IPP Software Navigation Tools IPP Links Communication Pan-STARRS Links

Changes between Version 1 and Version 2 of ThroughputDemo_20091104


Ignore:
Timestamp:
Nov 6, 2009, 10:50:08 AM (17 years ago)
Author:
eugene
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • ThroughputDemo_20091104

    v1 v2  
    11= Throughput Demonstration 2009.11.04 =
     2[[PageOutline]]
    23
    34== start @ 11:30am ==
     
    2526my fix to the camera paths had no effect: the paths are now of the form:
    2627
    27 neb://any/gpc1/throughput.20091104/o4999g0217o.78668/o4999g0217o.78668.cm.29757.smf
     28{{{neb://any/gpc1/throughput.20091104/o4999g0217o.78668/o4999g0217o.78668.cm.29757.smf}}}
    2829
    2930but they are still all going to ipp039, who is getting quite a load.
     
    7576At the moment, our limiting steps have used the following number of seconds:
    7677
     78{{{
    7779chip: 1,021,709.35 sec
    7880warp: 610,303.59 sec
     
    8183dist process: 177,430.2 sec
    8284camera: 69,067.08 sec
     85}}}
    8386
    8487=== update 1 ===
     
    117120I've gone through most errors and found the following:
    118121
    119 * all of the chip failures were caused by the issue with ipp014 ->
    120 these would be addessed by fixing the nebulous randomization.
    121 * all but one of the warp failures were caused by the chmod problem ->
    122 these can be addressed with one of the options I listed
    123 * the remaining warp failure was a more traditional NFS I/O problem,
    124 and would be naturally cleaned up with the revert pass
    125 * the diff failures I checked seemed to be the traditional NFS I/O
    126 problems, but we do not seem to have a diff revert task -- that is
    127 surprising!
     122 * all of the chip failures were caused by the issue with ipp014 ->
     123 these would be addessed by fixing the nebulous randomization.
     124 * all but one of the warp failures were caused by the chmod problem ->
     125 these can be addressed with one of the options I listed
     126 * the remaining warp failure was a more traditional NFS I/O problem,
     127 and would be naturally cleaned up with the revert pass
     128 * the diff failures I checked seemed to be the traditional NFS I/O
     129 problems, but we do not seem to have a diff revert task -- that is
     130 surprising!
    128131
    129132there are also some destreak failures which I did not investigate.