SYSTEM Cited by 4 sources
Amazon SNS¶
Amazon SNS (Simple Notification Service) is AWS's managed pub/sub fanout messaging service. Publishers post to a topic; subscribers (SQS queues, Lambda functions, HTTP endpoints, email, mobile push) each receive their own copy. Commonly paired with systems/aws-sqs as the fanout half of a topic → queue-per-consumer pipeline.
Role for this wiki¶
As external delivery channel for dead-man's-switch heartbeats (2026-05-05)¶
Distinct from the fanout / event-bus role: SNS here is used as a destination for continuous dead-man's-switch heartbeats from an internal observability stack. Airbnb's Observability team routes an always-firing Alertmanager alert to a dedicated SNS topic, and a CloudWatch rate alarm on the topic pages on-call when the message rate drops — i.e., when the internal pipeline that generates the heartbeat has broken. SNS's value here is being on a different control plane than the observability stack: a K8s-cluster or mesh incident does not affect SNS's ability to accept the heartbeat. See concepts/dead-mans-switch and patterns/heartbeat-absence-as-alert-trigger. This is the first canonical wiki disclosure of SNS in a meta-monitoring / external-watchdog-channel role.
SNS appears as the fanout primitive in AWS-native producer/consumer pipelines. The canonical shape is S3 event → SNS topic → SQS queue-per-consumer, letting each downstream subscriber control its own queue depth + retention without affecting others.
SNS/SQS pairs as anti-pattern at org scale¶
Per-integration SNS topics + SQS queues between services is a natural first step for introducing async decoupling, but Amazon Key explicitly names it as "implemented on an ad-hoc basis, lacking standardization and creating additional maintenance overhead". At fleet scale SNS/SQS pairs become pub/sub without governance — no shared schema registry, no content-based routing rules, no subscriber-provisioning scaffolding. EventBridge is AWS's answer for org-scale event-driven architectures: one bus with content-based routing rules + a schema registry + reusable subscriber constructs (patterns/single-bus-multi-account, patterns/reusable-subscriber-constructs, patterns/client-side-schema-validation). SNS remains the right primitive for narrower fanout pipelines (S3 event → SNS → SQS per consumer) and as an EventBridge target.
Seen in¶
- sources/2026-05-05-airbnb-monitoring-reliably-at-scale — canonical wiki instance of SNS as the external termination channel for a dead-man's-switch heartbeat. Airbnb's Alertmanager continuously pushes an always-firing alert to a dedicated SNS topic; a CloudWatch rate alarm pages on-call when the heartbeat stops. The design property that makes SNS load-bearing here is its different control plane — the internal observability stack runs on dedicated Kubernetes clusters, but the watchdog path exits via AWS managed services so that cluster-level / mesh-level incidents can't silence the watchdog.
- sources/2024-07-29-aws-amazons-exabyte-scale-migration-from-apache-spark-to-ray-on-ec2 — Amazon BDT's 2021 serverless-Ray job management substrate uses SNS alongside systems/dynamodb, systems/aws-sqs, and systems/aws-s3 for durable job lifecycle tracking and notifications.
- sources/2026-02-04-aws-amazon-key-eventbridge-event-driven-architecture — Named in the "ad-hoc SNS/SQS pairs" anti-pattern that motivated Amazon Key's migration to EventBridge as the shared-bus substrate; SNS still valid for narrower fanout, but superseded at org scale.
- sources/2026-05-04-netflix-democratizing-machine-learning-building-the-model-lifecycle-graph — SNS (alongside Kafka + SQS) carries thin notification-of-change events from six Netflix source systems into MDS, which then hydrates full state from source APIs. Canonical wiki instance of SNS as the fanout primitive feeding a metadata-graph ingestion pipeline — one source-system event reaches all downstream consumers (MDS + others) via per-subscriber SQS queues without the producer needing to know who's listening.
Related¶
- systems/aws-sqs — canonical downstream-queue pair.
- systems/amazon-eventbridge — the org-scale event-bus abstraction that supersedes ad-hoc SNS/SQS pairs.
- systems/aws-s3 — common upstream event source.
- systems/dynamodb — common state store in the same pipeline.
- systems/aws-cloudwatch — rate-watchdog for the DMS role.
- systems/alertmanager — upstream Prometheus-ecosystem sender of the DMS heartbeat.
- concepts/dead-mans-switch
- concepts/meta-monitoring
- patterns/heartbeat-absence-as-alert-trigger