Amazon DynamoDB

Overview

Amazon DynamoDB is a cloud-based, managed NoSQL database for applications that need distributed storage without operating database servers. It is strongest for workloads built around key-based access, including internet-scale, mission-critical, agentic and IoT applications. Its main decision point is capacity billing: pay for requests as they arrive, or pay for provisioned capacity whether you use it or not.

Key features

Capacity and access patterns

DynamoDB handles server provisioning and patching, and tables can scale automatically. On-demand capacity can scale to zero and avoids capacity planning, making it a sensible fit for variable request volume. Provisioned capacity gives teams control over specified read and write capacity, with optional auto scaling, but can cost more than usage warrants when demand falls below the amount reserved.

Primary keys anchor data access; global and local secondary indexes let applications query through alternate keys. Native vector search supports similarity-search and agent-memory use cases, extending the service beyond conventional key-value access. Values are limited to 0.4 MB, so applications with larger individual values need another approach.

Resilience, security and recovery

Global tables support multi-Region, multi-active operation. AWS says they are designed for up to 99.999% availability, a strong target for applications that must remain available across Regions, though this is not a guarantee. Continuous point-in-time recovery provides per-second granularity and can restore data within a configured window of 1 to 35 days.

Data is encrypted in transit and at rest by default. IAM policies can restrict access down to particular items and attributes; AWS also describes KMS encryption and VPC endpoints through PrivateLink among its service integrations. AWS says its controls and compliance standards address sensitive requirements such as those of governments and global banks.

The database stores data on disk, supports TTL and works through HTTP(S), JSON and PartiQL. AWS positions it for modernization as well as mission-critical, internet-scale, agentic and Internet of Things workloads.

Pricing

DynamoDB is freemium, with a free tier and two paid capacity models. The choice matters: on-demand bills per read and write request consumed, while provisioned capacity bills for the capacity specified even when actual use is lower.

• Free tier — free. The account-based allowance is 25 WCUs, 25 RCUs, 25 GB storage and 25 replicated WCUs. The free tier uses provisioned capacity and the Standard table class; its request volume is described as about 200 million per month, depending on item size. A separate monthly allowance lists 25 WCUs, 25 RCUs, 25 rWCUs for global tables across two Regions, 25 GB storage, 2.5 million stream read requests and 1 GB data transfer out. These are bounded allowances, not a substitute for checking expected workload against capacity and storage needs.
• On-demand capacity — custom pricing. It charges for consumed requests, scales automatically and requires no capacity planning. It suits variable or difficult-to-predict traffic, but request-based billing makes usage the cost driver.
• Provisioned capacity — custom pricing. Customers specify read and write capacity and can enable auto scaling. This suits workloads with more predictable needs, but charges continue against specified capacity when consumption is lower.

The published plan terms give no paid price, so budget decisions should account for the difference in billing basis rather than assume one model is invariably cheaper.

Platforms

DynamoDB is a cloud deployment, with platform support listed for Android, iOS, Linux, macOS, Windows and web, alongside an API. That breadth is useful for teams connecting varied clients to a managed database, rather than installing the database locally.

Who it's for

Choose DynamoDB when a team wants AWS-managed distributed NoSQL storage, alternate-key indexes, multi-Region operation or per-second recovery without provisioning database servers. On-demand is the more natural fit when request levels vary; provisioned capacity is more appropriate when teams can plan capacity and value specifying it. The 0.4 MB value ceiling and the distinction between consumed-request and allocated-capacity billing make it less suitable when larger values or a simpler predictable cost basis are central requirements.

Pros and cons

• Pros: Managed operation removes server provisioning and patching, while on-demand can scale to zero.
• Pros: Primary keys plus global and local secondary indexes cover direct access and alternate-key queries; native vector search adds similarity-search and agent-memory workloads.
• Pros: Multi-active global tables and continuous recovery with a configurable 1-to-35-day window support resilience and restoration needs.
• Cons: Provisioned billing applies to specified capacity even when the application consumes less.
• Cons: Values cannot exceed 0.4 MB, which rules out treating it as a store for larger individual values.
• Cons: Paid capacity prices are custom, so the free allowances do not establish what a production workload will cost.

Alternatives

Consider Riak KV when an open-source option is preferable: its free Apache 2-licensed plan includes a key-value module, masterless built-in replication, HTTP API and Protocol Buffers. Aerospike Database is another freemium choice, with a Community Edition described as supporting 8 nodes, 2.5 TB of data and 2 namespaces; its Enterprise Edition is also offered at custom pricing.

For a free Redis interface, Redis Insight is described as 100% free, though the plan gives no stated limits. Choose RocksDB instead when an open-source C++ library under GPLv2 or Apache 2.0 suits the project better. VMware Workstation Pro is a free Windows or Linux host application for 64-bit Intel or AMD PCs, rather than a direct database substitute.

Cloudflare Workers KV may suit a smaller, bounded workload: its free plan allows 100,000 reads, 1,000 writes, 1,000 deletes and 1,000 list requests per day, plus 1 GB storage. Upstash Redis offers a free plan with 256 MB data, 10 GB monthly bandwidth and 500K monthly commands for readers whose workload fits those caps. Infinispan is a free Apache 2.0 option delivered as an embedded library or standalone server.

Browse Key-Value Databases, Multimodel Databases or Database Software to compare more options.

Verdict

DynamoDB is a strong choice for teams building distributed applications that benefit from managed scaling, indexed key access, multi-Region tables and granular recovery. Choose it when those operational and resilience benefits justify working within its value limit and capacity-based pricing trade-offs. Look elsewhere if larger individual values or a free-standing, predictable paid price matters more than AWS-managed scale.