Amazon Keyspaces (for Apache Cassandra)
AWS-managed, serverless **Apache Cassandra–compatible** wide-column NoSQL database service. Zero infrastructure management, pay-per-request billing, automatic scaling, point-in-time recovery (PITR) with 35-day window. CQL-API compatible so existing Cassandra application code works without modification, but the AWS managed service has functional differences from open-source Cassandra (transaction semantics, batch sizes, secondary indexes, change-feed format). Backup-and-analytics integration with S3 is via export/streaming pipelines feeding Athena or EMR — not a native S3 backup target.
Definition
AWS-managed, serverless **Apache Cassandra–compatible** wide-column NoSQL database service. Zero infrastructure management, pay-per-request billing, automatic scaling, point-in-time recovery (PITR) with 35-day window. CQL-API compatible so existing Cassandra application code works without modification, but the AWS managed service has functional differences from open-source Cassandra (transaction semantics, batch sizes, secondary indexes, change-feed format). Backup-and-analytics integration with S3 is via export/streaming pipelines feeding Athena or EMR — not a native S3 backup target.
Self-managing Cassandra at scale is operationally expensive — node lifecycle, repair scheduling, GC tuning, capacity planning, compaction tuning — and most teams running Cassandra-shaped workloads end up with a dedicated team just to keep the cluster healthy. AWS Keyspaces' bet: eliminate the entire operational layer behind a serverless API while keeping CQL compatibility, accepting the trade-off that some Cassandra-internal semantics (multi-row LWT, secondary-index behavior, specific compaction strategies) work differently or not at all. The 2026 framing has matured — Keyspaces is the answer for teams that want Cassandra's wide-column data model without the operational tax.
Wide-column NoSQL workloads at variable scale (per-request billing matches), event-store and time-series application backends, gaming session-state stores, IoT telemetry ingestion, migrations off self-managed Cassandra clusters where operational cost dominates infrastructure cost, and any workload where the CQL data model is the right fit but Cassandra's operational complexity is the wrong trade.
Recent developments
- July 1 2026: range delete operations transition from synchronous → asynchronous. Major operational change — applications that previously assumed range-delete returns mean the data is gone need to revise that assumption. Per AWS Keyspaces Cassandra-compatible setup guide.
- 35-day point-in-time recovery (PITR) — continuous backups. Built-in PITR protects against accidental write and delete operations; you can restore to any second in the preceding 35 days. Per AWS Keyspaces features.
- Serverless + pay-per-request billing model. Zero infrastructure management, zero-downtime maintenance, instant auto-scaling, pay only per request. The CQRS-style read-throughput-meets-write-throughput billing is fundamentally different from self-managed Cassandra's capacity-provisioned model. Per AWS Keyspaces homepage.
- Functional differences from Apache Cassandra are formally documented. AWS publishes a comprehensive functional-differences page — multi-row LWT works differently, secondary indexes have different semantics, certain compaction strategies aren't available. Per AWS docs — Keyspaces vs Cassandra and AWS docs — Functional differences.
- CQLReplicator solution for near-real-time migration from Cassandra. AWS published a guidance solution + the open-source CQLReplicator tool for near-real-time data migration from Apache Cassandra to Amazon Keyspaces, lowering the lift on customers replacing self-managed Cassandra clusters. Per AWS — CQLReplicator migration solution and AWS Database Blog — CQLReplicator.
Connections 3
Outbound 2
scoped_to1alternative_to1Inbound 1
alternative_to1