably the best Linux distro going. They tend to be a little behind the curve, because they release slowly, but their releases are exceedingly solid. You typically need very few patches on the stable distro, even late in the 2-or-so year release cycle.

Now, there's lots of other reasons you might like Red Hat; if you have a lot invested in the RPM infrastructure, or if all your institutional expertise and code is in their filesystem layout and configuration, the pain of switching might not be worthwhile.

But it strikes me that you're doing an awful lot of work, each time you need to strip all the proprietary Red Hat bits out of their distro to distribute it freely. You wouldn't have to do any of that if you were based on Debian... just take the base OS, add whatever packages you want, and you're done. Further, if you get involved with the distro itself, you can probably get your preferred packages built right into the original archives, rather than having to maintain your own, separate infrastructure. I show twenty-eight thousand packages in their repos. In exchange for the help in tweaking what's already in there, I suspect they'd be happy to carry whatever extras you might need. Once you had the packages upstreamed, it'd become an issue of ongoing maintenance over the long release cycle, rather than a hurricane of work all of once.

Further, because your infrastructure should then be compatible, you should very easily be able to roll out Ubuntu on desktops, probably with the exact same packages, because Canonical will probably pick them up, too. I think the combination of Debian on servers, where stability is usually paramount, and Ubuntu on desktops (which gets an ongoing blizzard of updates, because they release whether it's ready or not), makes a heck of a lot of sense. Low maintenance burden on the really important stuff, higher burden on non-critical machines that need to be friendly. I just tried out the most recent Fedora, experimenting with using it as a KVM host, and was appalled at how poor the overall usability is in comparison with Ubuntu. It didn't last long on my test machine.

It just seems to me that the distro's entire focus would be much better suited to cooperating with the scientific community; their purpose is not to make money, but rather to make a great operating system, which strikes me as far more closely aligned with your goals than are Red Hat's. I'd argue, in fact, that this difference in focus would probably be more important to you, over the long term, than any theoretical benefits of RPM.

Just a thought.

Scientific Linux 6.0 released

Posted Mar 4, 2011 12:59 UTC (Fri) by lkundrak (subscriber, #43452) [Link] (3 responses)

> But it strikes me that you're doing an awful lot of work.

Not true, look at the release notes.

Red Hat splits most of proprietary artwork in separate packages which are easy to replace (you can pick unbranded replacements directly from Fedora if you wish). The changes they've done were rather minor -- look at the release notes.

The rest of your reply is sort of hard to argue with beyond what was already said, we'd be just repeating ourselves. For me (and trust me, I run ubuntu and debian as well, mostly to gain input about status and improvements done to them), I find el6 to be by far the best pick when it came to what I care about (sane configuration, ease of packaging, hardware support, stability and polish).

Scientific Linux 6.0 released

Posted Mar 4, 2011 13:05 UTC (Fri) by malor (guest, #2973) [Link] (2 responses)

Huh, from the CentOS threads, I thought it must be a pile of work. Is CentOS just dead on the vine?

Scientific Linux 6.0 released

Posted Mar 4, 2011 15:30 UTC (Fri) by ttrafford (guest, #15383) [Link] (1 responses)

The CentOS people have 5.6 to build first.

Scientific Linux 6.0 released

Posted Mar 4, 2011 16:18 UTC (Fri) by ESRI (guest, #52806) [Link]

SL also has 5.6. I think the two projects are actually fairly close schedule wise. SL has a smaller and different type of user base which isn't as prone to complaining like the CentOS user base is... and also, SL is honestly better at communication. Sharing a publicly available "rolling release" is part of just making everyone feel better that progress is being made (whether or not it's any faster or slower than CentOS's doesn't really matter).

Scientific Linux 6.0 released

Posted Mar 6, 2011 9:33 UTC (Sun) by Np237 (guest, #69585) [Link] (3 responses)

I dont understand the dichotomy you make between servers and desktops.

You need the same stability on (professional) desktops than you need on servers. Just because Ubuntu with its 6-month release cycle is popular among geeks, it doesnt make it suitable in a corporate environment. Windows doesnt change every 6 months, that is part of what makes its success in professional desktop usage.

Scientific Linux 6.0 released

Posted Mar 6, 2011 14:02 UTC (Sun) by malor (guest, #2973) [Link] (1 responses)

I think of that as the price to pay for the overall level of user-interface polish and easy usability of nearly all hardware. If your particular environment is as intolerant of change on desktops as on servers, and you also want a nicely polished desktop UI, I don't think there are any Linux distros that suit at all well.

Ubuntu LTS is probably the closest match, but with that set of requirements, I think you'd probably be better off just running Windows in most environments.

Scientific Linux 6.0 released

Posted Mar 6, 2011 16:43 UTC (Sun) by Np237 (guest, #69585) [Link]

Our experience with Debian is quite nice. It is only lacking on hardware support, which is the easiest of all these requirements to catch up with by ourselves.

Windows is a no-go in a scientific computing environment.

Scientific Linux 6.0 released

Posted Mar 7, 2011 16:57 UTC (Mon) by pboddie (guest, #50784) [Link]

You need the same stability on (professional) desktops than you need on servers. Just because Ubuntu with its 6-month release cycle is popular among geeks, it doesnt make it suitable in a corporate environment.

Yes, but you can get support on Debian and Ubuntu for longer than six months, and you don't have to upgrade when the next release comes out, although I have yet to be convinced that Ubuntu (or at least Kubuntu) can be trivially upgraded at the end of a long-term support release cycle to the next LTS version.

Windows doesnt change every 6 months, that is part of what makes its success in professional desktop usage.

This sounds like the argument about "old being good", which leaves users with ancient versions of popular desktop environments and having to manually install software from source because the distribution only packages ancient versions of applications. That's a solution which is only good for administrators who just want to push something out there and forget about it (and whether the users actually like it or not).

And yes, when I had to use Windows in a workplace, even being able to use a recent version was bad enough. So you can imagine how much fun it was to use an "old but good" version.