Big Data
Pub/Sub
Fully managed, real-time messaging and event streaming
AWS equivalent
SQS + SNS / Kinesis Data Streams
AWS → GCP: Key Differences
- ▸
Pub/Sub is closer to Kinesis Data Streams for streaming or SNS+SQS combined for pub/sub patterns.
- ▸
Fully global by default: messages published in one region can be consumed globally.
- ▸
At-least-once delivery by default.
Key Concepts to Know
- 1
Topics (producers publish to) and Subscriptions (consumers pull from or push to).
- 2
Push subscriptions: Pub/Sub delivers messages to your endpoint. Pull: your service polls.
- 3
Dead letter topics: route undeliverable messages for debugging.
- 4
Message retention: up to 7 days.
- 5
Ordering keys: guarantee in-order delivery for messages with the same key.
DCE Interview Tips
- ★
Core pipeline anchor: 'Every real-time data architecture on GCP starts with Pub/Sub. It decouples producers from consumers and handles any ingestion rate.'
- ★
'For your IoT factory scenario: sensors → Pub/Sub. It absorbs traffic spikes. Dataflow then processes the stream.'
Common Gotchas
- !
Pub/Sub does NOT guarantee message ordering unless you use ordering keys.
- !
Messages are acknowledged per-subscription — if you have 3 subscriptions, each message is delivered to all 3.