Running Cassandra Workloads as a Managed Service with Amazon Keyspaces - CQL Compatibility and Serverless

Wide-column data is managed with an Apache Cassandra-compatible CQL API, and serverless on-demand capacity brings the operational burden to zero. This article explains how to choose between this and provisioned mode, and multi-Region replication.

About 6 min readLast updated: 2026-10-31

Overview of Keyspaces

Amazon Keyspaces (for Apache Cassandra) is a fully managed wide-column database service compatible with Apache Cassandra. You can use CQL (Cassandra Query Language) and connect with existing Cassandra drivers and tools. It has a serverless architecture, and capacity management is not required after you create a table. In Keyspaces, AWS takes on tasks such as adding / removing nodes, compaction, repair, and backups that you would handle when running your own Cassandra cluster. Keyspaces uses a serverless model and provides CQL queries and compatibility with the Cassandra ecosystem.

Capacity Modes and Data Model

Keyspaces offers two capacity modes: on-demand mode and provisioned mode. On-demand mode is pay-as-you-go billing based on the number of requests and is suited to workloads whose traffic patterns are hard to predict. Provisioned mode sets read / write capacity units in advance and keeps costs down for predictable workloads. Combined with Auto Scaling, automatic adjustment according to load is possible even in provisioned mode. The data model requires the same partition key and clustering key design as Cassandra, and the choice of partition key influences performance and scalability. It supports the major Cassandra data types, such as static columns, TTL (Time to Live), and counter columns.

Multi-Region and Operations

With multi-Region replication, you place table replicas in multiple Regions and can read and write in a Region near your users. It is an active-active configuration in which both reads and writes are possible in each Region, and when performance drops in one Region, you can switch the connection to another Region and continue processing. Point-in-time recovery (PITR) lets you restore a table to any point in time in the past 35 days, making recovery from operational mistakes and data corruption possible. Encryption is enabled by default both at rest (AWS owned keys, customer managed keys) and in transit (TLS). You monitor request counts, latency, and throttling events with CloudWatch metrics and make use of them to adjust capacity.

Keyspaces Pricing

Keyspaces pricing consists of the capacity mode and storage. On-demand mode costs about $0.00000178 per write request unit (WRU) and about $0.000000356 per read request unit (RRU) (as of August 2026). Provisioned mode costs about $0.000742/hour per write capacity unit (WCU) and about $0.0001484/hour per read capacity unit (RCU). Storage costs about $0.25/month per GB. Multi-Region replication incurs additional charges for storage in each Region and for replication writes.

Operational Benefits of Serverless

When you run Apache Cassandra yourself, you need to build nodes, expand clusters, apply patches, and respond to failures. Keyspaces takes these on as a managed service and provides them serverlessly. Users can focus on table design and on reading and writing data without being aware that a cluster exists. Capacity is adjusted automatically according to the volume of access, so you do not have to secure capacity in advance to match peaks. With Keyspaces, you can leave cluster operation tasks to AWS while using Cassandra features.

Designing Multi-Region Replication

Keyspaces can use a configuration that replicates data across multiple Regions. If you configure it so that reads and writes are possible in each Region, users around the world can access a nearby Region, and you can keep latency low. When a failure occurs in one Region, you can continue processing in another Region. Because replication involves a slight delay, you design with an understanding of how consistency is handled when the same data is updated in multiple Regions at the same time. For services with users in many parts of the world, you use a multi-Region configuration when low latency and availability are required.

Integration with AWS Services

Keyspaces integrates with other AWS services. Authentication and access control are handled with IAM, and you can control permissions at the level of tables and operations. You monitor the operating status with CloudWatch metrics and can understand the state of throughput and throttling. Continuous backups let you restore to any point in time in the past, so you can prepare for operational mistakes and failures. With VPC endpoints, you can carry communication over a closed path without sending it out to the internet. You can handle authentication, monitoring, backup, and networking with standard AWS mechanisms.

Optimizing Capacity and Cost

Keyspaces lets you choose between two capacity modes: on-demand and provisioned. On-demand can be used without predicting access volume and is suited to workloads whose traffic fluctuates widely or is hard to read. Because you are billed only for what you use, charges increase when a high load continues steadily. Provisioned lets you set the expected load in advance and, combined with automatic scaling, follow fluctuations as well, so you can keep costs down for stable workloads. You understand your access patterns and choose a mode.

Summary

Amazon Keyspaces is a fully managed wide-column database compatible with Cassandra that you can query with CQL and that runs serverlessly. You can use an active-active configuration with multi-Region replication and restoration to any point in time within 35 days with PITR.

References (Official AWS Resources)

The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.

If this page and the official documentation disagree, treat the official documentation as authoritative.