PORRITTINC
( LAYTON, UT )
SPECIFICATION · HPC COMPUTE

Workload
meets plant.

A scheduler that knows plant state can move flexible work to match generation and thermal capacity rather than fighting them. That is only possible when one control layer sees both.

← Back to Run by Plant Intelligence

REFERENCE ARCHITECTURE · PROPOSAL

This describes a proposed reference architecture. It is not a shipping product and no vendor partnership is claimed or implied. Interfaces and standards conformance are specification intent; quantities are design targets.

Design basis
INPUTS

Plant state to scheduler

Generation availability, thermal headroom, and maintenance state are scheduler inputs rather than external surprises.

CLASSES

Flexible vs firm workloads

Work is classified by deferability. Flexible batch work absorbs variability; firm work is protected and scheduled against guaranteed capacity.

ACTIONS

Throttle, defer, migrate

Available responses are bounded and explicit. The scheduler operates within limits set by the plant, not the other way round.

BOUNDARY

Plant has priority

Where compute demand and plant safety conflict, plant constraints win. That precedence is designed in, not configured.

Design targets
1
Control layer
2
Workload classes
PLANT
Has precedence
PROPOSAL
Design stage
Integration notes

Requires the scheduler and the plant supervisory layer to share a data model. This is the principal software integration in the architecture.

Discuss the details.

Specifications evolve as the design matures. Contact us for current status and integration requirements.

CONTACT TIMOTHY → Back to Run by Plant Intelligence →