Another thing that would be useful would be to run with verbosegc output so we can understand what impact the garbage collector is having.

I found that increasing the eden size was very beneficial - since we want to minimise the number of full GCs. I found for my tests that a heap of 1GB with an eden size of 300MB worked well, e.g. -Xms1024m -Xmx1024m -XX:NewSize=300m. On jrockit the default eden size is better (for me) since it increases in line with the number of CPUs but to set it manually specify -Xns (n for nursery).

RG


Michael Bauroth <[EMAIL PROTECTED]>

01/05/2006 15:19
Please respond to mina-dev

       
        To:        [email protected]
        cc:        
        Subject:        Re: Performance - some interesting results



Sorry, I've made a mistake. The -server switch was toggeled on the whole
time. Another missing info: The buffer size (receive) was 32k.

Some more results:

All the measurements was made with SimpleByteBufferAllocator. Now I've
made the same tests with the default Allocator. In this context the
direct buffer memory consumption performs much better (continously) on
the same level like heap buffer. Additional I've seen, that heap buffer
works slower with Simple Allocator.


heap buffer (default Allocator)
min. / max. memory free (1740 kB)
5 - loop time: 0.2776
10 - loop time: 0.083
15 - loop time: 0.074
20 - loop time: 0.078
25 - loop time: 0.073
30 - loop time: 0.075
35 - loop time: 0.073


heap buffer (Simple Allocator)
min. / max. memory free (1740 kB)
5 - loop time: 0.2738
10 - loop time: 0.084
15 - loop time: 0.079
20 - loop time: 0.078
25 - loop time: 0.084
30 - loop time: 0.082
35 - loop time: 0.081


direct buffer (default Allocator)
min. / max. memory free (1740 kB)
5 - loop time: 0.2522
10 - loop time: 0.08
15 - loop time: 0.077
20 - loop time: 0.076
25 - loop time: 0.075
30 - loop time: 0.076
35 - loop time: 0.076


direct buffer (simple Allocator)
min. memory free (1290 kB)
max. memory free (1700 kB)
5 - loop time: 0.2576
10 - loop time: 0.080
15 - loop time: 0.073
20 - loop time: 0.074
25 - loop time: 0.075
30 - loop time: 0.075
35 - loop time: 0.074

One more: this measurements are for Debian 3.1 with Kernel 2.4. I know,
that the times can vary much from this from the Kernel 2.6!

Regards
Michael






This communication is for informational purposes only. It is not intended
as an offer or solicitation for the purchase or sale of any financial
instrument or as an official confirmation of any transaction. All market prices,
data and other information are not warranted as to completeness or accuracy and
are subject to change without notice. Any comments or statements made herein
do not necessarily reflect those of JPMorgan Chase & Co., its subsidiaries
and affiliates.

Reply via email to