Theory Computer
Data Disk Quotas
Below are the names of the theory data disks, their useable sizes and a
list of each person's quota on that disk. All numbers are in Gbytes.
To see a snapshot of the current usage instead look at this
automatically
generated table
.
Host Machine |
Disk Name |
Disk Size |
Quota |
cdml0: |
/gal/sr1 |
910 |
Adrian Jenkins 910 |
cdml0: |
/gal/sr2 |
921 |
Tom Theuns 200 Takashi Okamoto 200 Carlos Frenk 100 Cedric
Lacey 250 Carlton Baugh 100
Darren Reed 50 Rob Crain 50 Phil Bett 50 Raul Angulo 50
Angelo Fausti Neto 60 Valeria Coenda 35 Richard Bower 80
|
cdml0: |
/gal/sr3 |
921 |
Noam Libeskind 400 Tom Theuns 300 Claudio Dalla Vecchia 100 Rowena Malbon 50 Yan-Chuan Cai 100 Gabriel Altay 50
Craig Booth 50 Juan Gonzalez 100 Adrian Jenkins 10 Andrew Cooper 50 |
cdml0: |
/gal/r1 |
1300 |
Adrian Jenkins 300 VIRGO 200 Richard Bower 300 Vincent Eke 100 Hansik Kim 100 Nikos Fanidakis 200 Noam Libeskind 50 Rowena Malbon 100 Rob Crain 100 Cesario Almeida 300 Laura Silva 50 James Mullaney 30 Andrew Benson 15 Chiara Tonini 10 Yamila Yaryura 30 David Wake 20 Jim Geach 25 Mark Swinbank 10 Peder Norberg 20 Basil Dales 1 |
cdml0: |
/gal/r2 |
1300 |
Tom Theuns 300 Carlton Baugh 680 Shaun Cole
100 Cesario Almeida 200 Wong Tam 50 |
cdml0: |
/gal/r3 |
1300 |
Raul Angulo 400 Yan-Chuan Cai 150 David Murphy 150 Lydia Heck 20 Nelson Padilla 5 Julian Martinez 20 Alvaro Orsi 260 Violeta Gonzales 20 Claudia Lagos 250 Qi Guo 100 Jiaxin Han 100 Quan Guo 60 |
cdml0: |
/gal/r4 |
1300 |
Rowena Malbon 190 Darren Reed 100 Cedric Lacey 210 Ian Vernon 100 Adrian Jenkins 1 Ian McCarthy 200 Liang Gao 400 John Helly 175 Masahiro Nagashima 50 Juan Gonzalez 350 Owen Parry 50 Daniel John 50 Ben Lowing 50 Alex Merson 100 Lucia Guaita 5 Quan Guo 40 |
duomo: |
/gal/dsk1 |
427.8 |
Shaun Cole 70 Vicent Quilis 25 Vince Eke 30 Cedric Lacey 50
Rowena
Malbon 50 Stelios Kazantzidis 20 Pasquale Mazzotta 50 Nelson Padilla 50
Claudio Dalla Vecchia 30 Carlos Frenk 20 Ariel Sanchez 20 Noam
Libeskind
10-->0 Dajana Dzanovic 5 lrm? 10 |
duomo: |
/gal/dsk3 |
525.4 |
Peder Norberg 120 John Helly 200 Adrian Jenkins 100 Ben Moore
20 Andrew
Benson 10 |
duomo: |
/gal/perm |
3.8 |
Enforced quotas. See /gal/perm/Last_Dump for the time of the last
backup. |
* Others include Gus Evrard, Vince Eke, Julio Navarro, Bepi Tormen
Michiel
van Haarlem, Gustav Yepes, Mario Abadi, Enzo Branchini, Luis Teodoro,
Cesario
Lia, Vijay Narayan, Steve Hatton, Joerg Colberg, Ben Moore Wolfram
Freudling
and Nigel Metcalfe
CDM Cluster and qsub
Instructions
from
Lydia.
Good Practice
Here are few guidelines and suggestions that prevent users accidently
hogging
all the computing resources.
- All jobs that are expected to take more than 10 mins of CPU
should be
run at nice +18.
A useful alias that can be used for running such jobs can be setup
by inserting the following line in your .cshrc
alias run 'limit coredumpsize 0; nice +18 \!* >&! \!:1:r.out
&'
Then simply type
run progam.exe
and your program will be run and its output put in file progam.out
. - Low Priority Jobs: If you want to run a job at even
lower
priority,
i.e. you are in no hurry and don't mind if other people's jobs take
priority
over yours then run them at nice +19.
- Multiple Jobs: When you have several long jobs to run do
not
spread
them over all our machines. Please try to limit yourself to no more
than
3 or 4 CPUs in ordinary circumstances. Note that europa, callisto and
ganymede
each have two CPUs. Multiple jobs on one CPU should be set to run
sequentially
rather than simulataneously. Together these self-imposed restrictions
should
give all users a fair share of the CPU.
If you need more CPU than this and one or more processor is sitting
idle for 24 hours or more than do feel free to use it but be prepared
to
yield to others. In such cases run at nice +19 so that others get
priority
when using the default nice +18. - Urgent Jobs: If you need a
large chunk of CPU for a week
or less
to finish something urgently then follow this procedure. Send an email
to all the theorists (potential CPU users) explaining that you need
priority
on particular machine for a certain length of time. If no one complains
simply run your jobs there at high priority, ie at nice +17 rather than
nice +18.
- Beware that unattended emacs sessions have a habit of going
crazy and
grabbing all the CPU on a machine. An alias of the following form
prevents
it from using more than 30 mins of CPU before terminating.
alias edit 'limit cpu 30 min ; emacs \!* &' - Other
Resources: Note that the multiple CPUs on the IT
service's
Marvin machine are not heavily used and also have a reasonable
allocation
of disk space to go along with them.