Migrate to Apache Cassandra
This section provides practical, production-focused guidance for migrating to Apache Cassandra® from other distributions and platforms. Each migration source is documented as its own guide, tailored to its respective starting point.
The guides are written for experienced Cassandra operators, SREs, and DBAs. They
assume working familiarity with nodetool, JMX, and cassandra.yaml, and they
focus on real-world sequencing, decision points, and the edge cases that tend to
surface during production migrations. For reference documentation covering Apache
Cassandra itself, see the Apache Cassandra documentation.
Migrate with AxonOps support
AxonOps has planned and delivered migrations to Apache Cassandra from a wide range of source platforms, onto whatever target architecture and infrastructure fits the workload: any public cloud, on-premises, a hybrid of the two, or Kubernetes. AxonOps professional services can run the migration end to end or work alongside an in-house team, with 24/7 SLA-backed support in place once the cluster is live. To scope the work with an engineer who has run these migrations before, book an expert session.
Why migrate to open-source Apache Cassandra
Section titled “Why migrate to open-source Apache Cassandra”Organisations move from a proprietary or managed distribution to open-source Apache Cassandra for the following reasons.
-
Lower, predictable cost
No licence fees and no consumption-based billing. Infrastructure cost is bounded to what is actually deployed; migration is a one-time effort, whereas licence and consumption fees recur indefinitely.
-
No vendor lock-in
Apache-2.0 licensed and community-developed. Run any release, apply security patches directly, and choose tooling and support vendors freely.
-
Data sovereignty and control
Keep data exactly where residency and compliance regimes require, including regulated and air-gapped environments. Configuration and upgrade timing stay under operator control.
-
Multi-cloud, hybrid, on-premises
Run anywhere: across multiple clouds, in a hybrid cloud and on-premises topology, or on owned hardware. A single logical cluster can span datacentres and providers.
-
The latest Cassandra features
Cassandra 5.0 adds Storage-Attached Indexes (SAI), native vector search, the Unified Compaction Strategy, and trie-based memtables and SSTables, with Apache Cassandra 6.0 on the roadmap.
-
A mature, active ecosystem
In production across many industries, with a large contributor base, a public roadmap, and a choice of drivers, tools, and commercial support.
Operating Cassandra with AxonOps
Section titled “Operating Cassandra with AxonOps”Open-source Apache Cassandra removes licence cost and vendor lock-in. AxonOps supplies the operations tooling and SLA-backed support that a managed or proprietary distribution otherwise bundles, so leaving DSE does not mean losing day-2 operations or a support contract.
OpsCenter does not run on open-source Cassandra, and the usual replacement is an assortment of separate tools to wire together and keep maintained, where a team can spend more time on the tooling than on the database. A single AxonOps agent handles all of it, from repair and backup to metrics, dashboards, and AI-assisted operations, at lower overhead: it collects high-resolution metrics without the cluster load a JMX-based stack or OpsCenter impose, which compete with Cassandra for CPU, memory, and disk I/O and can fail to keep up at scale. See the DSE component mapping.
AxonOps also:
- Installation automation. AxonOps does not provision infrastructure, but supports installing Cassandra through established configuration-management tooling: Ansible, Chef, or Kubernetes (K8ssandra). See Installing Apache Cassandra.
- SaaS or self-hosted. Self-hosting keeps all telemetry and cluster data inside the operator's network, suited to regulated, secure, and air-gapped environments.
- 24/7 SLA-backed support. Vendor-independent production support, around the clock.
- One control plane for Cassandra and Kafka. The same tooling and support extend to Apache Kafka®, so teams running both manage them in one place.
- No lock-in. AxonOps runs alongside plain open-source Cassandra and can be removed at any time without moving data.
AxonOps runs production Apache Cassandra for organisations in regulated industries, including financial services, payments, and the public sector, among teams that have moved off DataStax Enterprise. Deployments range from small clusters to estates of hundreds of nodes across multiple datacentres and regions, including fully air-gapped environments. See the customer stories.
To discuss a migration, see AxonOps for Cassandra or contact AxonOps.
The business case: cost, risk, and support
Section titled “The business case: cost, risk, and support”For engineering leadership weighing the disruption of a migration against the cost of staying put, the table below sets out the trade-off.
| Category | Proprietary or managed platform | Open-source Cassandra + AxonOps |
|---|---|---|
| Database cost | Per-node licence (DSE) or consumption-based billing (Astra), set by the vendor | Apache Cassandra is open-source and free; infrastructure costs are bounded to what is run |
| Database version | DSE 6.x: built on a fork of Apache Cassandra 3.11, a release line that reached end of life in 2024. Astra: managed, vendor-controlled | Current Apache Cassandra 5.0, with 6.0 on the horizon |
| Operations tooling | OpsCenter (DSE-only) or the provider's console | AxonOps, SaaS or self-hosted, on any open-source Cassandra |
| Lock-in | Proprietary SSTable formats or managed endpoints | Open formats; run anywhere; remove any tool without migrating |
| Deployment choice | Coupled to the vendor's infrastructure, cloud, or managed endpoint | On-premises, any public cloud, hybrid topology spanning both, or Kubernetes; AxonOps operates the cluster consistently across all environments |
| Migration effort | Not applicable | The data migration window for a mid-sized cluster is commonly 1 to 5 days, driven by data volume and network throughput. Standard CQL workloads require no application downtime because the AxonOps ZDM Proxy, a dual-write proxy that mirrors application writes to both the source and target clusters, keeps the application serving traffic. Replacing proprietary DSE components (Search, Analytics, Graph, and similar) is separate, scoped work. |
| Vendor risk | Roadmap, pricing, and support set by a single vendor | Community roadmap; operator's choice of tooling and support vendors |
The migration is a one-time, bounded cost; licence and consumption fees recur for as long as the proprietary platform is retained.
Infrastructure choice is a second dimension. A proprietary or managed platform couples the database to its vendor's infrastructure, cloud, or managed endpoint; moving or spreading workloads across environments then requires re-platforming, not only migrating data. Apache Cassandra runs on-premises, in any public cloud, in a hybrid topology spanning both, and on Kubernetes. AxonOps operates the cluster in the same way across all of those environments, so workloads can be placed, moved, or spread without changing the database platform. The organisation retains full control over where its data and database run, tied to no single vendor's infrastructure or hosting model.
Commercial support and on-call coverage
Section titled “Commercial support and on-call coverage”Apache Cassandra is masterless and highly resilient: data is replicated across nodes with no single point of failure, so losing a node does not take the cluster down. The cluster keeps serving reads and writes while the failed node is replaced as a routine operation. When the cluster is correctly deployed, monitored, and managed, Apache Cassandra supports continuous, zero-downtime operation.
AxonOps makes that outcome more likely and easier to sustain. The tooling keeps the cluster correctly configured, repaired, backed up, and monitored, and AI-assisted operations surface and help resolve emerging issues early, often before they need a support ticket. Where deeper help is needed, 24/7 expert support backed by SLAs is available without the lock-in.
Available migration guides
Section titled “Available migration guides”| Source | Target | Guide |
|---|---|---|
| DataStax Enterprise (DSE) | open-source Apache Cassandra | Migrate from DataStax Enterprise to Apache Cassandra |
| DataStax Astra DB | open-source Apache Cassandra | Migrate from DataStax Astra DB to Apache Cassandra |
Additional migration sources will be added here as sibling guides over time.
Frequently asked questions
Section titled “Frequently asked questions”Is open-source Apache Cassandra cheaper than DSE or Astra?
Section titled “Is open-source Apache Cassandra cheaper than DSE or Astra?”Usually, yes. Apache Cassandra has no licence fees and no consumption-based billing; infrastructure costs are bounded to what is run. The migration is a one-time cost, whereas licence and consumption fees recur indefinitely.
Is DataStax Enterprise (DSE) end of life?
Section titled “Is DataStax Enterprise (DSE) end of life?”DSE 6.x is built on a fork of Apache Cassandra 3.11, a release line that reached end of life in 2024; that open-source foundation is no longer maintained by the upstream community. Migrating to an actively maintained Apache Cassandra release addresses that exposure. See the DataStax Enterprise migration guide.
Is self-hosted Cassandra production-ready?
Section titled “Is self-hosted Cassandra production-ready?”Yes. Apache Cassandra is masterless and highly resilient: data is replicated across nodes with no single point of failure, and the cluster supports continuous, zero-downtime operation when correctly deployed, monitored, and managed.
Are features lost when leaving DSE or Astra?
Section titled “Are features lost when leaving DSE or Astra?”CQL workloads carry over directly, and Cassandra 5.0 adds Storage-Attached Indexes, vector search, and the Unified Compaction Strategy. DSE's proprietary engines and Astra's Data API are the parts that need replacing or fall outside these guides.
How long does a migration take?
Section titled “How long does a migration take?”For a mid-sized cluster with a standard CQL workload, the data migration window is commonly 1 to 5 days, driven by data volume and network throughput. No application downtime is required because the AxonOps ZDM Proxy keeps traffic serving throughout. Replacing proprietary DSE components such as Search, Analytics, or Graph is separate, scoped work.
Does using AxonOps create vendor lock-in?
Section titled “Does using AxonOps create vendor lock-in?”No. AxonOps runs alongside standard open-source Cassandra without changing the database, and can be removed at any time without migrating clusters or moving data.