Quote:
Originally Posted by kebabbert
(Post 6191989)
I want to be able to install new software without fear of breaking things. That does sound like fragile to me: "only run these software that we have accepted, if you run other software then your environment might be stable". Not robust at all. Imagine if you only could run the software that Microsoft accepted - people would say that Windows is fragile.
|
If you mix lots of bloatware in the windows machine you will have bad surprises too.
Critical professional software needs always a more controlled IT environment/system to run safely. That's a rule for any IT system.
You can't install "anything" in a system without fear. Don't forget what "malware" is, or security holes. As an administrator you can do almost anything you want, even things that are dangerous for the stability and security of your system. This means that you have to be always careful otherwise no o.s. or antivirus s/w can save your system.
The more fancy and "easy to use" or "easy to administrate" the system, the more vulnerable is, for me. And I think it is true for security as well as for stability. "Robust" systems are usually more restrictive and less fancy. Critical systems are always restrictive. A server that has to work 24/7 for months or a production workstation can be a critical system for the business. A desktop computer that someone uses at home for news surfing, youtube and facebook isn't really that critical from an admin's or developer's point of view.
Quote:
Originally Posted by kebabbert
(Post 6191989)
As a user you need to be able to do whatever your work requires. Computational fluid dynamics, MCMC, gaming, word processing etc. You choose, and the OS should be stable enough to allow you to work with whatever your work is.
|
Those software categories belong to very specific/scientific software, out of the "general" user's software like the remaining. So, let me share my experience on this domain.
Most of the times, the developer tries to get the best for the calculations to be performed (usually a very big computational load), to solve the problem within the available RAM, to shrink as much as possible the execution time that can be from hours to days, etc. Factors like running under every o.s. version or any distro etc are often left behind. Very often, older language versions and/or libraries are used because they are already available and well tested. Hardware requirements can be very different than, say, a game.
My company bought an expensive commercial software (domain of the marine engineering), and we had to make decisions on the hardware specification. I did suggest that we need to know if the software makes use of a performant graphics card for the calculations. The man from the vendor/support confirmed (orally) that it makes use of the graphics card, and then my company bought a mswindows machine 32GB RAM, Xeon (or perhaps i9) 16 thread (=8 core) with a very expensive graphics adapter. However, running times seemed to be very long. I did inspect the system's activity and found: RAM at less than 50%, CPU at 115% for all cores, GPU activity zero! This meant that the software didn't use the graphics adapter, at least for its principal heavy task, and lots of money went to an unnecessary device instead of, say, another faster CPU! and that "support" staff was non competent at all.
By the way, looking in the user's forum I found many users that confirmed that the software was running faster under linux than under windows (it was originally a Unix software around 1990-92, at the time there was no windows, no linux, just unix or vms for serious applications).
Very often developers specify the linux distributions and versions that they did test with their software. It seems that distributions like Red Hat Enterprise or Suse or CentOS are well respected. Nevertheless I found forum reports about finite element computational software running under Ubuntu too. You can try to get info from the developers or users of those s/w packages on specific o.s. requirements/details.
To summarize, I think it would be safer to not "mix" heavy computational software with internet, gaming and multimedia user's software. Perhaps 2 separate VMs with separate OSs (or 2 separate computers) could solve the issue.