Organizing Workloads — PDE Study Guide

Bundled into the existing Professional Data Engineer premium course — no separate purchase.

Included in this chapter:

Capability	Standard	Enterprise	Enterprise Plus
Compute model	Autoscaling only	Autoscaling + baseline	Autoscaling + baseline
Slot commitments (1yr / 3yr)	Not supported	20% / 40% discount	20% / 40% discount
Idle-slot sharing	No	Yes	Yes
Set max concurrency target	No	Yes	Yes
Managed disaster recovery	No	No	Yes
Best fit	Light or experimental use	Steady production analytics	Regulated or DR-critical analytics

Cheat sheet

A slot is BigQuery's unit of compute, bought either on-demand or by capacity
Capacity-based billing comes in three editions chosen per reservation
Set the baseline to your steady floor, autoscaling to cover the spikes
Autoscaling adds slots in multiples of 50
Idle-slot sharing is ON by default; turn it off with ignore_idle_slots
Only baseline and committed slots are shareable, never autoscaled ones
Commit steady slots for a 20% (1-year) or 40% (3-year) discount
Assignments route a project, folder, or org to a reservation by job type
The most-specific assignment wins down the resource hierarchy
A project with no QUERY assignment silently falls back to on-demand
INTERACTIVE is the default and counts toward the concurrency limit
Run deferrable work as BATCH so it queues instead of competing
An unstarted batch job times out at 24 hours, it is not promoted
Queues cap at 1,000 interactive and 20,000 batch jobs per project per region
Isolate competing workloads with separate reservations and assignments
Buying more slots does not fix a single query that scans too much
Isolate BigQuery tenants by dataset-per-tenant, not project-per-tenant, to scale to thousands
Spread BigQuery concurrency across compute-tier projects sharing one reservation

Unlock with Premium — includes all practice exams and the complete study guide.

Cheat sheet