Hi,
I have a db2 udb v8.1 database on aix server.My front end is in Java. I need to store some xml files each of something like 600k. So we have to use clob/blob . The client does not want to use xml extender as he doesnt need to query the xml. But we need to retrieve the data and update it. So what should be the approach to be followed for performance tuning other than having a long tablespace? ie any query for selecting, inserting /updating should be completed in a very short time. Now lobs dont allow sorting/indexing. So what should be the approach?

Dpn't expect great performance for LOB's. The data is stored separately from the other data and does not use the DB2 bufferpools (data is not cached in memory). Use the fastest disk you can get for these.

The lob data is retrieved directly from the disk. In this case,would it be ok to use a single sms tablespace for the table and the lob column or should they be seperated into normal 4k ts and 4k bufferpool for all the other data and long tablespace and 32k Bufferpool for the lob data? Again is using blob advantagaeous over using clob?