MongoDB เทียบกับ DynamoDB: ข้อแลกเปลี่ยนของระบบคลาวด์เนทีฟ

AWS DynamoDB and MongoDB Atlas are both cloud NoSQL databases, but with very different philosophies. DynamoDB is AWS's fully serverless NoSQL service — no servers to manage, automatic scaling, pay-per-request pricing. MongoDB Atlas is a managed cloud service but still based on server clusters with richer operational control. The choice often comes down to: how much infrastructure control do you want, and how complex are your query patterns?

DynamoDB is truly serverless: there are no instances to choose, no cluster sizing decisions, no replica set configuration. AWS automatically scales read/write capacity based on demand, handles sharding invisibly, and guarantees single-digit millisecond latency at any scale. For teams that want zero operational overhead and tight AWS integration (IAM roles, Lambda triggers, Streams), DynamoDB eliminates an entire category of infrastructure decisions.

DynamoDB's primary access model is limited: every table has a partition key (required) and an optional sort key. Items are retrieved by exact partition key match, optionally filtered by sort key range. There is no general-purpose query engine — ad-hoc queries require either a Global Secondary Index (GSI) designed at table creation, or a full table Scan (expensive and slow). This forces access-pattern-first schema design, similar to Cassandra.

MongoDB Atlas supports the same rich query model as self-hosted MongoDB — aggregation pipelines, $lookup joins, text search, geospatial queries, and arbitrary field filters with compound indexes. Query patterns can evolve after launch by simply adding a new index. This flexibility is a major advantage for applications where requirements change frequently or are not fully known at design time.

DynamoDB offers two pricing modes: On-Demand (pay per read/write request — great for unpredictable traffic) and Provisioned (reserve capacity units in advance — cheaper for predictable workloads with auto-scaling). MongoDB Atlas charges by cluster tier (instance size + storage), making costs more predictable at consistent load but higher at low traffic. For bursty or event-driven workloads with zero traffic between events, DynamoDB's pay-per-request can be dramatically cheaper.

DynamoDB integrates natively with the broader AWS ecosystem. DynamoDB Streams emit change events that trigger AWS Lambda functions — identical to MongoDB Change Streams but within AWS. IAM authentication means no separate database credentials to manage. Point-in-time recovery and backup are configured with a few clicks. For teams already deeply invested in AWS (Lambda, API Gateway, Cognito, S3), DynamoDB minimises context switching.

MongoDB Atlas runs on AWS, Azure, and GCP — you are not locked into one cloud provider. You can deploy a multi-region cluster spanning AWS and Azure. Self-hosted MongoDB on your own servers is identical in behaviour to Atlas, so you can migrate between cloud and on-premises. DynamoDB is AWS-only; migrating away requires rewriting your data access layer. For organisations with multi-cloud strategies or regulatory data residency requirements, Atlas's portability is a significant advantage.

DynamoDB offers eventually consistent reads (default — lower latency, lower cost) and strongly consistent reads (higher latency, 2x cost). Writes are always strongly consistent within a single item. DynamoDB also supports transactions (TransactWriteItems / TransactGetItems) for atomic operations across multiple items — up to 100 items per transaction. These are similar to MongoDB's multi-document transactions but are billed at double the normal read/write unit cost.

Both databases support flexible, schema-free documents. DynamoDB items can be up to 400 KB; MongoDB documents up to 16 MB. For rich media metadata, long text content, or large nested structures, MongoDB's larger document limit is important. DynamoDB's 400 KB limit means large payloads must be split or stored in S3 with a DynamoDB reference — adding complexity. For typical structured data under 400 KB, the limit rarely matters.

Choose DynamoDB when: your team is AWS-native and wants zero infrastructure management; traffic is bursty or unpredictable (Lambda-backed APIs, event-driven systems); your access patterns are simple and well-defined upfront (key-value lookups and sorted ranges); or you need tight Lambda/IAM integration without managing connection pools. Single-page applications with a REST API and simple CRUD against DynamoDB are classic fits.

Choose MongoDB Atlas when: query patterns evolve or are complex (aggregations, joins, full-text search); your team needs multi-cloud or on-premises portability; documents may be larger than 400 KB; you need Atlas Search, Vector Search, or Data Federation; or your application already uses MongoDB locally and you want zero friction moving to cloud. MongoDB is also the stronger choice if rich analytics pipelines are a first-class requirement.

Test your understanding of MongoDB & NoSQL Databases concepts from this lesson.

In this lesson you learned: DynamoDB's serverless model and AWS-native integration make it ideal for bursty event-driven workloads with simple, pre-defined access patterns, MongoDB Atlas offers richer queries, larger documents, and multi-cloud portability that DynamoDB cannot match, and the key tradeoff is operational simplicity and AWS lock-in (DynamoDB) versus query flexibility and cross-cloud freedom (MongoDB). Next up we explore when to use a graph database like Neo4j instead of MongoDB.