|
|
Archives of the TeradataForum
Message Posted: Thu, 26 Dec 2002 @ 18:44:50 GMT
Subj: | | Re: Redistribute data |
|
From: | | Judge, James A |
One of those, "it depends" answers. Without knowing your configuration you could possibly just "join" new disks to existing vprocs.
This obviously increases the amount of GBs a vproc has available for new data rows. I suppose these new GBs could also be used for Fallback
or to meet spool requirements. But all these possibilities imply increased work per vproc, so the performance impact would probably not be
immediate but over time would cause each vproc to do more work, so to speak.
If the disks add "net new ranks" then you would be adding more vAMPs to the existing nodes. This causes more units of parallelism and so
you would have more amp worker tasks per node (and your current memory of your existing nodes). Performance impact would depend on the
"concurrent users" profile. If you have very few concurrent queries and utilities (not likely I would imagine) then more AMPs may increase
performance. The more likely impact though would be same/more concurrent queries with more AMPs would cause more "performance
opportunities" that need to be monitored and controlled.
| |