EditAttachPrintable
r9 - 02 Aug 2004 - 08:42:00 - LeoGalambosYou are here: TWiki >  Egothor Web  > EngineBenchmark

Environment

The hit list is produced from 2-term query - it shows I/O performance together with speed of evaluation formula. The total length of the list is about 10% of all documents which are saved in the index. Nothing has been printed to console or logs.

O/S, HW: Sun JDK 1.4.2_04, Gentoo Linux 2.4.22-r5, reiserfs 3.6.11, single CPU (P4), UDMA 100

Tested software: egothor 1.2.5-RC (64-bit engine), Lucene 1.3 (32-bit kernel)

The Lucene in this test did not call optimize routine during the indexing. The routine was called when all documents were already processed. See LuceneIssue.

Document collection

About 17,976 documents (432,411 kB) in HTML, XML, PDF (APIs, guides): css2, docbookXSL, j2sdk-1.3.1, jsguide13, tagref, xerces2, cyberneko, htmlpres, j2sdk-1.4.1, jsguide15, tdg (docbook guide), xml, cs-en dictionary, httpclient-2.0-beta1, java-tutorial, jsref13, uml.

Results

Batch of 1000 queries.

egothor 1.2.5 RC, NORMFACTOR = 1000
Config Searching (millisec per query) Indexing
indexSeq simpleBits sys user real sys user real index size (kB)
true true             32,578
egothor 1.2.5 RC, NORMFACTOR = 10000
Config Searching (millisec per query) Indexing
indexSeq simpleBits sys user real sys user real index size (kB)
true true 0.07 3.02 3.13 0'11.1" 6'23.3" 7'16.7" 40,857
true false 0.06 3.59 3.90 0'11.8" 6'25.8" 7'24.1" 37,237
false true 0.08 3.02 3.13 1'2.8" 7'0.8" 8'44.9" 41,337
false false 0.08 3.85 3.95 1'3.5" 7'5.3" 8'53.4" 37,717
lucene 1.3
N/A 0.10 3.70 3.79 0'23.6" 9'39.9" 11'4.1" 34,893

Egothor indexed all XML, PDF and HTML documents in the collection. Lucene indexed only HTML. The difference is less than 100 documents. Therefore it cannot affect the results.

Tests prepared by LeoGalambos. Comments and questions are welcome.

Edit | Attach | Printable | Raw View | Backlinks: Web, All Webs | History: r9 < r8 < r7 < r6 < r5 | More topic actions
Know.EngineBenchmark moved from Know.Benchmark on 13 Apr 2004 - 14:49 by LeoGalambos
 
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback