RE: SQL vs. traditional I/O?

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: Erich W. Schasse
> 
> Joe,
> 
> what's about SELECT INTO ?
> 
> I'm sure, the ratio would no longer be 6:1.

Excellent, Erich!  This is the kind of thing I like to see!  It indeed
lowers the time for the SQL 30%, down to seven seconds.  But that's
still much slower than the native I/O.

The interesting thing about my test is that the first time you run the
RPG test after a period of system quiescence, you get a value of about
20-30 seconds for 100,000 records.  The second time drops to 15 seconds.
The third time it drops to seven seconds.  My guess is that this is
related to how DB2 caches data, since the performance improvement is
shared by others jobs.

However, no such effect occurs for the SQL.  The result is seven seconds
for 10,000 records, no matter what.

So, with repeated calls, the native I/O increases in speed from three
times the speed of SQL to ten times the speed.  I have no good answer
explanation for this, I'm just relaying the information.

Joe