[mnet-devel] FEC Performance

[Hauke Johannknecht]

Hi.

some tests with the new FEC show problems with cpu-work scaling
superlinear to the filesize.

numbers are blocks generated per minute:

3MB file: (<1 sec init)
     95 2003-04-12_23:45
     59 2003-04-12_23:46
     54 2003-04-12_23:47

5MB file: (2 sec init)
    102 2003-04-12_23:53
     38 2003-04-12_23:54
     37 2003-04-12_23:55
     40 2003-04-12_23:56
     [...]

8MB file: (5 sec init)
    143 2003-04-13_00:01
     26 2003-04-13_00:02
     25 2003-04-13_00:03
     26 2003-04-13_00:04
     [...]

15MB file: (30 sec init)
    242 2003-04-13_00:41
     15 2003-04-13_00:42
     15 2003-04-13_00:43
     15 2003-04-13_00:44
     [...]

26MB file: (136 sec init)
    402 2003-04-13_00:28
      8 2003-04-13_00:29
      8 2003-04-13_00:30
      8 2003-04-13_00:31
      [...]

not that i dont like the new erasurecode, but waiting 24+ hours
for blockgeneration on a 26MB file has lead to derogatory
feedback from testers here ...

i commited my changes to publish.py that make it publish to the
publishingcache, so please test if the observed behavior is some
forgotten env var on my side again ...

Gruss,
  Hauke

-- 
Hauke Johannknecht        Berlin / Germany        HJ422-RIPE
Use PGP ! -> lynx -dump http://www.ash.de/ash.asc | pgp -kaf




-------------------------------------------------------
This SF.net email is sponsored by: Etnus, makers of TotalView, The debugger 
for complex code. Debugging C/C++ programs can leave you feeling lost and 
disoriented. TotalView can help you find your way. Available on major UNIX 
and Linux platforms. Try it free. www.etnus.com
_______________________________________________
mnet-devel mailing list
mnet-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mnet-devel