JMS and Database access

Question

hi all,
i have problem i have an application which use two database tables, our one external application store records in table A, my application continuous poll table for new entries as it get from it it do manipulation on it and store them in table B.
Someone suggest me that continuous DB table polling is not good approach rather then use JMS.
i am total novice in it do some one suggest me that is it good approach to use JMS?
if so please tell me some guide lines for programme.
Thanks & regards

Answer 1

By the sounds of it we implemented a similar solution. Clustered application read X number of records from the database using pesimistic locking (select for update in oracle), put them on a local queue and updated the status flag to prevent them being re-read.
Using pesimistic locking meant that only one node in the cluster could access the database at once, which is far from ideal for performance, but does at least ensure the messages were processed only once. Another component of the clustered application read messages from the queue and performed the CPU intensive bits. It was a complete lottery whether the messages were evenly distributed between the nodes, but the law of averages suggests that they will be.
If your database records have a predicatble sequence you could do something like...
1. Select next 10 keys from an identical sequence
2. Poll for records with only those keys
3. When you've run out keys get the next 10.
Providing you have lots of records being inserted that should allow you to have multiple clients processing them without duplicates, but you would obviously have to give the full solution some serious thought e.g. What do you do if one of the clients dies effectively orphaning up to 10 messages etc.