CSA07
From UF HPC Wiki
CSA07@HPC Information
The CSA07 Chowder samples have been transferred to the HPC. See Ronny's email below for information:
Chowder Announcement
To all,
The transfer of all officially produced CSA07-Chowder samples from Fermilab --> Tier2 --> HPC has just reached completion this morning. For those of you who are not aware, these samples consist of t-tbar, W + Jets, and Z + Jets events made with ALPGEN. These samples are obviously quite essential to anyone studying backgrounds for SUSY or Higgs related phenomena. Thus, speaking for Yuriy, Alexey, Mike, and myself, these samples will be used extensively. Certainly, others in the group who are doing different analyses (Z-prime, SM measurements, etc) are welcome and strongly encouraged to make good use of these valuable samples as well. We hope to be able to store and use these samples until the 2 or 3 months preceding the arrival of real LHC data, the e.t.a. of which is still uncertain, but hopefully will be towards the end of this year.
The transfer process was initiated in mid March by Yuriy, Bockjoo, and myself. Yuriy and Bockjoo had the samples transferred sequentially in 6 sets consisting of roughly 2600 root files (~10 TB) each to our Tier2 from Fermilab. Once, one set began to arrive at our T2, I began the transfer process to the HPC using a very effective shell script developed by Bockjoo. Most root files, which are approximately 3.0 GB, transferred in 90 seconds or less. Approximately, <1% of the root files failed to transfer within a ten minute time limit imposed by the shell script. These, were usually transferred successfully within a second or third execution of the script (each time relaxing the time limit by a few minutes). Ultimately, this process took about 7 weeks to complete. This includes the time that it took to debug the shell script, re-execute the script, make sure files existed on our T2, etc.
In the future, I think we can expect to reduce this transfer rate significantly, perhaps even by a factor of 1/2, with just a few modifications to the script. I have attached a text file that contains the timestamp of the first 60 root files that were transferred from the third set of blocks. From this you can get an idea of how many files can be transferred within a certain amount of time. My estimation yields roughly 38 root files/ hour which is ~114GB/hr. At this rate, one could presumably "aspire" to transfer 60 TB in 526 hrs or 22 days. This assumes the ideal scenario with a continuous transfer of one file after the other (i.e. no down time on the T2/HPC and an unencumbered grad student to babysit the process). Again, this is a naive estimation - perhaps there are other details to consider, which may expedite/slow the process.
In total we currently have 27 Million Events (~61 TB) located in the following directories on HPC:
3.6T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0003 3.3T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0002 3.2T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0001 3.9T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0011 3.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0012 3.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0013 3.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0010 3.8T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0006 4.5T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0007 3.4T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0008 3.5T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0009 3.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0004 2.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0000 3.7T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0005 3.6T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0014 3.9T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0015 2.5T /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0016 155G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0024 48G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0025 36G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0026 7.4G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0028 21G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0029 51G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0017 38G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0018 35G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0019 23G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0020 35G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0021 62G /scratch/crn/remington/store/CSA07/2007/11/29/CSA07-CSA07AllEvents-Tier0-A3-Chowder/0023
Chowder File Breakdown
A quick note about the Chowder file breakdown. There are 16,500 files, each of which contain a separate process (eg., W + 0 jets). Because some of these processes are a lot more rare than others (eg., t-tbar + 4 jets), in order to remain unbiased, you must run over many thousands of files (with more than 1k event/file) to ensure that some of these more rare processes are included in an analysis. As such, I've placed the file breakdowns, ordered by process id, here:
Each file contains the physical filepath of the specific processes as currently located on the HPC disk. The numbering scheme can be parsed by looking at the table here:
Note that my automatic process-parsing failed (for whatever reason) for a few of the files, and for now I put these in a separate txt file.
