>From: "bozon" <curtis@crowson1.com>
>I believe that you can unload to a named pipe with HPL and then load
>from that same named pipe. This is a very fast way of doing this. I
>think there might even be a program in the iiug repository that will
>set up the hpl job to do this so you don't have to go through the
>horrible HPL GUI. I think you can even create the jobs and run many in
>parallel.
>
Interesting.
Although you'd be using an SQL unload command to the named pipe and then use
the HPL to read from the named pipe.

In theory, you'd still be faster if you were to select and write the results
in to shareed memory, then have n readers from shared memory that write to a
tcp/ip socket on the remote machine and then each socket has x readers each
with a connection to the database.

I say in theory because the complexity of the code outweighs its value
unless you were just experimenting. You'd then be limited by hardware, and
the size of your network pipe.


But the question goes, how fast is fast enough?

Even unloading to a named pipe and then using HPL across a network may be
faster, you still have to consider the margin of speed vs the complexity.

Note: Some of the technology that I'm talking about could be found in the
RTL that may or may not still be marketed by IBM. (I think they discontinued
the Fin Foundation Bundle but not the individual products.)

But hey! what do I know? I'm paid to solve client's problems, not play and
develop tools that may or may not improve the quality of life for a DBA. ;-)

__________________________________________________ _______________
Communicate instantly! Use your Hotmail address to sign into Windows Live
Messenger now. http://get.live.com/messenger/overview