Benchmark test using XBench

[XBench]
http://db.uwaterloo.ca/~ddbms/projects/xbench/

XML Data that use in the experiment is generated by XBench.
Option is TC, MD, small.

Database Class is TC/MD
	*Why MD?
		XML Database manages lots of documents, so people use Database.
	*Why TC?
		Xbench creates good number of files for test.

documents - 26
size - 11M
minimum - 120K / maximum - 1.2M	
(The size is maybe suitable for actual use.)
 
724K article1.xml    36K article16.xml  748K article22.xml  316K article5.xml
208K article10.xml  736K article17.xml  160K article23.xml  688K article6.xml
476K article11.xml   16K article18.xml  640K article24.xml  276K article7.xml
648K article12.xml  272K article19.xml  576K article25.xml  424K article8.xml
208K article13.xml  1.2M article2.xml   120K article26.xml  168K article9.xml
108K article14.xml  888K article20.xml  104K article3.xml
264K article15.xml  644K article21.xml   28K article4.xml

------------------------------------------------
*** Experiment environment *** 

[Machine]
	IBM Thinkpad X23

[CPU]
	$cat /proc/cpuinfo
	model name : Intel(R) Pentium(R) III Mobile CPU 866MHz
	bogomips : 1723.59

[Kernel]
	Redhat Linux 8.0 kernel 2.4.20
	$uname -r
	2.4.20

[Memory]
	384MB
	$free -m | grep Mem
	Mem: 377

[HDD]
	hdparm -T -t /dev/hda
	/dev/hda:
	Timing buffer-cache reads: 128 MB in 0.58 seconds =220.69 MB/sec
	Timing buffered disk reads: 64 MB in 3.23 seconds = 19.81 MB/sec 

[PostgreSQL]
	version 7.4 beta 3
	
	$cat postgresql.conf
	shared_buffers = 3840	#30M
	sort_mem = 8168			#8M
	wal_sync_method = fdatasync
	effective_cache_size = 10000   # 96M = 12000
	
	Postgres uses Max 32M for query except sorting, cache etc.
	Postgres uses 16M for sorting.
	
	Command line paramator:
	/usr/local/pgsql/bin/postmaster '-F' '-i

[eXist]
	version 0.9.1
	
	Java Options
	    -Xms32000k  initial Java heap size (64M)
	    -Xmx64000K  maximum Java heap size (128M)

	both client and server uses heap.

	this is my present (-:
	seems to documents fit in buffer.
	
	We evaluated with embedded mode, this maybe cause some advantage for eXist.

------------------------------------------------

Q1 Simple path match
/article/prolog/title

Q2 Ordered access with position
/article/body/section[1]/@heading

Q3 Exact match, that has string comparation
/article/prolog[authors/author/name="Yuiko Nytro"]/title

Q4 Numeral comparation
/article/epilog/references[a_id>=25]

Q5 Query contains expression '//'
/article//author/name

Q6 Query contains plural expression '//'
//article//author

Q7 query contains axis that cannot abbreviate
/article/body/section/following-sibling::*

Q8 query contains nodeTest
//article/parent::node()

----------------------------------------------------

http://localhost:8088/?_xpath=document(*)/article/prolog/title
