Coexistence Rules for Processing Nodes - Active Enterprise Data Warehouse

6750 Platform Product and Site Preparation Guide

Product
Active Enterprise Data Warehouse
Published
March 2018
Language
English (United States)
Last Update
2018-04-18
dita:mapPath
ras1471890041833.ditamap
dita:ditavalPath
6750.ditaval.ditaval
dita:id
jvo1471649303173
Product Category
Hardware
Coexistence Rules for Processing Nodes
Component Rules
Software All nodes (TPA and non-TPA) in the coexistence system must be running Linux SLES 10 SP3 or SLES 11, depending on which Teradata Database you are running. Before you add TPA nodes, the software and firmware on the existing TPA nodes must be at the same level as the new TPA nodes.
Node Generation There must be at least two nodes of a generation to be included in a coexistence system.
Cabinet All cabinets in a 6750HX or 6750X coexistence system are 47 inches deep. The following cabinet configurations apply.
  • 6750H nodes in expansion cabinets are added to 36-port BYNET V5 InfiniBand switches in 6700H cabinets, or to BYNET V5 InfiniBand switches in BYNET V5 InfiniBand cabinets.
  • 6690 nodes are added to 36-port BYNET V5 InfiniBand switches in 6750X base cabinets, or to BYNET V5 InfiniBand switches in BYNET V5 InfiniBand cabinets.
Cliques Nodes within a clique must be of the same generation (that is, all 6750HX, 6750X, 6700H, or 6690 nodes).

Cliques within a system can be of different generations.

Hot Standby Nodes Coexistence supports hot standby nodes.

Clique rules on the node type also apply to hot standby nodes.

Memory Nodes within a system can have different memory sizes, but memory sizes cannot be different within a clique.
AMPs The database can only use AMP storage to the capacity of the smallest AMP. In some cases due to configuration differences, data integrity features, or use of different drive capacities, there may be unused space on some AMPs. Some or all of that unused space may be usable on a future upgrade/expansion if the size of the smallest AMP is increased.

Additional memory may be required on the nodes within the faster cliques that have more AMP VPROCs. The faster nodes can then have at least the same amount of FSG-cache benefit per AMP VPROC as the slower cliques that have fewer AMP VPROCs.

New nodes should be configured so that they have an appropriate number of AMP VPROCs per node, depending on configuration and workload mix.