-E-mail Our Group
Initiation to ATLAS Software
1) register at CERN and ATLAS by contacting the ATLAS Secretariat at firstname.lastname@example.org (name, institution, contact = Steve Errede, Tony Liss, Debbie Errede, purpose = register at CERN/ATLAS and get an lxplus.cern.ch account) This can take a week, so don't wait. Also, you will get a NICE/MAIL account (on your lxplus CERN account) which, yes, requires a password. See the link below for forms. (You don't have to do the safety courses for registration and computer access.)
2) get a DOE grid certificate. go to : DOE GRID Certificate Webpage . One certificate is good for all of the GRIDs, but you must be "authenticated" by CERN before accessing their GRID. (This is different than being registered at CERN/ATLAS.) More information will be added later.
3) get SSH keys:
Access to CERN (Tier 0):
Access to Brookhaven (Tier 1): "Access to all Tier 1 US Atlas Facility computing resources is made through the US Atlas Ssh Gateways, atlasgw.bnl.gov, using your Kerberos 5 password. You will need an Ssh program that uses Version 2 of the Ssh protocol to get interactive access to the gateways."
Access to UChicago, UIndiana, (Tier 2) facilities:
4) HELP : email@example.com, firstname.lastname@example.org, http://consult.cern.ch/ and the pertinent HELP page from the Atlas Workbook.
5) How to join an ATLAS VO. The first link should be accessible by everyone. The second link is only accessible once you have a GRID certificate. (Tell me if I am wrong about the later.)
6) Read The ATLAS Workbook and try the tutorials.
7) Another source of information is located at the agenda for the LBL ATLAS Software Workshop 23-27 October, 2006.
8) An ATLAS News Letter, ATLAS eNews, is informal with much interesting information.
E.g., web access to some nice historical physics lectures by Rolf Landua
(Landua is principal investigator for the ATHENA experiment at CERN on properties comparison between hydrogen and anti-hydrogen)
9) The ATLAS Experiment homepage at CERN.
10) Are you a Fortran afficionado and a C++ cucumber? Paul Kunz (Stanford University) has the answer. Check this out: C++ for old fortran physicists
11) A recent review paper LHC Primer on hard interactions of quarks and gluons may be useful for understanding jets for TileCal people.
12) The ATLAS software release 12.0.6 is installed centrally and can be run without installing your own kit by doing the following on LX5 (standard linux 3):
Use the Bourne-style shells only. The C-style shells have problems associated with $PATH length.
( to switch to another shell type "> ypchsh " and the appropriate responses)
(in principle the following commands can be put in your .bash_profile so that they are executed every time you log in)
source /home/atlas/grid/app/atlas_app/atlas_rel/12.0.6/cmtsite/setup.sh -tag=AtlasOffline,12.0.6,gcc323,32
kinit -4 yourusername@CERN.CH
Running AthenaHelloWorld for example:
then, for example, going to your own ~/testarea/12.0.6 area issue the command
> cmt -co -r UserAnalysis-00-09-10 PhysicsAnalysis/AnalysisCommon/UserAnalysis
> cd PhysicsAnalysis/AnalysisCommon/UserAnalysis/cmt
> source setup.sh
> cd ../run
> get_files HelloWorldOptions.py
and then run Athena by issuing the command
> athena.py HelloWorldOptions.py
(see the Atlas Workbook online for more information on AthenaHelloWorld)
** This information is not complete, and will be continually updated and corrected.
webpage managed by Deborah Errede: email@example.com (last updated: 12 April 2007)
This is the explanation for the approach we are taking to start learning Atlas Software. Presently the software packages are setup to be brought over from CERN (a Tier 0 site), BNL (a Tier 1 site) or the Tier 2 sites (example: UChicago, UIndiana, ...) as a "software release package" per individual account...that is, each user copies the package (called "kits") from one of these sites to their own account. In order to copy these kits the user needs to be recognized by the Atlas experiment. This requires registration at CERN+ATLAS. When registering you also should get a CERN linux account (lxplus.cern.ch) and then a secure SSH key for sftp-ing files and ssh-ing sessions back and forth between these accounts, for example. Whatever works, works best from and at CERN, because the packages originate there. The requirements changed recently due to the major BNL security upgrade, so I believe getting registered at BNL as a guest and getting a linux account (gateway machine: atlasgw.bnl.gov) there is also necessary (+ secure ssh-key, etc.) in order to do anything as a member of USATLAS, but I will have to check on that to be sure. Just so you know, I have an account at CERN, SLAC and Brookhaven (smile). USATLAS refers, at some level, to sites in the U.S. so I suspect that copying a kit from UIndiana (e.g.) requires the Brookhaven guest status plus BNL account. I would like someone to test that theory for me sometime. THANKS!
Why do we need to access CERN and Brookhaven? Why not just do everything here at UIUC? Each MC event takes about 1/2 hour to generate fully. (Yea, ouch.) People may look into Fast MC's later...maybe they can get it down to 3 minutes per event. The ESD data sets (Event Summary Data ) are nn kb/event, and the AOD (Analysis Object Data) data sets are ~1 kb/event. A million events means nn Gb per 1 million events and 1 Gb per 1 million events respectively. The conclusion is that we are going to use the GRID to produce root ntuples on large data sets and for producing MC data. Testing code on small data sets at our home institution is reasonable.
So, now you understand that you need a DOE GRID certificate in order to work on the OSG or the LCG (US grid and CERN grid respectively). Don't wait. Here is an email from CERN regarding GRIDs.
The certificate is the same for all the Grids, you may access both US- GRID (OSG) and LCG with the same certificate. You only need to have the correct membership in the ATLAS VO, for both grids (i.e. choosing /atlas, /atlas/lcg1 and /atlas/usatlas in the registration pages, when asked).
Yes, this means that getting a GRID certificate is not sufficient to access the GRID. You must have membership in the ATLAS VO (Virtual Organization) also. Since I haven't completed all of this yet, I can't tell you everything you want to know about all of this.
The system for getting information on how to do all of these things is not half bad. The Atlas-Secretariat is an excellent source of help, and they quickly redirect you to the CERN help-desk (firstname.lastname@example.org) for questions they can't answer.
You may have noticed that you will be having many, many, many passwords, pass-phrases, software keys, etc. You may have to write them down and lock them away somewhere easily accessible to you. A program exists for encrypting passwords and then you can download them onto a memory stick. That program is accessible at ://nsd.dyndns.org/pwsafe/ . I haven't done that myself yet, but a colleague of ours does, and this works. Eventually ATLAS is going to have encrypted cards in order to log into all accounts.
I highly recommend reading the ATLAS Workbook. It is specific to CERN, which is another reason why having a CERN account is useful. See 6) above.
This is just the explanation for the approach that I recommend. The "how" will eventually be listed above. Much of this is available in the Atlas Workbook.