Need opinions on data model

[cesco75]

Hi.
I'm designing a data model with following requirements:

1) Documents should be associated to a "person", that in Alfresco will by just a numerical ID. All other infos are stored in an external DB.
2) Some docs can be aassociated to more than one person.
3) I will need to manage 3-4 millions of docs
4) Documents associciated to a single person usually are below 100
5) I need to retireve documents just for a fixed person (ID).
6) Documents have other properties (like date, category, …) that I eventually use for stats, not for search.

I think the straight way to model it is a multi-value property on docs, but I have some concerns about performance.
Without requirement 2, a good trick  would have been to store contents in a folder structure where all docs of a single person are filed on same folder: in that a case a path search would have make all the work with no concerns on performance…

Any better solution?
Thanks,
Cesco

[jpotts]

I see three options:

1. Use a multi-value property that stores the person identifier (as you propose).

2. Use an association between the document object and an object representing the person.

3. Use a multi-value property on an object representing the person that stores the node refs of the docs.

The multi-value property approach will be more efficient from a search standpoint. Plus, it doesn't sound like you are storing anything about a person other than their identifier, so no need to waste an object on a person, which rules out options 2 and 3.

You said that a give person will only have about 100 documents related to them and that a given document could be related to more than one person. But how many people might a specific document be related to?

Jeff

[cesco75]

Sorry I did't answer before and thanks for your suggestions:

You said that a give person will only have about 100 documents related to them and that a given document could be related to more than one person. But how many people might a specific document be related to?
Jeff

98% of times 1 person is related to just one document but sometimes 1 document is related to a large number of people: tens to thousands in my current understanding

Cesco

[mrogers]

Your multi-valued field containing tens to thousands should be O.K.   But if it goes much above that I'd question the requirements.  In addition there is a Lucene limit of 1000 docs for a search.   I'd not like to see 10,000 or more values in a multi-valued field.

Where large numbers of users are associated with a document tends to be because they have a role rather than individual responsibility.  You could perhaps model that through group membership.

I'd be tempted to try to model the relationship with assocs (Jeff's option 2).   Just my preference

[unknown-user]

Hi,

How did you solve this finally? I have an exact same requirement. We are building a Policy Management site in Alfresco. Each policy document needs to know who all agreed to that policy, so that when a user logs in, he will only see the pending policies for him.

The nof users who agreed to a policy could be equvilaent to the nof employees in the company which would be 5K to 10K. From a performance stand point, whats the best option multi-valued property or an association between person object? Another suggestion I got the from the forum was to use Preferences service to store the list of policies agreed by a user but it has some cons.

Thanks