Introduction

Hey, I know. These are micro benchmarks and can't be taken as unique meter to compare products. However it's an index of the speed of the engine. Take it for what it is!

The tests aim to demonstrate the performance of all Database types.

All the speed test cases are located under the source folder: trunk/tests/src/test/java/com/orientechnologies/orient/test/database/speed

Scenario

1 DELL Notebook model XPS M1530 with Intel(r) Core Duo T7700 2.40Ghz, 3 GB RAM and HD 5.400rpm, O.S. MS Windows Vista, JRE 1.6.0_20

JRE Startup configuration: -server -XX:+UseParallelGC -XX:+AggressiveOpts -XX:CompileThreshold=200

NOTE: After some comparison with similar HW I found that OrientDB runs better on Windows and MacOsX systems. Linux seems to be slower than others because of EXT3 file system (EXT4 is much more slower again!) and maybe the JVM implementation even if it's the same release on the other systems

Massive insert of records

This test inserts 1,000,000 (one million) of records.

Comparison between local, remote and distributed

Source	Database type	Database location	Transaction mode	Total time elapsed (less is better)	Requests per second (more is better)	Tested version
Source	ODatabaseFlat	memory	no tx	1.91 seconds	510,464	0.9.9
Source	ODatabaseFlat	memory	optimistic tx	2.31 seconds	422,475	0.9.9
Source	ODatabaseFlat	local	no tx	4.72 seconds	209,248	0.9.8
Source	ODatabaseFlat	local	optimistic tx	10.15 seconds	98,173	0.9.8
Source	ODatabaseFlat	remote	no tx	75.82 seconds	13,189	0.9.8
Source	ODatabaseFlat	remote	optimistic tx `#1`	13.23 seconds	75,437	0.9.8

#1 Remote connection benefits of optimistic transaction since it compress all the data and send it at the commit in one shot.

Comparison between Database types

Source	Database type	Database location	Total time elapsed (less is better)	Requests per second (more is better)	Tested version
Source	ODatabaseFlat	local	4.72 seconds	209,248	0.9.8
Source	ODatabaseDocumentTx	local	9.24 seconds	107,677	0.9.8
Source	ODatabaseObjectTx	local	14.98 seconds	66,431	0.9.9

Massive insert of pairs key/value

This test inserts 1,000,000 (one million) of pairs Key/Value. Consider that inserting pairs require much more effort than inserting records since records are indexed by the key.

Source	Database type	Database location	Total time elapsed (less is better)	Requests per second (more is better)	Tested version
Source	ODatabaseFlat	memory	16.89 seconds	59,017	0.9.9
Source	ODatabaseFlat	local	86.93 seconds	11,502	0.9.8