2VB1: Difference between revisions
| Line 31: | Line 31: | ||
JOBS=DEFPIX INTEGRATE CORRECT | JOBS=DEFPIX INTEGRATE CORRECT | ||
and run "xds_par" again. It completes after about 5 minutes on a fast machine, and we may inspect CORRECT.LP . | and run "xds_par" again. It completes after about 5 minutes on a fast machine, and we may inspect CORRECT.LP . | ||
== timings for processing sweep "e" as a function of MAXIMUM_NUMBER_OF_PROCESSORS and MAXIMUM_NUMBER_OF_JOBS == | == timings for processing sweep "e" as a function of MAXIMUM_NUMBER_OF_PROCESSORS and MAXIMUM_NUMBER_OF_JOBS == | ||
| Line 63: | Line 64: | ||
total cpu time used 2833.4 sec | total cpu time used 2833.4 sec | ||
total elapsed wall-clock time 566.5 sec | total elapsed wall-clock time 566.5 sec | ||
but please note that this actually only uses 10 processors, since the default DELPHI=5 | |||
and the OSCILLATION_RANGE is 0.5°. | |||
Using | Using | ||
| Line 82: | Line 85: | ||
MAXIMUM_NUMBER_OF_JOBS=6 | MAXIMUM_NUMBER_OF_JOBS=6 | ||
performs best for a 2-Xeon X5570 machine with 24GB of memory and a RAID1 consisting of 2 1TB SATA disks. It should be noted that the dataset has 27GB, and in 296 seconds this means 92 MB/s continuous reading. The processing time is thus limited by the disk access, not by the CPU. And no, the data are not simply read from RAM (tested by "echo 3 > /proc/sys/vm/drop_caches before the XDS run). | performs best for a 2-Xeon X5570 machine with 24GB of memory and a RAID1 consisting of 2 1TB SATA disks. It should be noted that the dataset has 27GB, and in 296 seconds this means 92 MB/s continuous reading. The processing time is thus limited by the disk access, not by the CPU. And no, the data are not simply read from RAM (tested by "echo 3 > /proc/sys/vm/drop_caches before the XDS run). | ||
== Comparison of XDS and published results of data processing == | == Comparison of XDS and published results of data processing == | ||
Revision as of 19:20, 20 February 2011
This reports processing of triclinic hen egg-white lysozyme data @ 0.65Å resolution (PDB id 2VB1). Data (sweeps a to h, each comprising 60 to 360 frames of 72MB) were collected by Zbigniew Dauter at APS 19-ID and are available from here. Details of data collection, processing and refinement are published.
XDS processing
- use generate_XDS.INP to obtain a good starting point
- edit XDS.INP and change the following:
ORGX=3130 ORGY=3040 ! for ADSC, header values are subject to interpretation; better inspect the table in IDXREF.LP! TRUSTED_REGION=0 1.5 ! we want the whole detector area ROTATION_AXIS=-1 0 0 ! at this beamline the spindle goes backwards!
- for faster processing on a machine with many cores, use (e.g. for 16 cores):
MAXIMUM_NUMBER_OF_PROCESSORS=2 MAXIMUM_NUMBER_OF_JOBS=8
For all the sweeps, processing stopped with an error message after the IDXREF step. By inspecting IDXREF.LP, one should make sure that everything works as it should, i.e. that a large percentage of reflections was actually indexed nicely:
... 63879 OUT OF 72321 SPOTS INDEXED. ... ***** DIFFRACTION PARAMETERS USED AT START OF INTEGRATION ***** REFINED VALUES OF DIFFRACTION PARAMETERS DERIVED FROM 63879 INDEXED SPOTS REFINED PARAMETERS: DISTANCE BEAM AXIS CELL ORIENTATION STANDARD DEVIATION OF SPOT POSITION (PIXELS) 0.53 STANDARD DEVIATION OF SPINDLE POSITION (DEGREES) 0.12
It may be possible to adjust some parameters (for COLSPOT) so that the error message does not occur, but it is not worth the effort. So we just change
JOBS=XYCORR INIT COLSPOT IDXREF DEFPIX INTEGRATE CORRECT
to
JOBS=DEFPIX INTEGRATE CORRECT
and run "xds_par" again. It completes after about 5 minutes on a fast machine, and we may inspect CORRECT.LP .
timings for processing sweep "e" as a function of MAXIMUM_NUMBER_OF_PROCESSORS and MAXIMUM_NUMBER_OF_JOBS
The following is going to be rather technical! If you are only interested in crystallography, skip this.
Using
MAXIMUM_NUMBER_OF_PROCESSORS=2 MAXIMUM_NUMBER_OF_JOBS=8
we observe for the INTEGRATE step:
total cpu time used 2063.6 sec total elapsed wall-clock time 296.1 sec
Using
MAXIMUM_NUMBER_OF_PROCESSORS=1 MAXIMUM_NUMBER_OF_JOBS=16
the times are
total cpu time used 2077.1 sec total elapsed wall-clock time 408.2 sec
Using
MAXIMUM_NUMBER_OF_PROCESSORS=4 MAXIMUM_NUMBER_OF_JOBS=4
the times are
total cpu time used 2102.8 sec total elapsed wall-clock time 315.6 sec
Using
MAXIMUM_NUMBER_OF_PROCESSORS=16 ! the default for xds_par on a 16-core machine MAXIMUM_NUMBER_OF_JOBS=1 ! the default
the times are
total cpu time used 2833.4 sec total elapsed wall-clock time 566.5 sec
but please note that this actually only uses 10 processors, since the default DELPHI=5 and the OSCILLATION_RANGE is 0.5°.
Using
MAXIMUM_NUMBER_OF_PROCESSORS=4 MAXIMUM_NUMBER_OF_JOBS=8
(thus overcommitting the available cores by a factor of 2) the times are
total cpu time used 2263.5 sec total elapsed wall-clock time 320.8 sec
Using
MAXIMUM_NUMBER_OF_PROCESSORS=4 MAXIMUM_NUMBER_OF_JOBS=6
(thus overcommitting the available cores, but less severely) the times are
total cpu time used 2367.6 sec total elapsed wall-clock time 267.2 sec
Thus,
MAXIMUM_NUMBER_OF_PROCESSORS=4 MAXIMUM_NUMBER_OF_JOBS=6
performs best for a 2-Xeon X5570 machine with 24GB of memory and a RAID1 consisting of 2 1TB SATA disks. It should be noted that the dataset has 27GB, and in 296 seconds this means 92 MB/s continuous reading. The processing time is thus limited by the disk access, not by the CPU. And no, the data are not simply read from RAM (tested by "echo 3 > /proc/sys/vm/drop_caches before the XDS run).
Comparison of XDS and published results of data processing
| unique reflections (30-0.65Å) | completeness (%) | data redundancy | R merge | avg I/sigma | completeness in highest resolution range (0.67-0.65Å) | data redundancy in highest resolution range | R merge in highest resolution range | avg I/sigma in highest resolution range | |
| published | 187165 | 97.6 | 7.1 | 0.04 | 36.20 | 67.3 | 2.7 | 0.18 | 4.20 |
| XDS | 190017 | 99.4 | 7.5 | 0.033 | 33.20 | 94.1 | 3.9 | 0.27 | 4.49 |