I would suggest you don't use the iBATIS cache and instead write your own. It's been proven to me numerous times that an application (i.e. business domain) level cache cannot be beaten in terms of performance. This is especially true considering your complex requirements.
Alternatively, you could try taking one of the open source caches (OSCache, EHCache) or even a commercial one (Tangosol Coherence) and modify it or extend it to your needs. Cheers, Clinton On Thu, 17 Mar 2005 15:33:23 +0200, Abdullah Kauchali <[EMAIL PROTECTED]> wrote: > Hi folks, > > We are piloting a few stateless services in our organisation and would > like to create > a Security Service that will satisfy requirements for User Session > Management. > > The idea is that once a user is authenticated against the Security > Service, a GUID token > is provided to the client. This token is then used by the client system > to negotiate services > across the different organisational public "API's" (thereby obviating > the need for > re-authentication at each call to a service.) The Session GUID is > persisted in the backend > database. > > One of the requirements of the Session Management architecture is that > of verifying > the session for: > 1) Validity (does the server recognise the GUID as one that it provided > in the recent past?) > 2) A timeout period (has the Session expired beyond a specified TIMEOUT > period.) > > We have already created the components that answers the two requirements > above without caching. > However, we would now like to optimise the process by using caching. > In this regard, we would like to > avoid/minimise the number of unnecessary roundtrips to the database to > "refresh" the Session details > for the user. As the number of concurrent users increases, the amount > of update hits against the > Session table will be considerable. > > To this end, the attached diagrams are what's being proposed. Please > see the attached images. > > Ok. Long winded. :) The question regarding iBatis: > > How do we achieve the process of "updating the cache" directly? > > Other general opinions and crits are also welcome, > > Kind regards > > Abdullah > > >
