I'm currently planning cache strategy for my content delivery application with high scalability requirements. The basic operation goes like this: first a list of headlines is shown to the user. Once the user clicks one of the headlines, the application fetches the actual content item from the db and displays it to the user. I am using ehcache for content items, which is working fine. Headlines-lists are fetched using createQuery-method and since they are often requested with the same parameters, I'd like to cache them as well. From the documentation I have understood that query cache is the way to go but I have difficulties understanding how it actually works. Could someone please explain what the excerpt below means:

"Note that the query cache does not cache the state of any entities in the result set; it caches only identifier values and results of value type. So the query cache is usually used in conjunction with the second-level cache."

And also, how do I combine query cache with second level cache?

if you want to take advantage of cache even for


then use iterate() method instead of list()

iterate will ask the db only the id of the resulted objects, then it will iterate on each to obtain full data, it will first hit cache and then the db if the object is not cached.

Note that if the objects are not in cache... this strategy is very expensive.

Just take a look at sql generated...

Thanks for you reply. I was thinking about using iterator(), but was concerned about the performance. My hibernate code is hidden inside DAO objects, which contain methods that return a List object. The user of DAO object may then do several tricks with a List object such as sort it with certain criteria or combine it with another List object. So returning itarators from DAOs is out of question, instead, I would need to go through the Iteration inside DAO and put objects into a List, which would then be returned to the caller. It might be a bit ugly, but I'm limiting the size of Lists with setMaxResults() so I guess that would not cause much performance penalty though. However, I'm still concerned about iterator() hitting the database on each request for fetching ids, since that is not really necessary, I would rather return the already cached results and refresh the cache periodically.

I may be asking too much and would probably be best off by implementing my own caching for headlines lists, but letting hibernate to do the job would make life much easier (and code much cleaner). Anyway, suggestions and comments are welcome.

what you can write on your DAO layer is two methods:
getResults() which calls .list()

getCacheableResults() which call iterate() and transform it to a list...