Message boards :
Number crunching :
Why Red Hat?
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
I was looking at the /stats/ directory last night and noticed something odd. Apache reports that CERN uses Red Hat for this server. The reason that is odd is that CERN and Fermilab together support Scientific Linux - a distro that aims to be functionally equivalent to RH but without the fees. So, if CERN are already putting in the effort to support Scientific Linux, why are they paying fees to RH? There is even a CERN based sub-distro called Scientific Linux (SLC3) One answer may be that this is in preparation for the move to QMC, who may prefer to be paying fees in return for the live support RH give (which I am told is excellent, but should be for the money). Edit: Another answer may be that CERN have stopped supporting SL - the SLC3 website seems not to have been updated for a while. SL website + SLC3 website + SL Wikipedia entry River~~ ![]() |
Send message Joined: 13 Jul 05 Posts: 170 Credit: 15,020,549 RAC: 39 ![]() ![]() |
I was looking at the /stats/ directory last night and noticed something odd. High-Energy Physics involves a lot of people, and even more machines (think "1000 node compute farms") that need to run specialised software, and some of the code like Geant4 and ROOT also gets used in other fields. There aren't the resources to port and debug on every possible platform, so the drive is towards a common one. Windows is expensive and seen as unsuitable for heavy-duty number crunching; IIRC there was an attempt at a deal with Red Hat for a HEP-wide (LHC-collaborator-wide?) RHEL licence but it fell through. So basically what's been done is to re-compile selected parts of the RHEL source and bundle it up with some other bits deemed "useful". So, if CERN are already putting in the effort to support Scientific Linux, why are they paying fees to RH? There is even a CERN based sub-distro called Scientific Linux (SLC3) I don't know; I can suggest three reasons: - Some commercial software (e.g. Oracle) may not be supported on SLC. Most of the critical databases seem to be on Oracle; don't know about LHC@home. Having paid the fees they may want to use RHEL anyway for services that are deemed "mission-critical" or at high risk of being hacked (not sure where we fit into that, either). (and they may not be paying Red Hat for long!) - Paying for Red Hat gets them expert support which is useful in understanding the innards of SLC - Giving them cash simply stops Red Hat getting too miffed about SLC... Another answer may be that CERN have stopped supporting SL - the SLC3 website seems not to have been updated for a while. SLC 3.0.8 came out a couple of weeks back - that would be about the only update that page would need since last year. They may be concentrating on SLC4... I suggest normal people AVOID the CERN-specific version - it's designed for use inside the lab, so the install leaves you with lots of configuration issues to tidy up before you can print to your printer instead of somewhere near Geneva! Play with the standard SL - less brain pain! Thanks Henry SL website + SL Wikipedia entry |
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
[quote]... The self adopted mission of the SL team is that anything that runs on RHEL will also run on SL. When it comes to updates, there is a lag of a few days while SL test the updates, this is bad for major security holes, but good for packages that are HEP specific, as the SL folk will bias their testing that way and thus sometimes HEP packages will run on an SL update *before* they run on the updated RHEL. I suggest normal people AVOID the CERN-specific version - it's designed for use inside the lab ... Play with the standard SL Sounds like good advice, and it explains the SL use of the word "site" for downstream customizations. R~~ |
©2025 CERN