jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Nüßler <mark.nuess...@9elements.com>
Subject Re: session save and performance question
Date Sat, 24 Jan 2009 20:48:20 GMT
Hi Kevin,

got no expirience with mapping xml to a possible jackrabbit structure.

Think you should give your last thought a try. If you don't need every
information from your xml-file, just map these you are interestet in.
If you save the xml-file as additional property (Stream) to the 
content-node you are able to analyse the xml again. Maybe you
could have same kind of background-worker.

just a thought:
1. create a content-node and add the xml-file as blob,
    this shouldn't take much time
2. run a background-worker to do the hard stuff

got also no expirience with more than ~30 properties,
maybe you could have a look at the benchmark classes
of jackrabbit and write a performace test.

 > I'm now doubting if my case is the right use case for Jackrabbit.
^^ if your xml files have all the same structure and you query
always the same properties, a relational db is much faster.
but if you are dealing with different semistructured files
and want to be as flexible as possible - i don't know any better
than jackrabbit. i my usecase i must say i have just 'raped'
the rabbit ;-)


best regards

derMark



Cheng Zhang schrieb:
> Hi Mark & all,
> 
> What I'm doing currently is to use Jackrabbit to store XML document. Each XML element's
attribute will be mapped to a node property. And each XML text element, for example "<age>13</age>",
will be also mapped to node property. All other XML elements are mapped to a jackrabbit node.
> 
> I'm now doubting if my case is the right use case for Jackrabbit. In my case, one XML
document will be mapped to 500 node/child-nodes. If I have 1 million xml documents, it means
500M nodes will be created and indexed. 
> 
> Or I only map those nodes/properties I'm interested and in addition store the xml in
a file node.
> 
> Best,
> Kevin
> 
> 
> 
> ----- Original Message ----
> From: Mark Nüßler <mark.nuessler@9elements.com>
> To: users@jackrabbit.apache.org
> Sent: Friday, January 23, 2009 1:38:49 PM
> Subject: Re: session save and performance question
> 
> Hello Kevin,
> 
> i think you are right, when saying that saving the session is cost
> intensive. but ... it depends on ...
> 
> 1. how your application works internally
>     - do you really have to save the session every time
>       or can you work with a transaction concept ?
> 2. what kind of structure you use
>     - flat or hierachical
> 3. if you need/use mix:versioning
> 4. .... maybe others i've not mentioned here ?
> 
> i made some tests regarding 'invisible' structure nodes
> vs a flat hierarchy when adding 50k content nodes.
> [50k is a small amount within my current project]
> 
> it is always better to have a kind of structure above
> your content nodes, when you exceed x numbers of child-nodes.
> 
> because i haven't done all of the tests, i am not really sure
> what i would suggest for x - the userlist says ~10k, anywhere
> else (or was it an old entry ?) mentioned 4k.
> 
> @Kevin, if not done yet, read some of the old list-entries
> 
> 
> best regards
> 
> derMark
> 
> 
> Cheng Zhang schrieb:
>> I saw session saving costs too much time. I guest it might be caused by too many
nodes to be saved. what's the best practice to save the session to get the best performance?

>> Thanks a lot,
>> Kevin
>>
>>
>> ------------------------------------------------------------------------
>>
>>
>> No virus found in this incoming message.
>> Checked by AVG - http://www.avg.com Version: 8.0.176 / Virus Database: 270.10.12/1911
- Release Date: 23.01.2009 07:28
>>
>>
>> ------------------------------------------------------------------------
>>
>>
>> No virus found in this incoming message.
>> Checked by AVG - http://www.avg.com 
>> Version: 8.0.176 / Virus Database: 270.10.13/1912 - Release Date: 23.01.2009 18:54
>>

Mime
View raw message