> ## Content Index > Fetch the complete content index at: https://www.edgewisely.com/llms.txt > Use this file to discover other available public pages before exploring further. # DynamoDB vs Cassandra: Which to Pick in 2026? - URL: https://www.edgewisely.com/dynamodb-vs-cassandra/ - Published: 2026-11-12T06:30:22.000Z - Updated: 2026-11-12T06:31:48.000Z - Description: DynamoDB vs Cassandra, compared on data model, pricing, consistency and ops. A neutral 2026 verdict on which NoSQL database to pick, and when. - Author: John Karpentar - Tags: Engineering, Cloud - **DynamoDB vs Cassandra** is a managed-versus-self-hosted call. **DynamoDB** is AWS's fully managed, serverless key-value/document store. **Cassandra** is an open-source, self-managed wide-column database. On-demand DynamoDB bills **$0.625 per million writes** and **$0.125 per million reads**; Cassandra has **no license fee** but you pay for nodes and operations. - **No, DynamoDB is not based on Cassandra.** Both descend from Amazon's 2007 Dynamo paper, but they are separate systems with different data and consistency models. - Current releases: **Apache Cassandra 5.0.9** (Aug 2026), adding Storage-Attached Indexing, vector search and Unified Compaction Strategy. DynamoDB offers global tables, Streams and free default warm throughput. - **Pick DynamoDB** if you live inside AWS and want zero ops. **Pick Cassandra** for multi-cloud, no lock-in, or predictable write economics at very high scale. DynamoDB vs Cassandra comes down to one choice: a fully managed AWS service versus open-source software you run yourself. DynamoDB is serverless and wins on zero operations inside AWS. Cassandra wins on portability, no vendor lock-in, and avoiding per-request costs at extreme scale. Neither is based on the other. Most pages ranking for this query sell a competing product or a tooling layer. This one doesn't. Edgewisely has no database to push, so the verdict below is about fit, not sales. Both sit in the broader [NoSQL field we mapped separately](https://www.edgewisely.com/top-7-nosql-databases-2026/). ## Is DynamoDB based on Cassandra? No. This is the most common misconception about the two, and the answer is a clean no. Both trace their lineage to one document: Amazon's [2007 "Dynamo" paper](https://www.allthingsdistributed.com/files/amazon-dynamo-sosp2007.pdf?ref=edgewisely.com), which described a highly available key-value store built on consistent hashing and eventual consistency. That paper influenced a generation of distributed databases. Cassandra came out of Facebook. Its creators took Dynamo's distribution and replication ideas and married them to a Bigtable-style wide-column data model, then donated it to the Apache Software Foundation. DynamoDB is a different animal. AWS launched it in 2012 as a fully managed service. It shares the Dynamo name and some ancestry, but it is proprietary, managed, and architecturally distinct from both the original Dynamo and from Cassandra. So: shared grandparent, not parent and child. They solve similar problems with different trade-offs. ## DynamoDB vs Cassandra: the differences that matter The core split is **managed service versus software you operate**. DynamoDB is a serverless AWS product; Cassandra is a database you deploy on your own nodes (or pay a vendor to run). | Dimension | DynamoDB | Apache Cassandra | | -------------- | ----------------------------------------------- | ------------------------------------------------ | | Type | Fully managed, serverless (AWS only) | Open-source, self-managed (any cloud or on-prem) | | Data model | Key-value + document | Wide-column | | Latest version | Continuously updated service | 5.0.9 (Aug 2026) | | Consistency | Eventually or strongly consistent (per request) | Tunable per query (ONE, QUORUM, ALL, etc.) | | Scaling | Automatic; on-demand or provisioned | Add nodes to a peer-to-peer ring | | Query/indexing | Secondary + global secondary indexes | Storage-Attached Indexing (SAI), vector search | | Ops burden | None (AWS runs it) | You run it, or buy managed (Keyspaces, Astra) | | Pricing model | Per-request + storage | Node/infra cost + operational effort | | Lock-in | High (AWS proprietary) | Low (open source, portable) | Cassandra 5.0, [GA since September 2024](https://cassandra.apache.org/%5F/blog/Apache-Cassandra-5.0-Announcement.html?ref=edgewisely.com), added the features that close its biggest gaps: **Storage-Attached Indexing** for flexible querying, a **vector data type with approximate-nearest-neighbor search** for AI workloads, and the **Unified Compaction Strategy**. It also moved to **JDK 17**, which the project says brings performance improvements "of up to 20% in some cases." The latest patch is [5.0.9, released 2026-08-07](https://cassandra.apache.org/%5F/download.html?ref=edgewisely.com); the 4.1.12 and 4.0.21 lines are still maintained. DynamoDB's [current feature set](https://aws.amazon.com/dynamodb/pricing/?ref=edgewisely.com) covers on-demand and provisioned capacity, multi-region **global tables**, change capture via **DynamoDB Streams**, and **warm throughput** values that are free by default and only charge when you pre-warm. ## What does each one cost? The pricing models are not comparable line for line, so compare the shape, not a single dollar total. **DynamoDB** charges per request plus storage, with no servers to run. On the [official AWS pricing page](https://aws.amazon.com/dynamodb/pricing/on-demand/?ref=edgewisely.com), on-demand in US East (N. Virginia), Standard table class: - **$0.625 per million write request units** (1 WRU = one write up to 1 KB) - **$0.125 per million read request units** (1 RRU = one strongly consistent read up to 4 KB) - **$0.25 per GB-month** of storage - Always-free tier: **25 GB storage, 25 provisioned WCUs, 25 RCUs** per region Provisioned capacity runs **$0.00065 per WCU-hour** and **$0.00013 per RCU-hour**, and reserved capacity can cut that further for steady workloads. **Cassandra** has **no license fee** — it's Apache-licensed open source. Your cost is the infrastructure (nodes, storage, network) plus the engineering time to run a distributed cluster: deployment, repairs, upgrades, monitoring, on-call. If you don't want that burden, managed Cassandra (Amazon Keyspaces, DataStax Astra) shifts it back to per-use billing, which narrows the gap with DynamoDB. The honest summary: at low-to-moderate, spiky traffic, DynamoDB's pay-per-request model and zero ops usually win on total cost. At sustained, very high request volumes, per-request pricing adds up, and a well-run Cassandra cluster on your own nodes can be cheaper — if you have the team to run it. ## Which is better for high write throughput? Both are built for heavy writes; the right pick depends on who operates it. Cassandra's masterless, peer-to-peer architecture means every node accepts writes, and you scale by adding nodes with no single bottleneck. That makes it a strong fit for write-heavy, globally distributed workloads where you control the hardware. DynamoDB also absorbs massive write volume, and on-demand mode scales automatically with no capacity planning. The trade-off is cost: at sustained extreme write rates, **$0.625 per million writes** becomes a real line item, whereas Cassandra's write cost is bounded by your node count. Verdict: for raw write scale with a capable ops team, Cassandra. For write scale without an ops team, DynamoDB. ## What this means for you Match the database to your constraints, not the hype. - **If you're a solo dev or small team on AWS:** Pick **DynamoDB**. Zero ops, a generous free tier, and it disappears into your stack. Don't run Cassandra yourself for this. - **If you're a platform team optimizing cost at high, steady volume:** **Cassandra** on your own infrastructure can undercut per-request pricing — but only budget it if you have the people to operate a distributed cluster. - **If you need multi-cloud or want to avoid AWS lock-in:** **Cassandra**, every time. DynamoDB does not leave AWS. - **If your real question is relational scale:** this is the wrong pair — compare [CockroachDB vs Postgres](https://www.edgewisely.com/cockroachdb-vs-postgres/) instead. - **If you want Cassandra's model without the ops:** Use managed Cassandra (**Amazon Keyspaces** or **DataStax Astra**) and accept per-use billing. - **If you're building AI/vector features:** Cassandra 5.0's native vector search is worth a look; otherwise evaluate DynamoDB alongside [purpose-built vector databases](https://www.edgewisely.com/pgvector-vs-qdrant/). ## Frequently Asked Questions ### What is the difference between Cassandra and DynamoDB? DynamoDB is a fully managed, serverless key-value and document database that runs only on AWS. Cassandra is open-source, wide-column, and self-hosted on any infrastructure. DynamoDB removes operational work in exchange for lock-in and per-request pricing; Cassandra gives portability and control in exchange for running it yourself. ### Is DynamoDB based on Cassandra? No. Both descend from Amazon's 2007 Dynamo paper, but they are separate systems. Cassandra originated at Facebook and is open source under Apache; DynamoDB is AWS's proprietary managed service, launched in 2012\. They share design ancestry, not code, and differ in data model and consistency. ### Is DynamoDB cheaper than Cassandra? It depends on scale and team. DynamoDB has no license fee and no ops cost, so at low or spiky traffic it's usually cheaper overall. At sustained very high volume, DynamoDB's per-request charges ($0.625 per million writes) can exceed the cost of self-run Cassandra nodes — if you have the staff to operate them. ### Can DynamoDB replace Cassandra? Often, yes, if you're committed to AWS and can map your access patterns to DynamoDB's key-value model. It won't replace Cassandra where you need multi-cloud deployment, on-premises hosting, no vendor lock-in, or Cassandra's tunable per-query consistency. Migration also means rewriting data models and queries, which is non-trivial. ### Which is better for high write throughput, Cassandra or DynamoDB? Both handle heavy writes well. Cassandra's masterless ring lets every node accept writes and scales linearly as you add nodes, which suits write-heavy workloads where you control the hardware. DynamoDB scales writes automatically with zero ops, but per-request pricing makes sustained extreme write volume more expensive. **Editor's note — sources:** DynamoDB on-demand prices, storage rate, and free tier: AWS DynamoDB on-demand pricing page (aws.amazon.com/dynamodb/pricing/on-demand/). Capacity modes, global tables, Streams, and warm throughput: AWS DynamoDB pricing page (aws.amazon.com/dynamodb/pricing/). Cassandra 5.0.9 (2026-08-07) and maintained 4.x lines: Apache Cassandra download page (cassandra.apache.org/\_/download.html). SAI, vector search, Unified Compaction Strategy, JDK 17 and the September 2024 GA: Apache Cassandra 5.0 announcement (cassandra.apache.org). Shared lineage: Amazon's Dynamo paper, SOSP 2007 (allthingsdistributed.com). Prices are US East (N. Virginia), Standard table class, read 2026-11-12, and change over time.