Changes between Version 4 and Version 5 of Performance/Sep2011


Ignore:
Timestamp:
2012-05-20T18:03:59Z (13 years ago)
Author:
warner
Comment:

doubts about MDMF download "linear in k" slowdown

Legend:

Unmodified
Added
Removed
Modified

  • Performance/Sep2011

    v4 v5  
    9595Complete benchmark toolchain and data included in
    9696attachment:atlasperf.git.tar.gz
     97
     98Zooko raised the question (20-May-2012) on IRC:
     99{{{
     100<zooko> warner: this page says that K=60 takes twice as long as K=30:
     101<zooko> https://tahoe-lafs.org/trac/tahoe-lafs/wiki/Performance/Sep2011
     102<zooko> But these graphs seems to show those two taking about the same time as
     103each other:
     104<zooko>
     105https://tahoe-lafs.org/trac/tahoe-lafs/attachment/wiki/Performance/Sep2011/MDMF-100MB-partial.png
     106<zooko>
     107https://tahoe-lafs.org/trac/tahoe-lafs/attachment/wiki/Performance/Sep2011/MDMF-1MB-partial.png
     108<zooko> I hope the latter is true. :-)
     109}}}
     110
     111Those partial-read graphs certainly don't show a linear difference
     112between k=30 and k=60. At best there might be a 4% difference between
     113k=3 and k=30, and k=30/k=60 look identical.
     114
     115I think I was prompted to write that sentence by looking at the yellow
     116"trunk-MDMF" line in the attachment:CHKMDMF-100MB-vs-k.png graph, which
     117*does* show a linear slowdown with increasing k (with a lower multiplier
     118than the new-immutable-downloader). I don't know how to reconcile the
     119two. So something is funny, and I need to review the data, and possible
     120run new tests.