[TM4J-users] About implementing a hybrid memory-persistent topicmap backend

[Bent A. S. <be...@us...>]

Hi TM4J developers,

I use the TM4J library intensivly in one of the projects I'm working on,  
and I'm coming to you with a problem I know have been discussed on this  
list before; performance of the Hibernate backend. I realize that there is  
not very much that can be done about this on a short perspective, but I  
have suggestion. How about implementing a hybrid Topic Map Provider  
consisting a high-speed (read and write) memory backend, and lower speed  
persistent hibernate or ozone backend? When reading from a topic map with  
this hybrid backend, the reading is done accessing the memory backend.  
Writing operations are done first against the memory backend, and then  
pushed to some queue for writing to the persistent backend in an own  
background thread. I believe the persieved performance of such a solution  
will be similar to that of the memory backend. When loading the topicmap  
on application startup, the topicmap is first loaded from the persistent  
backend, and then pushed into the memory backend (do not know if there are  
any easy ways to do this).

I'm just airing my thoughts here, and I do not know if this is a major  
undertaking as I'm not all that familiar with the TM4J code base. Anyways,  
keeping two topicmaps synchronized the way I suggest might turn out to be  
challenging.

I would really like to hear your thoughts - and the thoughts of anybody  
else reading this list - about my suggestion. Am I way off?


Best regards,
Bent