Domingo, 8 de Fevereiro de 2009

Alternative serialization methods for WSTest

Since I've posted the WSTest results I've been wondering what the results would be when using serialization methods other than XML and what is really the difference between Aalto and SJSXP XML parsers

I've set up a test using JSON and native Java serialization. The JSON bits are handled by the Jackson processor because it's claim to fame is that it's fast and in addition to that it supports easy binding to Java objects

As always the code is available for download

WSTest (results in tps, higher is better)

Test Namexml - sjsxpxml - aaltojson - jacksonjava serialization
GetOrder-208662116321165011465
GetOrder-1004509512656517077
EchoStruct-209441127731478812914
EchoStruct-1004133538967967575
EchoList-208610129861388913649
EchoList-1003846512960677238

What are the conclusions?

For this tests Aalto seems to be about 30-40% faster than SJSXP (the bundled StaX parser).

Jackson is screamingly fast, it gives Java serialization a run for it's money, specially for small message sizes.

published by luisneves às 18:31
perm link | comment | add to favourites
1 comment:
De anikin a 17 de Fevereiro de 2009 às 17:25
It should not be impossible to beat Java serialization for simplistic requirements.

All the "infrastrsucture" Java puts up in order to serialize complete object graphs (not duplicating instances which are referenced twice and so on) sure must weight.

Comment post

.search this blog

.Fevereiro 2009

Dom
Seg
Ter
Qua
Qui
Sex
Sab
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

.recent posts

. Alternative serialization...

. WSTest, some numbers - Up...

. A fast implementation of ...

. European PGDay 2008

. Need for speed

. Receiving events/notifica...

.archives

. Fevereiro 2009

. Janeiro 2009

. Dezembro 2008

. Outubro 2008

. Maio 2008

.tags

. todas as tags

.subscrever feeds