 |
 |
Archives of the TeradataForum
Message Posted: Mon, 08 Jul 2002 @ 21:35:56 GMT
| Subj: | | Re: Access Modules performance - NamedPipes(on NT) Vs MQ Series (on MVS) |
| |
| From: | | Jim Downey |
We do not have Tpump (yet) so I can not help you there. We do read MQ on the mainframe but we just read and clear the queue then
multiload during our batch window. We also have a process that exports from the Teradata data warehouse to an Oracle server using fast
export on 'our' side, a Unix named pipe in the middle then Oracle loader on the mart side.
The main reason I am responding is to share a lessons learned when faced with a similar decision. We were predominately a mainframe shop
but we had a new source to load (web log) that was on NT servers. We decided to leverage our skill sets and push the NT data to the
mainframe so it could pick up the data train. This decision turned out to be problematic as we had to move data back and forth (NT, MF,
Unix). We changed to process to load directly from the NT environment and have been happier ever since. I can't tell where your source
data is coming from but I suggest you weigh the number of times you have to move the data...sometimes the fewer number of moving parts, the
better....
Jim
| |
