Hi All,

I'm using Oracle 10g and Hibernate Core 3.2.6 on a complex app with quite a few classes (~90 at present, with 13 abstract classes [most near tree top], and ~200 classes planned). (The hibernate version used does not affect the memory usage.)

I switched to union-sub from table per class at around 30 classes and had to increase the heap space to 385. With the present configuration the heap must be greater than 768mb just to build the session factory. I had to pare down the set of entities in order to profile, but I confirmed that the space is not gc'd after factory creation. The vast majority of the space directly after factory creation was taken up by char[] elements (>90%).

I'm just wondering if this is a design flaw on my part (is union-subclass not meant / allowed to be 'deep') or if this is an issue with Hibernate itself.

With every new set of class additions the factory seems to grow by 10s of mb. I've had to increased heap space to 1g in order to cope with this issue (>768 to make the factory, plus space for usage), but I'm guessing this is not the way it was meant to operate!

I'm in the process of trying to debug this and / or create a set of hibernate objects to illustrate it and post a bug, but was just wondering....

Has anyone else run into this?

Thanks,
-Mike

I think you're really just running into some real issues that you get with mapping an object oriented system into a relational database. It gets tough. The fact that there are three different mechanisms for mapping inheritance just demonstrates that there is no, single, right answer for mapping inheritance, just wrong answers, with some wrong answers being less wrong than others, given a certain scenario.

Mapping Inheritance in Hibernate Tutorial: Benefits and Drawbacks