Hi,

I want to enable second-level caching for my application and it appears very straightforward to configure in NHibernate. I am not sure about the different values for the usage property of the cache element.

Nearly all of my domain objects can be read and updated so does that mean that I should always use "read-write"? What are the actual differences between each of the settings and what impact will they have on the application?

Also, I am a little unclear as to how the different cache regions work with Session.Get() and Session.Load(). When using the ICriteria, you can set the cache region to use but I am not seeing how this is possible with the Get or Load so which of the cache regions will the object be added to?

Thanks,

Jason

partial answers...

there are three settings in NH:

nonstrict-read-write - no guarantee of consistency between cache and db. used if data doesn't change much (like days or weeks) and you can afford a small amount of risk of data inconsistency. impact: improves performance by reducing the database access overhead for objects that don't change often if used properly.

read-write - read committed isulation, non clustered envirnoments. used when the state of the object changes frequently and you need to make sure the cache is consistent with the db.mpact: improves performance by reducing the database access overhead but not as much as nonstrict. still makes a lot of sense to employ caching. i use this for Member (Person) objects.

read-only - used only for data that never changes. i use these for "month" and "country" objects as an example. They just don't change at all really and if they do, there is usually a new app deployed which means a restart. impact: all but eliminates the database access overhead for these types of objects.


session.get and session.load add their objects to the first-level cache. you cannot change that as far as i know.

hth,

-devon

Thanks Devon, that helps a lot. Is there a doc somewhere that explains the logic behind each of these settings?

In relation to the Get and Load, are you saying that these objects do not get added to the second-level cache? I do not want to prevent them from being added to the first-level cache but I assumed that they would also go into the second-level cache and wanted to know how to control which cache region they were added to as I was intending on having a couple of different regions.

Cheers,

Jason

cache info from the docs:

http://www.hibernate.org/hib_docs/nhibe ... he-mapping

also, there is a good description in the "Hibernate in Action" book. It is java-based but much of the information is invaluable for any Hibernate/NHibernate work.

You can configure the object for storage in the second-level cache using the <cache usage=""> property in the mapping file. Otherwise it simply gets stored in the first-level cache.

-devon

Yeah...I have already added the <cache> element to the entities but there doesn't seem to be any way to control which cache region they go into in the same way that you can for queries. I assume that they therefore go into their own cache region but it isn't clear if that is really the case. If they do go into their own region, how do you control the cache expiration?

BTW, I am using NHibernate.Caches.SysCache.

Thanks for your help.

Jason

oh, (doh!) now i see what you're asking. sorry to be so dense. each object has its own region, and each collection has its own region. (cache:classname or class:collectionName.) configuring them is dependent upon the cache provider. reffering to this documentation:

http://www.hibernate.org/hib_docs/nhibe ... s.SysCache

you would need to configure for each object and/or collection.



hth,

-devon

Aha! That's the ticket. Couldn't find this hidden gem in the doco's but maybe I'm just not looking in the right place!

Thanks a lot for all your help...I think we're done now :-)

Jason