SQLServer jdbc批量执行性能

Question

I am using following pattern to execute jdbc batch on the SQLServer 2008 R2. The batch execute is simple MERGE statement on one table with proper indexes as per MERGE join clause.

batchSize = 50;
for(.....){
...
...
...

    //add to batch
    ps.addBatch();
    if (count >= batchSize) {
        result = ps.executeBatch();
        count = 0;
    }
}
if (count > 0) {
    result = ps.executeBatch();
    count = 0;
}

I have very large data pumping into for-loop. What i have observed is initially execution of batch of size 50 takes approx. 150 msec, and then it increases exponentially to 2.5 miunutes when reaches to counts of 400k records!

Is there some sqlserver specific tunning needed?

Answer 1

I got it to @30 minutes time for Insert data volume of 400k!!! I have to do following:

• Change MERGE to DELETE/INSERT (If kept MERGE it takes 1.5 hrs)
• Increase batch size to 1k
• Commit after every batch execute
• Call ps.clearBatch() after batch execute

That is it... :)