I'm having a general performance tuning problem and I'm wondering if any of you have had a similar experience. We have a deeply nested object model, 4 or 5 levels deep in some places, with several objects having 5 or more one-to-many relationships and a handful of many-to-one. The number of queries that it is taking to pull in all the data we need is staggering. We have turned off lazy fetching where appropriate and tried batch-sizes of 10 and 100. We are still struggling for acceptible performance. My guess is that we are simply dealing with too many one-to-many relationships. Is Hibernate really limited to eager fetching only one of these per query?

If you have such a use case, could you please share some tips? Are there other strategies that we should be exploring, like a stored procedure or view combined with a flattened object?

I should also point out that most of the one-to-many relationships are more like one-to-a-few. It is more a problem of how many of these relationships there are at each level (thus driving up the number of queries).

No, Hibernate does not any longer have this limitation.

So you can use multiple outer join fetches.

Or you can try subselect fetching.


(Batch fetching is not a good way to solve this problem, as you can see.)

Analyze your model for relationships which should (nearly) always be eagerly loaded, and use the fetch="join" strategy for them. Stick with select fetching for relationships which are only rarely used.

Ensure that your cache is configured correctly.

Consider providing multiple load methods in your access layer; some will use the default (mostly lazy) loading, and some will eagerly fetch the relevant data (using Hibernate.initialize). You must have some context information that you can use to tune your mappings.

Just wondering, in what version was this added? I tried it before and could not get it to work. Or is there still a limitation in HQL to only one one-to-many fetch?