jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ard Schrijvers <a.schrijv...@onehippo.com>
Subject Re: combining data from relational db and jcr
Date Mon, 29 Nov 2010 10:08:24 GMT
2010/11/27 Fabián Mandelbaum <fmandelbaum@gmail.com>:
> Well, indeed I've heard that recommendation over and over again... but
> did you make tests with thousands of nodes and got bad performance? Or,

You don't need to test this as this is commonly know already. You will
get bad performance.

> I don't know about the 'cost' of doing such tests though, but I'd try
> 'em quickly because using the ID directly as the 'key' is far simpler
> than manipulating the ID everytime (which is also time consuming, even
> if computers do compute very fast).

I cannot say that it doesn't cost time, because obviously it does cost
some cpu cycles, but this won't be noticeable at all. Creating the
idea is like 1 billionth of the time it takes to actually create the
folders. I don't think the random folder structure is a bad idea. You
can also create the user 123 below

/users/1/2/3

Furthermore, I don't think you should store the jcr path back into the
database of the created user! You can much better store the identifier
of the created user node.

Regards Ard

>
> Just an idea. I'll let other, more experienced, JackRabbit guys jump in
> here if they have better ideas (I'm sure some will).
>
> El 27/11/10 08:50, danisevsky escribió:
>> Users may be tens of thousands and I think it is recomended "go deep"
>> instead of "go wide".
>>

Mime
View raw message