jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Klimetschek <aklim...@adobe.com>
Subject Re: Jackrabbit - storage documents ( file system / Mysql)
Date Fri, 16 Mar 2012 12:35:09 GMT
Is the original question about the DataStore (for binaries/files) or generally about the persistence
manager?

For (larger) binaries the file system based datastore will be a lot more efficient than databases,
as relational DBs are usually not made for binaries and the overhead they introduce does not
give any benefit for binaries, since you can't query for them (Jackrabbit has a separate full
text search index).

For the persistence manager, which also stores all the fine granular data such as small properties,
this is very different. The only simple file-system based PM was the XMLPersistenceManager,
which is very, very inefficient. The Bundle DB PMs are the most efficient ones in Jackrabbit
(but note that they also don't use much of the DB queries, they only ask for node bundles
based on the node's UUID as primary key, that's all).

Cheers,
Alex

On 16.03.2012, at 00:19, Mark Herman wrote:

> Agreed, in general it is accepted that the filesystem is going to be faster.
> Be aware of premature optimization though. There may be features given by
> the overhead of a DB that you'd want, but give it up for a difference that
> the users will never notice.
> 
> Also, make sure you avoid high number of children under one node if
> performance is important.  They recommend less than 10k per node, but I'd
> structure it so you're never even close to that limit.
> 
> --
> View this message in context: http://jackrabbit.510166.n4.nabble.com/Jackrabbit-storage-documents-file-system-Mysql-tp4474715p4476719.html
> Sent from the Jackrabbit - Users mailing list archive at Nabble.com.

-- 
Alexander Klimetschek
Developer // Adobe (Day) // Berlin - Basel


Mime
View raw message