I have just done some test of index with and without
"IncrementalCitation" option which appeared in 1.1.2
In both cases I created a database with about 4200
URLs (according to var/aspseek/total file) and index
more 1000 URLs (by running index -N50 -n1000). After
that, I run index -D two times (second run is actually
unneeded but it shows coolness of IncrementalCitations).
Here are the results (time index -D output):
1. Without IncrementalCitations
first run: 3.85u 2.53s 0:23 elapsed
second run: 2.33u 1.31s 0:13 elapsed
2. With I.C.
first run: 2.46u 1.62s 0:12 elapsed
second run: 0.56u 0.07s 0:01 elapsed
So you see that with I.C. index -D runs
twice faster. This is because calculations
that has been done before are not repeated.
So, I.C. is not for conserving memory, it is also
faster, and it is seen even with such small number
of URLs. If you have 100.000 URLs the difference will
be substantially larger. If you have a million URLs,
this is killer feature for you. So, consider testing
and seeing it yourself.
NOTE although that in 1.1.2 IncrementalCitation realisation
have some bugs, so please wait for 1.1.3, which we will release
shortly (hmm, there surely can be other bugs in 1.1.3:-)
NOTE that format of data used by I.C. is not compatible
with the old one, so you need to reindex everything from
scratch, and set "IncrementalCitation yes" in both
aspseek.conf and searchd.conf files.
PS: I tested it on dual PIII/800 with 512 megs of RAM
and MySQL with 256 Kb key_buffer. YMMW.
-- [EMAIL PROTECTED] http://kir.sever.net ICQ 7551596 --
If you can't stand the heat, sit down or leave the sauna
Now listening to Cranberries "Ode to My Family"
