With 4 Millions Documents, is better use some folders or one?

Hello,

We are developing a Documentum-based solution that would initial load about 4 million documents (mostly pdf, but also doc and jpg).

We are assessing the possibility of using folders to group documents, possibly we would have a folder with 2.5 million and 1.5 million in 15 folders in a more or less homogeneous.

Is it more efficient to use multiple folders or just one?

The recording and access to documents we do it from a J2EE application with DFC's libraries, and always by the object_id.

Thank you very much for your help

Find more posts tagged with

Note:

If there are Accepted Answers, those will be shown by default. You can switch to 'All Replies' by selecting the tab below.

Accepted answers

sivarry

Hi,

If you have the object ids at your end, my humble suggestion is to keep your documents in those 15 folders only. Folders are used for classification and for faster browsing but in your case documents are accessed with object id and hence lesser number of folders is good. Also number/level of folders need not be related to the efficiency of application

-Siva

All comments

sivarry

Hi,

-Siva

vmcicm

Thanks Siva,

Please, one question more ... Are any limitation abaut the number of documents into one folder?

Thanks in advance,

Victor.

DCTM_Guru

There is no limitation

sivarry

Hi Victor,

No limitation.

One thing to be noted is, it will take long times to view the documents in WebTop or DA (as DQL and ObjectList might take long time to create records) but I am sure that is a rare case.

-Siva

stephanemarcon

Hi,

even if you do not need any categorization logic, I can bet you will access your repository with some webtop- like client at some point. I would recommend splitting those docs in a minimal folder hierarchy. if you do not need any hierarchy, you may go for a technical one, using for example the object_id or the chronical_id of the docs to classify them. i personnaly split the id by section of two chars, after Having removed the first 8 characters that identify the object_type and docbaseid.

but, ok, those are my 50 cents as there is nothing that really forces you to classify, technically, now I would consider whether you'll ease debugging and support of your future application with classifying a little bit.

the drawback of the id-based structure is one also has to consider a future migration to another docbase, in which case one cannot expect migration tools to keep original ids. Now, personnally this would not make me leave out such a classification but would make me think the best way to implement such a classification in the correspoding TBo.

those were my 50 cents which aren't certainly worth more than that.

Regards,

Stephane

lastnitescurry

take a look at

http://wordofpie.com/2007/08/13/taxonomies-good-bad-or-ugly/