Some more results...
I have plaid with various page sizes and various number of elements per
page. Here are the time it takes to write 100 000 elements, and the file
size we get :
512 256
128 64 32
16 8 4
4096 : 31.919/1 499 164 696 | 29.979/1 233 539 096 | 23.469/1 200 701
464 | 25.508/1 357 205 528 | 28.369/1 625 350 168 | 34.221/2 051 842 072 |
2048 : 29.746/1 249 300 504 | 20.987/996 016 152 | 13.594/677 605
400 | 17.851/678 602 776 | 18.018/812 675 096 | 17.607/1 025 921 048
| 27.490/1 433 583 640 | 40.937/2 164 510 744
1024 : 32.207/1 153 515 544 | 28.237/817 985 560 | 16.001/542 615
576 | 9.217/389 012 504 | 8.658/406 337 560 | 12.735/512 960 536
| 22.671/716 791 832 | 31.206/1 082 255 384
512 : 37.644/1 109 965 848 | 29.288/738 782 744 | 15.568/441 060
376 | 11.156/332 613 144 | 8.514/242 526 232 | 8.560/256 480 280
| 10.404/358 395 928 | 16.470/541 127 704
256 : 51.904/1 100 778 264 | 37.151/706 175 000 | 18.463/407 867
928 | 14.122/287 093 784 | 10.836/206 497 560 | 9 705/169 157 912
| 9.942/179 197 976 | 13.469/270 563 864
The best results are obtained for pages containing 512 bytes, and
storing 32 elements per page :
512/32 : 11 745 write/s
The reads are also better :
239 808 read/s (first read)
1 612 900 read/s (second read)
This is *way* better than the original tests I did yesterday and this
morning.
That also means we will get very variable results depending on the
system we run the test, and the data we manipulate. Here, the key is a
Long, and the value is a small String, so it' not really characteristic
of what we will get in real life...
--
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com
---------------------------------------------------------------------
To unsubscribe, e-mail: labs-unsubscr...@labs.apache.org
For additional commands, e-mail: labs-h...@labs.apache.org
