DAF Home Page

What is the DAF?

The DAF is the "Drell-Alexander Farm" of computing facilities, named for its original and subsequent benefactors (substantial sums from the LEPP computing budget have also been made for the DAF). The DAF is for the use of the Cornell CLEO group only, except in the case of explicit exceptions made in consultation with a Cornell faculty member.

The DAF currently consists of the following Linux nodes: for CPU use, lnx187 and lnx189; hosting a 3TB RAID disk, lnx112. (For CMS work, please try lnx235 first.)

Scope of this page

This page is only meant to act as a guide for usage of the DAF, not as a guide to CLEO computing in general.

Questions?

Questions of general policy and practice should be directed to a Cornell faculty member and/or Brian Heltsley (bkh@lns.cornell.edu). Questions of a technical nature should be directed to Ray Ng (crn@mail.lns.cornell.edu) or "service" (service@mail.lepp.cornell.edu).

General Policy

Like most computing facilities, the DAF is a shared and limited resource. Please cooperate with your colleagues by using it carefully. Automated features which facilitate fair allocation of resources can always be defeated by ingenuity or ignorance. Think before you compute!

Usage Recommendations

• Compilations and linking are still more efficiently done elsewhere because the related library disks are mounted closer to those nodes: compile and link elsewhere!
• Jobs of more than a few minutes duration should generally be submitted to the batch queues; if you must run for longer, reduce the priority of long interactive jobs with nice (as with nice +10 myjob).
• Batch jobs (see below) should, where possible, not be excessively long; break up your job into smaller pieces if it's not too much of a hardship. Here "long" is defined as 8 hours. The time limits are 12 hours cpu time, 24 hours wall clock time. Monte Carlo jobs should be run in much shorter pieces.
• Jobs submitted to the general laboratory Linux batch queues will automatically allow Cornell group members' jobs to run on DAF nodes.
• Disk space: /nfs/cor, (approx 3TB total) is partitioned into four sections: user, temp, an1, and an2 (see more on these below). As a Cornell group member you should find a directory under your username in each area (you can make your own in temp). Please strive to store the files you create and use under a directory with your username; otherwise, bookkeeping gets difficult. In particular, once someone leaves the group, the files owned by that individual are free to be removed after 6 months absence. Please find a long-term home for your files when you leave!

• The user partition (500GB) allocates a fixed, enforced quota, set at 25GB as of this writing, for each user. Write privileges will cease when the quota is exceeded. Files in this directory are backed up automatically by the Computer Group, BUT ONLY ONCE PER MONTH! Archive your own files if you are concerned about losing important files. Please manage your space wisely!

• The temp partition (300GB) is for temporary staging of large files. Be warned that an auto-delete program runs on this disk when it gets above a "high water mark" of 80% full. It then looks to remove files, oldest first, until usage gets below 70%. This disk is NOT BACKED UP BY ANYONE!

• The an1 and an2 partitions (1.3TB each) are for long-term analysis use. FILES ON THESE DISKS ARE NOT BACKED UP BY ANYONE! Files are managed by the users. Please notify me if you see a disk filling up. Self-policing works best, but, if necessary, informal quotas may be imposed.