When specifying the number of sessions and instances to meet the goal of best overall job performance, consider these other factors:
- Number of nodes
Larger database systems provide more AMPs, and by default make available more sessions.
- Number and speed of CPUs
- The greater the available processing power, the less need there may be for a large number of parallel sessions. A few connections (sessions) to a very powerful system can handle the same amount of throughput as a larger number of connections to a less powerful system.
- An imbalance between the throughput capability of the source and target system may reduce the benefits of using parallel sessions. The operator that interfaces with the more powerful system may spend excessive time waiting for the operator that interfaces with the less powerful system to complete its tasks.
- Workload
Always attempt to run a Teradata PT job at times when the workload leaves enough available AMPs for the optimal number of sessions.
- I/O bandwidth
- Structure of the source or target tables/records
- Volume of data being loaded