SNS cost optimization (reduce deliveries and retries)

SNS costs are usually delivery-driven. The fastest way to save is to reduce deliveries (fan-out) and reduce retries (failures and timeouts). This playbook focuses on changes that reduce spend without breaking message semantics.

SNS cost levers

Reduce fan-out: tighten subscriptions and filters.
Retry control: fix failing endpoints to cut retries.
Message size: smaller payloads reduce transfer.

Step 0: baseline the two volumes (publishes vs deliveries)

Publishes/day by topic
Deliveries/day by topic and protocol
Failure rate and retry behavior (incident windows matter)
Average matched fan-out after filter policies

If you need a quick estimate first: estimate deliveries

1) Reduce fan-out (deliveries) without losing routing clarity

Use filter policies: deliver only relevant messages to each subscriber.
Split topics by intent: narrow high-fan-out topics so subscribers don’t receive irrelevant events.
Avoid broadcast payloads: send a pointer to an object instead of large payloads to many subscribers.
Remove dead subscriptions: stale endpoints create failures and retries.

2) Reduce retries by fixing endpoints (retries are a cost + reliability incident)

Fix slow/erroring HTTP endpoints that cause repeated delivery attempts.
Use dead-letter queues for failed deliveries where appropriate.
Set timeouts and backoff intentionally; don’t let failures keep retrying forever.
Alert on increased failure rate early; it’s cheaper to fix the endpoint than to pay for retries.

3) Control alert storms

Dedupe: one incident should produce a bounded number of notifications.
Rate-limit: cap notifications per key (service, customer, incident id).
Group: prefer “digest per minute” for noisy signals instead of per-event alerts.

4) Validate that savings didn’t move elsewhere

Optimizing SNS can shift load to downstream systems (SQS consumers, HTTP endpoints). Validate the end-to-end cost:

SQS request volume (if SNS fans out to SQS)
Downstream API costs and logging volume (retries often multiply both)
Error rates and latency (don’t “save” by dropping messages silently)

Quantify before/after

Deliveries saved = deliveries/day before − deliveries/day after
Retry reduction = failed deliveries before − after
Fan-out reduction = matched subscribers per publish before − after

Tool: SNS cost calculator

Common pitfalls

Reducing publishes but leaving fan-out unchanged (deliveries stay high).
Adding filter policies but not validating match rate (no real fan-out reduction).
Ignoring failures; retries can dominate “quiet” topics during incidents.
Building alert systems without dedupe/rate limits (inevitable storm later).
Not removing stale subscriptions (they create ongoing failure/retry cost).

Sources

A practical workflow to estimate SNS delivery volume: start from publishes, model matched fan-out (after filter policies), add a retry multiplier for failures, and validate with metrics so budgets survive incidents.

SNS pricing: what to model (publishes, deliveries, fan-out)

A practical SNS pricing checklist: publish requests, delivery requests by protocol mix, fan-out after filter policies, and the retry/alert-storm patterns that create surprise delivery volume.

AWS SQS cost optimization (high-leverage fixes)

A practical playbook to reduce SQS costs: reduce requests per successful message with batching and long polling, prevent retry storms and poison loops, and validate savings with sent/received/deleted metrics.

Estimate SQS requests (from messages and retries)

A practical workflow to estimate billable SQS request volume: start from messages/month, model requests per successful message (Send/Receive/Delete), and add the multipliers (retries, empty receives, poison loops) that cause spikes.

SES cost optimization (reduce volume, retries, and payload)

A practical playbook to reduce AWS SES costs: prevent duplicate sends, control retries and alert storms, reduce non-prod waste, and keep payloads small when they matter — with validation steps to protect deliverability.

SQS vs SNS cost: how to compare messaging unit economics

Compare SQS vs SNS cost with a practical checklist: request types, retries, fan-out, payload transfer, and the usage patterns that decide the bill.

FAQ

What's the biggest lever for SNS cost?

Reduce deliveries: lower fan-out, use filter policies, and avoid broadcasting to many endpoints when only a subset needs the message.

How do I prevent incident-driven SNS cost spikes?

Control alert storms (dedupe, rate limit) and fix delivery failures that trigger retries. Treat high retry volume as both a reliability and cost incident.

What commonly increases deliveries?

High fan-out topics, duplicated notifications, and retries to failing endpoints. A small failure rate at high volume can create a large delivery count.

How do I validate optimizations?

Track publishes, deliveries, failure rate, and retry volume before/after. Validate that filter policies and topic splits don’t break required consumers.

Last updated: 2026-02-07