Linux on a Scientific Workhorse
I'm building a workstation for scientific codes, utilizing the new Intel Xeon Nehalem processors. I won't list full system specs, because this isn't really a check my build type thread. However, if you're interested, they are listed here:
Newegg.com - Once You Know, You Newegg
A couple linux specific questions for building this machine:
1.) I'm trying to work out an optimal HD configuration to both protect the simulations that I'm doing, as well as prevent hangups to the system while simulations may be dumping large amounts of data in the background. So my current thinking is this:
2x ocz vertex 30 GB SSD (raid 0) - will store OS and all programs
1x Velociraptor 300 GB HD (/home and /swap, as well as some program writes, if speed is required)
2x 1TB Maxtor Caviar Black (raid 1) - safe data storage for simulations
Is this a good configuration? It's a little bit on the expensive side, and I don't know if the Velociraptor part is overly necessary, but I might be overly worried about the problems with frequent writes to ssd's)
So Question 1 is Is having /swp on a ssd a bad thing? I will have 12 GB of RAM on the computer, so usage of swp will be rare, unless one of the simulations decides to go psycho. Question 2: What filesystem should I be running on the ssd's in order to utilize their full capabilities?
2.) Are there any obvious linux problems with the system as constructed? Also, what variant of linux should I be using which can both maximize the power of this hardware, and also maintain system stability (this system must have some pretty consistent uptime, since some of the codes take multiple days to run)
Thanks for your help,
Best Linux Distribution for my Needs
I've worked with linux for a couple years on machines I work with, but haven't had it installed on my own machine in some time.
I'm currently building a scientific workhorse. It will be using cutting edge hardware (Dual Xeon Nehalem e5520, 12 GB RAM, a SSD raid 0 array, as well as a velociraptor HD and raid 1 backup array).
The two most important issues to me are:
1.) Stability - Some of my simulations take up to a week to run. If the system needs a reboot every other week, then that kills off a large percentage of my simulations, and sets me back a lot. I think a reboot every other month or so would be reasonable, as long as these were planned reboots and not crashes.
2.) Ease of configuration - I want to be able to set up all this hardware without too much personal tinkering of the settings. I'm experienced with linux, but I don't find it too much fun poking around configuration settings in order to get everything to work. Also, whenever you start doing that, something inevitably doesn't work quite right. Thus, the distribution should be made to correctly handle SSD disks, raid arrays, dual processors, 64 bit OS, etc.)
3.) Speed - Not too many background processes etc. Especially want to limit hd writes to the ssd disks, as that tends to clog them and lead to a huge speed deficit over time. No linux distribution should actually be clogging a system such as this, so I don't see it as a huge deal.
I've currently been looking at Red Hat, OpenSUSE, Gentoo, and the Scientific Linux distribution....but I don't know much about the advantages and problems with each.
Thanks for your help,