Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.naic.edu/~phil/usrproj/usrproj.html
Дата изменения: Fri Feb 19 16:14:45 2016
Дата индексирования: Sun Apr 10 03:56:40 2016
Кодировка:
usrProjects

User projects

aug,2003

     This page has lets me communicate with users about their projects. It has example code on reducing various kinds of data as well as any notes on the projects.
Mock spectrometer i/o rates for various projects (100% duty cycle)
project
chan /
172 band
Pols
Bytes/
  smp
Smp Tm
  (usec)
I/O (Mb/sec)
1box / all
(*1)
Gb per Hour
1box / all
Typical
duty cycle
    (3*)
Gb/Hour
@dutyCyle
1box/all
a2130 galfacts
4096
4
2
1000
65.5 / 459
235.9 / 1652.
100%
same
a2130 after compression(*2)




 5.4 / 37.8
19.45 / 136.2

same
p2030 palfa
512
1
2
65.14
31.4 / 220.1
113.2/ 792.3
80%
90 / 633
p2473 (palfa commensal
with alfalfa)
512
1
2
128
16  / 112.
57.6 / 403.2
100%
same
p1693 327 pulsar search
b0s1g0 only
1024/
57Mhz Band
1
2
125
16.4/16.4
59 / 59.
100%
same

Notes *:
  1.  the factor of 2  ((4096*4*2/.001)=32mb/sec) is because there are two 172 Mhz bands dumped .
  2. The galfacts compression creates 2 output files: decimate in time  by 100, and decimate in frequency  by 32 chan. The bytes/smp increases from 2 to 4 bytes (floats).
    1. Processing times for galfacts compression:
      1. Input 120Mb/sec per box: 337*2=674 2gb files takes: 3hr,20min (this is a 6 hour run)
      2. output: 674 * (43mb + 135.4mb = 178.4 mb/input file) in 3hr,20min= 9.97 mb/sec output
  3. palfa duty cycle will vary depending on how many test pulsar they look at for each observing run. The more test pulsars, the more time spent slewing..65% was for a 3 hour run on 17apr09

home_~phil