I am currently working on a project that utilizes the EAV model. This turns out to work quite well, but like many others I am now stumbling upon some performance issues.

The data set in this particular case consists of aproximately 2500 entities, each with aprox. 150 attributes. Each entity and each attribute is represented by a PHP-object.

Since most parts of the application only iterate through a filtered set of entities, we have not had very large issues yet. Now, however, I am working on an algorithm that requires iteration over the entire dataset, which causes a major impact on performance.

This information is perhaps not very much to work with, but since this is an architectural problem, I am hoping for a architectural pattern to help me on the way as well.

Each entity, including it's attributes takes up aprox. 500KB of memory.