Database ManagementAmazon Web Services

Amazon DynamoDB Review 2026

4.4/ 5.0 from 5,640 verified reviews
Vendor
Amazon Web Services
Pricing
On-demand: $0.625 per million writes; reserved capacity available
Deployment
AWS regions globally; DynamoDB Local for development
Best For
Serverless apps, gaming, IoT, ad tech, session stores
Industries
Gaming, Media, Ad Tech, SaaS, Mobile
Implementation
Minutes to provision; weeks to model access patterns

Overview

Amazon DynamoDB is the fully managed NoSQL key-value and document database that AWS first released in 2012, evolved from internal experience with the original Dynamo paper. It is among the most scalable databases in commercial use, supporting Amazon.com's flagship shopping events with single-digit millisecond latency at any scale. DynamoDB pricing now extends from on-demand (no capacity planning required, post-November 2024 price reduction of up to 50%) through provisioned capacity with reserved discounts.

DynamoDB is the default operational database for serverless AWS architectures and is tightly integrated with Lambda, API Gateway, AppSync, and Step Functions. The 2024 introduction of Global Tables multi-Region strong consistency, the broader on-demand price cut, and ongoing improvements to PartiQL query support have widened its applicability. The hardest part of running DynamoDB in production has always been data modelling — single-table design and access-pattern-first schemas remain the steepest learning curve.

Key Features

  • Single-digit millisecond latency at any throughput with predictable scaling
  • Serverless on-demand mode with no capacity planning or provisioning
  • Provisioned capacity mode with auto-scaling and reserved capacity discounts
  • Global Tables with multi-Region active-active replication (strong consistency option, 2024)
  • DynamoDB Streams for change data capture into Lambda, Kinesis, or EventBridge
  • Point-in-time recovery (35-day rolling) and on-demand backups to S3
  • DAX (DynamoDB Accelerator) for in-memory caching of read-heavy workloads
  • Transactional reads and writes across multiple items with ACID guarantees
  • PartiQL SQL-compatible query language as an alternative to the native API
  • Item-level TTL for automatic expiry of session and ephemeral data
  • Encryption at rest with AWS-managed or customer-managed KMS keys
  • Native integration with IAM for fine-grained access control

Pricing

ModeModelCost (us-east-1)
On-demand writesPer million write request units$0.625 per million standard writes
On-demand readsPer million read request units$0.125 per million eventually-consistent reads
Provisioned write capacityPer WCU/hour$0.00065/WCU/hour
Provisioned read capacityPer RCU/hour$0.00013/RCU/hour
Reserved capacity (1-year)WCU/RCU, prepaid~53% discount versus provisioned
StoragePer GB/month$0.25/GB/month
Global Tables replicated writesPer million rWCU$0.9375 per million

Pricing verified May 2026 from AWS public pricing for us-east-1. On-demand pricing reflects the November 2024 reduction of up to 50%. Cross-region data transfer, PITR, and backup storage are additional. Reserved capacity requires 1-year or 3-year commitment.

Strengths

  • Operationally unmatched — true zero-administration database at internet scale
  • Predictable single-digit millisecond p99 latency under variable load
  • Tightest integration of any database with the AWS serverless stack
  • Global Tables with strong consistency is rare in commercial multi-region databases
  • On-demand mode eliminates capacity planning for variable workloads
  • Reserved capacity and the 2024 on-demand price cut materially improve cost predictability

Limitations

  • Steep data-modelling learning curve — single-table design is non-intuitive
  • No native joins, aggregations, or ad-hoc analytic queries (export to S3 for analysis)
  • Item size limited to 400 KB; secondary index limits constrain access pattern flexibility
  • Migration off DynamoDB is non-trivial — locks customers into AWS
  • Cost can rise rapidly with unindexed scans or poorly distributed partition keys

Alternatives

Multi-cloud document store with richer query semantics
4.5
Globally consistent relational alternative on GCP
4.3
Distributed SQL with PostgreSQL compatibility, cloud-portable
4.3
In-memory key-value with sub-millisecond latency
4.5
Distributed document store with built-in caching
4.2

Compare DynamoDB

DynamoDB vs MongoDB → DynamoDB vs Cosmos DB → DynamoDB vs Aurora →

Frequently Asked Questions

When should we choose on-demand versus provisioned capacity?
On-demand fits unpredictable or spiky workloads, dev/test, and most early-stage applications. Provisioned with auto-scaling becomes more economical at steady-state utilisation above ~30% — and Reserved capacity adds a ~53% discount on top. Many production teams use on-demand for the first 6–12 months and switch to provisioned once access patterns stabilise.
Can DynamoDB handle relational workloads?
With single-table design and careful access pattern planning, yes — many large applications use DynamoDB exclusively. The trade-off is design effort upfront. For workloads where ad-hoc reporting or evolving relational queries dominate, Aurora or PostgreSQL is a better fit. DynamoDB excels when access patterns are known.
How do we run analytics on DynamoDB data?
Three established approaches: export to S3 (full or incremental) and query with Athena; stream changes via DynamoDB Streams to Kinesis Firehose to S3; or use zero-ETL integrations to OpenSearch and Redshift. The 2024 DynamoDB zero-ETL to Redshift integration removed most of the historical operational overhead.
What is the realistic blast radius of a DynamoDB outage?
Single-Region DynamoDB outages are rare but have occurred (notably us-east-1 events in 2015, 2017, and 2023). For applications where multi-Region availability is required, Global Tables with strong consistency (launched 2024) is the recommended pattern. Active-passive with automated failover remains the most common topology in regulated industries.
Last updated: May 2026
Last updated: