non-representative DB benchmark

Matthias Andree matthias.andree at gmx.de
Sun Sep 7 15:17:24 CEST 2003


Hi,

I have some results of a NON-REPRESENTATIVE benchmark (a sample for ONE
particular case) for a data base engine shoot-out.

NOTE that tdb natively supports concurrent access, in contrast to qdbm
(locks itself) and DB (we lock its files and don't use it in "concurrent"
mode yet).


WARNING: THESE RESULTS ARE _NOWHERE_ NEAR REAL-WORLD DATA WHEN MORE THAN ONE
READER IS ACCESSING THE DATA BASE. THIS IS SINGLE-USER, SINGLE-THREADED
ACCESS AND QUITE DIFFERENT FROM THE PROFILE ON A BUSY MAIL SERVER!


#1. all benchmarks single run with warm cache

#2. AMD Duron/700, VIA KT133 chip set, 320 MB SDRAM, Linux 2.4.22-pac1
    SYM53C875 SCSI Adaptor, Fujitsu MAH3182MP (7200/min) with WCE=0
    File system: reiserfs v3.6 via LVM1.

#3. I dumped my current 43,812 key combined DB data base and loaded it
    into each data base. (bogoutil.)

#4. I registered with bogofilter -n <~/Mail/spam-bf -d /tmp -v another
    mail box with 12,602 words and 195 messages, resulting in 50,813
    keys in the data base.

#5. I then run bogofilter -eMTB ~/Mail/lk -d /tmp -v >/dev/null to
    check how fast a different mailbox (linux-kernel, 2,700 messages) is
    scored.

#6. bogolexer -q <~/Mail/lk takes 4 seconds.

Results: QDBM is quick and has a lower footprint than DB, but DB is easier
on the system time with comparable times. TDB is slower in all respects, at
least in single-threaded mode.

** BerkeleyDB 4.0.14:
* load:
0.68user 0.04system 0:01.16elapsed 61%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (214major+369minor)pagefaults 0swaps

* add:
0.73user 0.05system 0:00.98elapsed 79%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (467major+1123minor)pagefaults 0swaps

* score:
13.80user 0.14system 0:14.13elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (1215major+314minor)pagefaults 0swaps

** TDB 1.0.6:
* load:
7.25user 0.62system 0:08.42elapsed 93%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (76032major+298minor)pagefaults 0swaps

* add:
3.80user 0.23system 0:04.09elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (27111major+359minor)pagefaults 0swaps

* score:
98.11user 0.56system 1:39.48elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (1040major+305minor)pagefaults 0swaps

** QDBM 1.6.15:
* load:
0.47user 1.62system 0:02.54elapsed 82%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (147major+299minor)pagefaults 0swaps

* add:
0.61user 0.47system 0:01.29elapsed 83%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (399major+360minor)pagefaults 0swaps

* score:
11.00user 5.65system 0:16.87elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (419major+306minor)pagefaults 0swaps

-- 
Matthias Andree

Encrypt your mail: my GnuPG key ID is 0x052E7D95




More information about the Bogofilter mailing list