LDAS LHO (Hanford, WA) Database Insertion Rates
Test Period December, 1999
Input data Ilwd with dimensions 10, 100, 200
Platform metaserver.ligo-wa.caltech.edu (SUN Ultra 1, RAM=256MB, Solaris 2.7)
LDAS software ldas-0.0.6 
Results are displayed in the order inserted.
Table #rows per block insert CPU time (secs) Wall Time (secs) Rate (#rows/sec)
process 10 0.20 1 33
process 100 1.37 3 33
process 200 2.61 6.6 30.26
process_params 10 0.10 0 > 10,see note*
process_params 100 0.73 2 50
process_params 200 1.52 4.68 42.74
frameset_chanlist 10 0.11 0 > 10,see note*
frameset_chanlist 100 0.61 2 50
frameset_chanlist 200 1.22 5 40
frameset_writer 10 0.14 0.374 26.74
frameset_writer 100 0.96 2.78 35.93
frameset_writer 200 1.80 5.15 38.84
frameset 10 0.15 0.31 32.25
frameset 100 0.93 3 30.30
frameset 200 1.88 4.95 40.40
frameset_loc 10 0.12 0.34 29.41
frameset_loc 100 1.01 2.77 36.10
frameset_loc 200 1.97 5.11 39.14
segment_definer 10 0.10 0.34 29.41
segment_definer 100 0.80 2.49 40.16
segment_definer 200 1.62 4.60 43.48
segment 10 0.09 0.24 41.67
segment 100 0.68 2.26 44.24
segment 200 1.26 4.10 48.78
summ_value 10 0.15 0.41 24.39
summ_value 100 1.15 3.28 30.49
summ_value 200 2.42 5.84 34.25
summ_statistics 10 0.17 0.43 23.25
summ_statistics 100 1.42 3.41 29.33
summ_statistics 200 3.00 6.60 30.30
summ_comment 10 0.16 0.48 20.83
summ_comment 100 1.49 3.44 29.07
summ_comment 200 2.82 7.02 28.49
summ_spectrum 10 0.17 1.17 8.55
summ_spectrum 100 1.25 5.96 16.78
summ_spectrum 200 2.47 11.21 17.84
filter 10 0.10 0.32 31.25
filter 100 0.71 2.35 42.55
filter 200 1.36 5.55 36.04
filter_params 10 0.13 0.33 30.30
filter_params 100 0.80 2.43 41.15
filter_params 200 1.56 4.53 44.15
gds_trigger 10 0.20 0.34 29.41
gds_trigger 100 1.36 3.36 29.76
gds_trigger 200 2.83 6.40 31.25
sngl_inspiral 10 0.14 0.33 30.30
sngl_inspiral 100 1.12 2.98 33.56
sngl_inspiral 200 2.24 6.36 31.45
sngl_burst 10 0.14 0.33 30.30
sngl_burst 100 1.03 2.60 38.46
sngl_burst 200 2.01 5.52 36.23
sngl_ringdown 10 0.17 0.36 27.78
sngl_ringdown 100 0.98 2.83 35.53
sngl_ringdown 200 1.95 5.71 35.03
sngl_unmodeled 10 0.13 0.40 25.00
sngl_unmodeled 100 1.04 2.66 37.59
sngl_unmodeled 200 2.04 6.07 32.95
sngl_unmodeled_v 10 0.08 0.20 50.00
sngl_unmodeled_v 100 0.57 2.24 44.65
sngl_unmodeled_v 200 1.06 3.23 61.92
sngl_dperiodic 10 0.18 0.36 27.78
sngl_dperiodic 100 1.21 3.17 30.91
sngl_unmodeled_v 200 2.47 6.47 30.91
sngl_datasource 10 0.12 0.29 34.48
sngl_datasource 100 0.93 2.33 42.92
sngl_datasource 200 1.85 4.38 45.66
sngl_transdata 10 0.14 1.12 8.93
sngl_transdata 100 1.29 5.16 19.38
sngl_transdata 200 2.61 11.34 17.64
multi_inspiral 10 0.22 0.45 22.22
multi_inspiral 100 1.32 3.21 31.15
multi_inspiral 200 2.75 7.10 28.17
multi_burst 10 0.14 0.39 25.64
multi_burst 100 1.18 2.96 33.78
multi_burst 200 2.23 6.10 32.79
coinc_sngl 10 0.26 0.48 20.83
coinc_sngl 100 2.20 3.92 25.51
coinc_sngl 200 4.31 9.55 20.94

Summary:

  1. Insertion rate is faster on the hanford machines than on deneb.ligo.caltech.edu, due to faster processor and larger memory.
  2. Smaller tables insert faster. Tables with blobs such summ_spectrum and sngl_transdata insert slowest.
  3. No significant memory leaks were seen during the insertion. The size of metadataAPI process remains around 12MB
    4. Note:
    1. The wall time for these insertions were not precise when measured in seconds because the insertion rate is 10x faster than for CIT database; the code measuring wall time was modified to record in milli-seconds to get better precision for the rest of the insertion.
    2. Since unique Ids have to be generated for certain tables e.g process, filter, sngl_xxx tables, multi_xxx tables etc. via canned sql in the LDAS metadataAPI, the sql statement size restricts the ilwd dimension to <= 200.
      3.