[jcifs] 0.6 speed!

Michael B.Allen mballen at erols.com
Fri Dec 7 07:33:21 EST 2001


On Thu, 06 Dec 2001 11:05:23 -0800
Rob Wygand <rob at filefish.com> wrote:

> (All times in mSec. All 10 user numbers are averages of the 10 result times)
> 
>                  jCIFS 0.5             jCIFS 0.6b3          GAIN
> # files:     1 user / 10 users     1 user / 10 users  (1 user / 10users)
> ------------------------------------------------------------------------
> 1            22     / 119          9      / 15.7       2.4x   / 7.6x
> 3            466    / 1251         238    / 89.6       1.96x  / 13.97x
> 20           241    / 2564         115    / 218.2      2.10x  / 11.75x
> 57           541    / 6065         241    / 1707.4     2.24x  / 2.55x
> 155          1923   / 23614.2      1154   / 4761       1.67x  / 4.96
> 212          3434   / 39117.2      722    / 5651.7     4.75x  / 6.92
> 
> I'm suspicious of the 14x and 12x numbers, but everything else looks 
> great! To get a 6x increase under load makes me so happy I could spit, 
> to coin a phrase.
> 
> Just wanted to share the perf numbers with you all...

Wow, that 39 seconds is pretty bad. I think caching attributes for a
fraction of a second really helps here. Incedentally, 10 threads is a
lot for one host. The most dramatic results would come from 10 users
all accessing different hosts. And these numbers will be left in the
dust when I implement listFiles() (1 user could list 10000 files in a
comparable amount of time to 212 files). Ultimately, I am confident that
jCIFS will be faster than NT.

Mike

-- 
May The Source be with you.




More information about the jcifs mailing list