Cloud Pub/Sub Complete Guide (2026)

Cloud Pub/Sub is GCP's global, serverless messaging service that loosely couples publishers and subscribers. It's the core of microservice-to-microservice communication, streaming pipelines, and IoT ingestion.

Core Concepts

Topic: The aggregation point for messages
Subscription: A subscription attached to a topic (Push / Pull)
Publisher: The message sender
Subscriber: The message receiver
Message: Up to 10 MB of data plus attributes
Ack Deadline: Processing grace period (default 10 s, up to 600 s)

Push vs Pull Comparison

Item	Push	Pull
Delivery	Pub/Sub -> HTTPS endpoint	Subscriber fetches
Scaling	Controlled on the endpoint side	Parallelized by subscriber count
Main use	Cloud Run / Cloud Functions	Custom apps, Dataflow
Rate control	Difficult (slow down via HTTP 429)	Easy
Exactly-once	Not supported	Supported (Pull only)

Pub/Sub vs Pub/Sub Lite

Item	Pub/Sub	Pub/Sub Lite
Scaling	Global, automatic	Zonal, capacity-reserved
Pricing	Pay-as-you-go (higher)	Reserved (cheaper, ~10x lower)
SLA	99.95%	99.5%
Main use	Microservices, IoT	Kafka alternative, log aggregation

Key Features

Ordering Key: Order guaranteed within the same key
Exactly-once delivery: Pull only; fully eliminates duplicates
Dead Letter Topic: Isolates messages that have failed n times
Message Retention: Up to 7 days (Lite supports up to 90 days)
Replay (Snapshot/Seek): Redeliver past messages
Schema Registry: Schema validation with Avro / Protocol Buffers
Filtering: Attribute-based delivery filtering on subscriptions
BigQuery Subscription: Writes directly to BigQuery (no Dataflow required)
Cloud Storage Subscription: Batches writes directly to GCS

Comparison with Other Messaging Systems

Item	Pub/Sub	AWS SNS+SQS	Kafka (Confluent)
Architecture	Serverless, global	SNS + SQS separate	Managed cluster
Ordering	Ordering Key	FIFO Queue	Within a partition
Throughput	High	Medium	Highest
Pricing	$40/TB	SNS $0.5/M + SQS $0.4/M	VM cost + license
Operations	None	None	Requires management

Typical Use Cases

Event-driven microservice communication (Order -> Email)
IoT telemetry ingestion (Device -> Pub/Sub -> Dataflow -> BigQuery)
Streaming ETL (Pub/Sub -> Dataflow -> Spanner/BQ)
Log aggregation (App -> Pub/Sub -> Cloud Logging)
Multi-subscriber fan-out (1 Topic -> multiple Subscriptions)
Cross-cloud integration (Webhook -> Pub/Sub -> AWS Lambda)

Should I choose Push or Pull?

Push delivers directly to an HTTPS endpoint (Cloud Run / Functions); Pull lets the subscriber fetch messages actively. Pull wins on scaling flexibility, Push wins on simplicity.

How does it differ from Pub/Sub Lite?

Standard Pub/Sub is serverless, global, and auto-scaled; Pub/Sub Lite uses reserved zonal capacity and is roughly 10x cheaper. Lite is a candidate Kafka replacement.

What is an Ordering Key?

Messages sharing the same key are delivered in order. Different keys are processed in parallel; within a key, order is preserved. Essential for use cases like chat messages.

Is exactly-once delivery possible?

Yes, with a Pull subscription and exactly-once enabled. Duplicates are fully eliminated at the cost of slightly higher latency.

What is a Dead Letter Topic?

A mechanism that forwards messages a subscriber has failed n times to a separate topic. Essential for isolating and reprocessing problematic messages.

How does it compare to AWS SNS/SQS and Kafka?

Pub/Sub unifies SNS (pub/sub) and SQS (queue). Versus Kafka, it wins on serverless, global, auto-scaling; Kafka still leads on raw throughput.

What is the pricing model?

Billed on ingest + delivery data volume ($40/TB). Pub/Sub Lite is billed per reserved throughput unit. Includes a 10 GB monthly free tier.

Is there a Schema Registry?

Yes, Pub/Sub Schemas supports Avro and Protocol Buffers schemas, guaranteeing schema consistency between publishers and subscribers.

Check what you learned with practice questions

Practice with certification-focused question sets

View GCP exam prep page

Author

NicheeLab Editorial Team

NicheeLab editorial team focused on data engineering and cloud certification learning. Content is structured around practical study needs and official exam domains.

Pub/Sub Complete Guide: Pricing, Push vs Pull, Ordering Key, Kafka Comparison