Power Failures and such interruptions are something which only clustered app servers can handle and tide over.


In your case a power failure would rollback the transaction automatically because you would never have issued a commit and the broken database connection will cause the database itself to abort the transaction on its part.


The advantage of using a clustered setup using app servers is that the cluster manager would detect a failed app server on one node and migrate the transaction to a mirror which would go ahead with the transaction. This clustering concept is almost a science perfected by appservers and it would be too much to try and replicate all that effort by yourself.


Hope this helps,

Anin