[jboss-dev-forums] [Design of JBossCache] - Re: Space versus time, e.g. JBCACHE-1157

[jason.greene@jboss.com]

"manik.surtani at jboss.com" wrote : Eh?  1157 has nothing to do with a CHM in UnversionedNode.  I did try using a CHM at one point, which did "hide" the issue again, but this posed other problems (inability to store nulls in the node).  
  | 
  | It's back to HashMap in UnversionedNode now, and the correct fix for 1157 is in the PessimisticLockInterceptor.
  | 
  | 

I believe he was referring to the child node map which has to be a CHM, even with Pessimistic locking, since we do not lock the parent node for inserts / removes. We could alternatively use a synchronized map.

With MVCC we could restore parent node locking on insert / remove, and thus switch to a pain hashmap, since reads are non blocking and would no longer affect traversal.

-Jason

View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4081431#4081431

Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4081431