Skip to content

SerializationStrategyBenchmark

Jump to bottom Edit New page
Martin Grotzke edited this page Aug 24, 2015 · 1 revision

This page shows the results of the serializer-benchmark that compares java, javolution based and kryo based serialization/deserialization times for a sample object graph (some persons, addresses, calendar objects etc.).

The benchmark was performed on my T61 (Core Duo T7300, 2x2GHz) running Fedora Core 12 with Sun JDK 1.6.0_18-b07:

java -Xms512M -Xmx512M -classpath [..] de.javakaffee.web.msm.serializer.Benchmark

The plain numbers are published here: spreadsheets.google.com (all number are millis, also in the charts below)

A short explanation of the labels:

  • Ser - serialization time
  • Deser - deserialization time (you guessed it :-))
  • Min, Avg, Max - minimum, average and maximum time in millis required for 500 serializations/deserializations (10 cycles)

Benchmark small (10 person samples, 7 component samples)

http://spreadsheets.google.com/oimg?key=0AspDnQzTrkhddDRWeEl3LWs5R0UxbFB6MURmdHY1cGc&oid=12&v=1273186036253&foo=bar.png

Benchmark medium (100 person samples, 31 component samples)

http://spreadsheets.google.com/oimg?key=0AspDnQzTrkhddDRWeEl3LWs5R0UxbFB6MURmdHY1cGc&oid=10&v=1273186069983&foo=bar.png

Benchmark big (500 person samples, 261 component samples)

http://spreadsheets.google.com/oimg?key=0AspDnQzTrkhddDRWeEl3LWs5R0UxbFB6MURmdHY1cGc&oid=8&v=1273186101696&foo=bar.png

Serialized data size

The following chart shows the size of the serialized data for the different serialization strategies for 10, 100, 500 samples (here's the spreadsheet with numbers):

Add a custom footer
Add a custom sidebar
Clone this wiki locally