Glossary

AI Workload Placement

The decision about which system, operating environment, and partition executes an AI model or agent, separately from the IBM i application supplying its data.

What it decidesWhich system, environment and partition runs the model
Kept separate fromThe IBM i application that supplies the data
OptionsExternal service, Power CPU environment, Linux with an accelerator
Decided byMeasurements, not the presence of IBM i

What AI workload placement decides

Placement is the question of where the model actually runs. The IBM i application holds the data. The model that reads it can live somewhere else. Keeping those two decisions apart is the whole idea.

Three places an IBM i shop can run the model

The runtime options placement chooses between

  1. An external serviceThe model runs outside your data center and your data travels to it.
  2. A supported Power CPU environmentThe model runs on the same Power system using the processor, including on-chip Matrix Math Acceleration.
  3. A supported Linux and Red Hat partition with an acceleratorThe model runs beside IBM i on a Spyre card driven from Linux.

What the placement choice depends on

  • Measured latency between the data and the runtime
  • Data controls, meaning who may see the data and where it may go
  • Model requirements such as memory and framework
  • Power headroom left after production

Having IBM i on the box does not decide placement. The measurements do.

Related Terms

AI Inference Partition

A separate supported operating-system partition used to execute an AI model while an IBM i partition may remain the production application and data source.

Matrix Math Acceleration (MMA)

An on-chip acceleration feature in IBM Power10 and Power11 processors that speeds up the matrix math operations used in AI inferencing, without requiring a separate GPU.

IBM i

IBM's operating system for Power servers running traditional AS/400-lineage business applications, with Db2 for i built directly into the OS.
Back to Glossary