NHibernate Performance Testing - Revision history

MakoBot: Removed 'Template:' prefix from template includings and/or changed external-link into internal-link

2011-06-10T23:11:54Z

Removed 'Template:' prefix from template includings and/or changed external-link into internal-link

2009-05-21T22:47:37Z

2009-01-03T08:27:10Z

‎Local MySQL/MyISAM Analysis

2009-01-03T08:18:15Z

‎Local MySQL/MyISAM Analysis

2009-01-03T08:15:06Z

‎Local MySQL/MyISAM Analysis

2009-01-03T08:13:50Z

‎Local MySQL/MyISAM Versus SQLite Comparison

2009-01-03T08:13:19Z

2009-01-03T08:12:33Z

2009-01-03T08:11:49Z

2009-01-03T08:11:26Z

@@ Line 1: / Line 1: @@
 __NOTOC__
-{{Template:Quicklinks}}
+{{Quicklinks}}
 <br />

← Older revision		Revision as of 22:47, 21 May 2009
Line 1:		Line 1:
		+	__NOTOC__
		+	{{Template:Quicklinks}}
		+	<br />
		+
	Here are results from NHibernate performance tests run on SQLite and MySQL database.		Here are results from NHibernate performance tests run on SQLite and MySQL database.

@@ Line 27: / Line 27: @@
 [[Image:mysql_performance_no_session_recycling.png]]
-The update and individual select operations were most expensive with 400 micro seconds cost per object.
+The update and individual select operations were most expensive with 400 microseconds cost per object.
-Inserts and deletes had 200 micro seconds cost due to the lack of indexing work required. Only primary key (id) index had to be updated.
+Inserts and deletes had 200 microseconds cost due to the lack of indexing work required. Only primary key (id) index had to be updated.
-Fastest were mass select which measured only the streaming speed from the database and ended up with 10 micro seconds cost per object.
+Fastest were mass select which measured only the streaming speed from the database and ended up with 10 microseconds cost per object.
 As conclusion MySQL performance with table sizes up to 1M of rows were not bound by table size dependent operations but factors like network delay and data transfer rate between application and database.

@@ Line 33: / Line 33: @@
 Fastest were mass select which measured only the streaming speed from the database and ended up with 10 micro seconds cost per object.
-As conclusion MySQL performance with table sizes up to 1M of rows were not bound by table size dependent operations but factors like network lag and streaming rate of data from database to the application.
+As conclusion MySQL performance with table sizes up to 1M of rows were not bound by table size dependent operations but factors like network delay and data transfer rate between application and database.
 == Performance Test Code and Configuration ==

@@ Line 23: / Line 23: @@
 [[Image:mysql_performance.png]]
-With session recycling there is only transaction cost to win which gives yelds 2 times better performance when updating 30 or more objects at once.
+With session recycling there is only transaction cost to win which yields 2 times better performance when updating 30 or more objects at once.
 [[Image:mysql_performance_no_session_recycling.png]]

@@ Line 13: / Line 13: @@
 === Local MySQL/MyISAM Versus SQLite Comparison ===
-Here is more in depth comparison with local MyISAM and SQLite using 1000 test objects and variable amount of initial objects varying from 1 to 1 000 000. SQLite insert and delete performance seems to degrade almost linearly.
+Here is more in depth comparison with local MyISAM and SQLite using 1000 test objects and variable amount of initial objects varying from 1 to 1 000 000. SQLite insert and delete performance degrades almost linearly.
 [[Image:mysql_sqlite_performance_comparison.png]]

@@ Line 7: / Line 7: @@
 * Performance test is execute with NAnt and .NET-framework 2.0.
-In the first test 1000 rows are inserted, updated, selected and deleted against tables with 0 and 10000 rows initially. Remote MySQL is added to demonstrate the network lag effect when batch operations are not used (It seems NHibernate does not support batch inserts, updates and deletes for MySQL). According to the results MySQL with MyISAM has best performance:
+In the first test 1000 rows are inserted, updated, selected and deleted against tables with 0 and 10000 rows initially. Remote MySQL is added to demonstrate the network lag effect when batch operations are not used. NHibernate does not support batch inserts, updates and deletes in combination with MySQL. According to the results MySQL with MyISAM has best performance:
 [[Image:nhibernate_performance_test.png]]

@@ Line 2: / Line 2: @@
 * Local machine is Windows Vista 64 bit with Quad Core processor and 4G memory. MySQL has factory settings.
-* Remote mysql database machine is Ubuntu Linux virtual machine with one virtual processor and 2G memory. MySQL has factory settings.
+* Remote MySQL database machine is Ubuntu Linux virtual machine with one virtual processor and 2G memory. MySQL has factory settings.
 * Local machine ping is 10 ms to the remote machine.
 * There are no other clients loading the database and this comparison does not reflect reliably to real load with mutiple simultaneous connections.