Re: svn commit: r1050283 [1/2] - in /archiva/trunk: ./ archiva-modules/metadata/metadata-repository-api/src/test/java/org/apache/archiva/metadata/repository/ archiva-modules/plugins/ archiva-modules/plugins/metadata-repository-file/src/main/java/org/

[Wendy Smoak <ws...@gmail.com>]

On Fri, Dec 17, 2010 at 1:26 AM,  <br...@apache.org> wrote:
> Author: brett
> Date: Fri Dec 17 06:26:09 2010
> New Revision: 1050283
>
> URL: http://svn.apache.org/viewvc?rev=1050283&view=rev
> Log:
> [MRM-1327] very rough prototype of a JCR based storage mechanism. Passes tests, but needs to be cleaned up and properly integrated, then tested for performance

Can you talk a bit about what you're doing here and how it's going?
Based on the commit message I thought, "You're doing *what* with my
artifacts?!" but I see from the issue description that it's for the
metadata and it's still file based. :)

-- 
Wendy

Re: svn commit: r1050283 [1/2] - in /archiva/trunk: ./ archiva-modules/metadata/metadata-repository-api/src/test/java/org/apache/archiva/metadata/repository/ archiva-modules/plugins/ archiva-modules/plugins/metadata-repository-file/src/main/java/org/

[Brett Porter <br...@apache.org>]

On 24/12/2010, at 12:22 PM, Wendy Smoak wrote:

> On Fri, Dec 17, 2010 at 1:26 AM,  <br...@apache.org> wrote:
>> Author: brett
>> Date: Fri Dec 17 06:26:09 2010
>> New Revision: 1050283
>> 
>> URL: http://svn.apache.org/viewvc?rev=1050283&view=rev
>> Log:
>> [MRM-1327] very rough prototype of a JCR based storage mechanism. Passes tests, but needs to be cleaned up and properly integrated, then tested for performance
> 
> Can you talk a bit about what you're doing here and how it's going?
> Based on the commit message I thought, "You're doing *what* with my
> artifacts?!" but I see from the issue description that it's for the
> metadata and it's still file based. :)

Thanks for asking!

There's nothing new here, just finally getting around to doing what I said I'd do earlier and experimenting a bit:

http://mail-archives.apache.org/mod_mbox/archiva-dev/201009.mbox/%3C6DD59AB2-2290-40D2-B632-8042621B9768@apache.org%3E
http://mail-archives.apache.org/mod_mbox/archiva-dev/201006.mbox/%3C64B9757E-33AC-42AE-A7B7-6D25929E3223@apache.org%3E
http://mail-archives.apache.org/mod_mbox/archiva-dev/201002.mbox/%3C574FF7F2-45B2-4B4B-9545-048D2B95FEDC@apache.org%3E

I haven't switched the "production" implementation in the webapp yet, so this is optional. I'm not convinced we can get the performance/memory characteristics we want out of it - but I want to try that with the two implementations and present the results here to decide. I'm sure we could get some help in tuning it as well.

My guess is that the result will be:
- much easier to work with than coding things ourselves
- better long running performance and memory characteristics
- a large initial hit / base requirement for the heap and higher "young generation" objects needing to be collected

If the latter is the only issue, then we might have a different deployment option for low-memory installations (eg. a proxy-only local machine installation that has no content repository).

I'm also rolling in some API changes that were driven by this. I don't think they're controversial, but I'll write them up and review it here. I was just doing that after the fact since I knew folks would be away for the holidays.

- Brett

--
Brett Porter
brett@apache.org
http://brettporter.wordpress.com/
http://au.linkedin.com/in/brettporter