[seam-dev] Page fragment cache with JBoss Treecache, ehcache and a FileSystemCache

[Sebastian Hennebrueder]

I uploaded the path to the current trunk

http://jira.jboss.com/jira/browse/JBSEAM-1891

The FileSystemCache is not faster than ehcache configured with 
diskoverflow, so I threw it away.

There are three cache provider: ehcache, jboss treecache 1.x, jboss cache 2
PojoCache is intended to be used with complex java objects. It is slower 
as compared to a plain treecache and we don't need the pojocache 
functionallity. So I did not write an implementation.

I added the dependency required to compile the patch to the core

There is one issue with the JBoss Cache 2 provider. I needed to use a 
deprecated API to make this run when libraries of JBoss Cache 1 and 2 
are provided. This is commented in the code.

JBoss Cache changes the API a lot. If I want to program this 
independently, I need to reflect everything (, instantiate the cache, 
build of cache node, put something in the node). This would impact 
performance.

Best Regards / Viele Grüße

Sebastian Hennebrueder

-----
http://www.laliluna.de
Laliluna.de, Berliner Strasse 22, 61118 Bad Vilbel, Germany

* Java Software Development, Support
* Training for Hibernate, EJB3 and Spring
* Tutorials for JSP, JavaServer Faces, Struts, Hibernate and EJB


Sebastian Hennebrueder schrieb:
> Christian Bauer schrieb:
>>
>> On May 03, 2008, at 11:48 , Pete Muir wrote:
>>>> FileSystemCache
>>>> Manually written, idea: a large site cannot be hold in a cache being 
>>>> present in memory. This cache has a small in memory area holding 
>>>> very frequent files (size can be configured) and fetches files from 
>>>> disk.
>>>> Though the operation system caches access to disk, this is still a 
>>>> lot slower than memory access but probably faster than having 
>>>> complex queries to the database (see Seam Wiki queries)
>>>
>>> I'm not quite sure why we need this?
>>
>> Given that ehcache supports configurable overflow to disk, I 
>> concur._______________________________________________
> 
> I will test the behaviour of EHCache using the Disk Overflow and give 
> you a feedback on this.
> 
> 
> 
> JBoss Cache 1.x and 2
> This won't be compilable at the same time. How shall I provide this, 
> because I would break the build.
> 
> I will provide the abstraction for the Pojo as well, though I do not 
> perfectly see the need. It is slower than the normal JBoss Cache (alias 
> TreeCache).
> 
> Best Regards
> 
> Sebastian
> 
> 
>