Archives of the TeradataForum
Message Posted: Wed, 09 Jul 2003 @ 16:55:22 GMT
I'm having some trouble getting my head around a Unix (node) connection vs. a Teradata connection and impact to the system, and I'm hoping to vet my ideas with this group.
Please validate, or explain the flaws in my thinking of the following:
For the purpose of the example, let's say I have a 10 node system, with a GIGE connection on node 10.
For users to access the GIGE, they need to connect via the IP for node 10. Other users connect via ODBC with all 10 IP addresses defined in their ODBC mapping. This will result in more users physically connected to node 10.
My understanding is that, regardless of what node you connect to, you have access to every PE in the system when connecting to Teradata, and Teradata manages workload via a round robin technique to decide which PE my session would actually use. So if I have 1 Lan PE per node, I can have 1200 sessions. I could have more than 120 Unix connections via node 10, and not run out of PE connections within Teradata.
So even though, there would be more connections to node 10, it would not impact the parallelism within Teradata.
My load balancing issues would only be at an I/O level - if say 100 users were each trying to return 2 million rows each (an exaggerated example to make my point) all connected via the same node 10.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|