Hi Mike, I retested and results are the same:
1/ I did not use sort (so FieldCache should not enter picture?) 2/ I created indexed data from scratch separately for 361 and 43 based on same text (text files), and I ran test from command line separately against each index folder, so seems a pretty fair test. 3/ Each test I created searcher from scrath (to measure creation time). I did not include JVM start time in each case. The tests are in same box. >From indexed data it seems that 43 generated a lot more data in folder, below I listed (ls -ltr) result (always pass in LUCENE_43 version, so lucen 42 codec should be used, why lucene41?). Thanks very much for helps, Lisheng /////////////////// 36: total 1228 -rw-r--r-- 1 root root 68 Jul 31 15:50 _0.fdx -rw-r--r-- 1 root root 44 Jul 31 15:50 _0.fdt -rw-r--r-- 1 root root 132 Jul 31 15:50 _0.tvx -rw-r--r-- 1 root root 260207 Jul 31 15:50 _0.tvf -rw-r--r-- 1 root root 20 Jul 31 15:50 _0.tvd -rw-r--r-- 1 root root 345803 Jul 31 15:50 _0.tis -rw-r--r-- 1 root root 4899 Jul 31 15:50 _0.tii -rw-r--r-- 1 root root 539098 Jul 31 15:50 _0.prx -rw-r--r-- 1 root root 12 Jul 31 15:50 _0.nrm -rw-r--r-- 1 root root 61703 Jul 31 15:50 _0.frq -rw-r--r-- 1 root root 29 Jul 31 15:50 _0.fnm -rw-r--r-- 1 root root 252 Jul 31 15:50 segments_1 -rw-r--r-- 1 root root 20 Jul 31 15:50 segments.gen 43: total 1828 -rw-r--r-- 1 root root 45 Jul 31 15:51 _0.fdx -rw-r--r-- 1 root root 66 Jul 31 15:51 _0.fdt -rw-r--r-- 1 root root 60 Jul 31 15:51 _0.tvx -rw-r--r-- 1 root root 176845 Jul 31 15:51 _0.tvd -rw-r--r-- 1 root root 10980 Jul 31 15:51 _0_Lucene41_0.tip -rw-r--r-- 1 root root 401339 Jul 31 15:51 _0_Lucene41_0.tim -rw-r--r-- 1 root root 1007621 Jul 31 15:51 _0_Lucene41_0.pos -rw-r--r-- 1 root root 216711 Jul 31 15:51 _0_Lucene41_0.pay -rw-r--r-- 1 root root 12048 Jul 31 15:51 _0_Lucene41_0.doc -rw-r--r-- 1 root root 46 Jul 31 15:51 _0.nvm -rw-r--r-- 1 root root 34 Jul 31 15:51 _0.nvd -rw-r--r-- 1 root root 205 Jul 31 15:51 _0.fnm -rw-r--r-- 1 root root 395 Jul 31 15:51 _0.si -rw-r--r-- 1 root root 69 Jul 31 15:51 segments_1 -rw-r--r-- 1 root root 20 Jul 31 15:51 segments.gen /////////////////// -----Original Message----- From: Michael McCandless [mailto:luc...@mikemccandless.com] Sent: Wednesday, July 31, 2013 11:31 AM To: Lucene Users Subject: Re: lucene 4.3 seems to be much slower in indexing than lucene 3.6? On Tue, Jul 30, 2013 at 6:13 PM, Zhang, Lisheng <lisheng.zh...@broadvision.com> wrote: > Hi Mike, > > I did more tests with realistic text from different languages (typical > text for 8 different languages, English one is novel "Animal Farm"). > > What I found seems to be: > > ## Indexing: > 36 and 43 comparable (your previous comment is very correct). > > ## Search: > 43 seems to be slower (30%), checking details, it seems it all due to > initial searcher creation and first search (warming), as if 43 did much > more in warming? Hmm, I'm not sure off hand why searcher warming would be slower in 4.3. Are you relying on FieldCache (e.g. sorting by a field instead of by relevance). Switching to doc values should make warming much faster. Are you sure the test was fair? Ie, in both cases the index was either hot or cold? For your 4.3 test you fully reindexed right? Ie, searched against a 4.3 (not 3.6) index? Mike McCandless http://blog.mikemccandless.com --------------------------------------------------------------------- To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org For additional commands, e-mail: java-user-h...@lucene.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org For additional commands, e-mail: java-user-h...@lucene.apache.org
