Hi Robert,
For workstations, I like openSUSE very much and I'd probably not give it
away for anything else (including Ubuntu). Although this is probably
also because I know it so well now that I can do much faster
troubleshooting than in others.
I have to say though that during my postdoc, I was not happy with Debian
.....
At the end of the day, for a workstation I believe between
openSUSE/Ubuntu/RedHat is a matter of taste ... Probably each of them
will work pretty well .... And in the end you should also remember that
the distribution is just the packaging of a Linux system .... Just pay
attention at installation defaults, these days they are made for
windows-like using (both Ubuntu and openSUSE).
For the cluster, we use CentOS, but this is probably just because out IT
Department insisted on that. .... SUSE has its own server OS but the
drawback is that its not free ....however, I guess one can use openSUSE
as well depending on features and complexity needed ....
Best
Vlad
On 05/20/2016 04:51 PM, Ross Walker wrote:
> Hi Robert,
>
> This is a bit of a religious question so take all the answers including mine with a grain of salt.
>
> In my experience you have two real choices based on what most people use:
>
> 1) Ubuntu
>
> 2) Redhat / Centos
>
> Note, redhat and centos are identical - centos is just the free unsupported version compiled from the redhat source - so if you know redhat centos is the obvious choice if you can't afford redhat. Some people also use fedora. In my opinion Ubuntu and Fedora are beyond the bleeding edge, I refer to it as haemoraging edge - if you like to have the latest and greatest everything on your desktop use Ubuntu. But for a cluster it is a ridiculously bad choice - it is unbelievably fragile, especially when you start putting in things like NVIDIA drivers etc. Having designed built and shipped over 50 clusters I would never ever ever recommend Ubuntu. The only reason I ever recommend Ubuntu is for things like deep learning development boxes where the people doing the development insist on using the latest version of everything they possibly can so Ubuntu is the only place it works. It is one complete world of pain. Pretty much all the support requests for I installed XYZ and it bricked my machi
> ne come from Ubuntu systems.
>
> So if you want to build a cluster use either Redhat 6 or 7 (Centos 6 or 7 is perfectly acceptable as well). Stay away from Ubuntu like the plague. I'd also advise you not to let your students etc use it - that way at least they won't be tempted to use tons of 'k00l' new features that are utterly unbackwards compatible etc.
>
> Note there are pre-wrapped cluster management systems as well - none based on Ubunutu. I'd highly recommend Bright (http://www.brightcomputing.com) although it isn't free or alternatively Rocks (http://www.rocksclusters.org/) which is free although updates can be a little sporadic and the learning curve is harder. For reference Rocks is what we use at SDSC to run our major supercomputers.
>
> Hope that helps.
>
> All the best
> Ross
>
>> On May 20, 2016, at 06:03, Robert Molt <rwmolt07.gmail.com> wrote:
>>
>> Good morning in the USA, good afternoon in Europe...
>>
>> I have to choose a flavor of Linux for a new cluster my research group
>> is building, intending to run Amber with GPU execution (among other
>> things). I am not used to making system admin-style choices and am
>> soliciting advice.
>>
>> I am sure Amber can be properly compiled on many forms of Linux, with
>> varying ease. I cannot get RedHat, unfortunately. Some of the free forms
>> of Linux suffer from a lack of stability (constant updates). Do Amber
>> veterans have strong opinions about a "best" overall choice in
>> Linux-style to go with for computers intended to compile Amber? I am
>> only experienced in RedHat, Ubuntu, and Mint but am happy to adapt to
>> any flavor.
>>
>> --
>> Dr. Robert Molt Jr.
>> r.molt.chemical.physics.gmail.com
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Dr. Vlad Cojocaru
Computational Structural Biology Laboratory
Department of Cell and Developmental Biology
Max Planck Institute for Molecular Biomedicine
Röntgenstrasse 20, 48149 Münster, Germany
Tel: +49-251-70365-324; Fax: +49-251-70365-399
Email: vlad.cojocaru[at]mpi-muenster.mpg.de
http://www.mpi-muenster.mpg.de/43241/cojocaru
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri May 20 2016 - 09:00:04 PDT
