KNIME · Infrastruttura · YoctoIT tech page

Executor

The elastic execution of the workflows: the heavy jobs scale on the nodes, separated from those who design them.

FOCUS · THE POWER WHERE NEEDEDDedicated execution contexts and scaling: the heavy workflow doesn't block the others
Request the Free Assessment Learn more
YoctoIT material for clients and partners · KNIME and the products mentioned are trademarks of KNIME AG.
01 · What it is

Executor scalabili, made clear.

In the Hubs, the execution is separated from the design: the executors are the engines that grind the workflows — sizable, dedicable per team or workload, scalable on the Kubernetes nodes. The 30-million-row data model runs on its big executor; the light report doesn't queue behind it.

Dedicati
execution contexts per team/workload: the separate lanes
Elastici
the executors scale on K8s: the power follows the queue
Isolati
the runaway job doesn't sink the platform: the blast radius contained
Executor scalabili
OFFICIAL KNIME BRANDING · EXECUTORS
CONSOLE REALE · EXECUTION CONTEXT · FONTE: KNIME DOCS
REAL CONSOLE · EXECUTION CONTEXT · SOURCE: KNIME DOCS
02 · How to use it well

The things that make the difference.

The engine

The workflows in the queuereports, ETL, models
Executor S
Executor L
GPU/dedicated executor
the sizes for the workloads
Scheduling & queueswhat runs where, when
Kubernetes nodesthe iron underneath, elastic
Every workload in its own lane

Workload profiling

The workflows measured (RAM, duration): the executor sizes tuned on facts.

Lanes by priority

The controlling's month-end doesn't wait for the data scientist: the queues with precedence.

An organized night shift

The heavy at night, the interactive by day: the calendar that exploits the iron.

The cluster's capacity

The K8s nodes observed: the executors' growth planned with the trends.

03 · In depth

The executors: where the flows run

The executors run the workflows on the Hub: they run on Kubernetes with sized execution contexts (CPU/RAM per team or use case), they scale horizontally with the queue, they get dedicated to the workloads (the heavy nightly ETL doesn't touch the interactive data apps), the colocation with the data cuts the transfers; the sizing is a craft: few fat executors or many thin ones, it depends on the flows — the queue gets monitored and tuned.

  • Execution context — the resource pen: CPU and RAM guaranteed per team
  • Scala orizzontale — more executors under load: the queue doesn't pile up
  • Dedicati — the interactive separated from the batch: the data apps always responsive
  • Colocation — the executor close to the data: the gigabyte doesn't cross the WAN
  • Sizing per profilo — in-memory vs streaming flows: the RAM gets designed
  • Monitoring — queues and times observed: the tuning with numbers
04 · Numbers and lifecycle

The numbers that matter.

K8s
the execution platform
n
executors per context: you scale by queue
RAM
the KNIME flows' critical resource: it gets sized
0
contention with the separated contexts
The flows' performance is sizing: contexts, resources and queues tuned by us — the analysis that finishes when it should.
05 · Use cases

Where it really pays off.

Voluminous ETL

The heavy transformations, without collateral victims.

Month-end peaks

The extra power when the calendar tightens.

Multiple teams

Each team its own engine, the platform at peace.

The platform is judged under load: the right executors keep it fluid — we tune them.