Skip to content

Kundera Performance: With Indexing

vishri edited this page Dec 13, 2011 · 1 revision

Release: 2.0.4

Below mentioned performance numbers generated with indexing support over Kundera (which comes with extra cost).

Enviornment: Amazon AWS ubuntu large instance (ami-09965860).

Bulk Load:

Number Of Records(Single Thread) Pelops Time (in ms) Hector Time (in ms) Kundera (in ms)
1 25 40 93
1000 677 957 1719
4000 2131 2429 3901
40000 13009 13075 22158
100000 28884 30097 44876
1000000 256986 262067 375137

Concurrent Load:

Number Of Threads (1 record) Pelops Time (in ms) Hector Time (in ms) Kundera (in ms)
10 53 56 118
100 232 281 552
1000 1560 1823 2762
10000 9120 9776 14104
40000 28869 32331 39549
50000 36454 37172 50099
100000 68586 74304 95410

Concurrent Load + Bulk Load:

Number Of Threads (1000 rec/ thread) Pelops Time (in ms) Hector Time (in ms) Kundera (in ms)
10 4343 4738 9546
100 26010 24619 35196
1000 227232 219247 308975

Source Code: Source code for producing these results is available here.

Steps to run:

  1. Download jar.
  2. java -jar {Number of records} {client type} {process mode} { number of threads} -Xmx1024m -XX:MaxPermSize=128m (e.g. java -jar /home/impadmin/workspace/Perf_I81211_c.jar 1 pelops b -Xmx1024m -XX:MaxPermSize=128m

{client type} => 1) Pelops 2) kundera 3) Hector {process mode} => 1)c(Cnoncurrent) 2) b(Bulk) 3)cb(Concurrent+Bulk)

Disclaimer:

Above mentioned results are on the specified environment and use case and might have different outcome in other cases.

Clone this wiki locally