Cassandra Trainings

Cassandra in 2026: where does distributed NoSQL stand?

In 2026, Apache Cassandra remains one of widely used NoSQL databases for very high availability and horizontal scalability use cases. Cassandra 5.x brings major improvements: native vector search for AI workloads, Storage-Attached Indexing (SAI), storage engine enhancements and tighter Kubernetes integration via the K8ssandra operator. Historical use cases (time-series, IoT, logs, recommendation, fraud detection) are now complemented by new workloads tied to RAG and generative AI pipelines.

On the competitive side, Cassandra coexists with other NoSQL engines: MongoDB for general document stores, DynamoDB on the AWS managed side, ScyllaDB as a high-performance C++ alternative, Couchbase and HBase on specific niches. Comparative knowledge of these engines has become indispensable to position Cassandra correctly and avoid unsuitable use.

Why Cassandra remains relevant for Swiss workloads

Apache Cassandra today powers massive databases at globally known platforms (commerce, telecom, media, industrial IoT) needing multi-datacenter availability, linear scalability and a data model designed for heavy writes. In French-speaking Switzerland, this workload profile appears in organisations aggregating industrial telemetry, running high-traffic media platforms, capturing continuous application logs, or industrialising product recommendations.

Cassandra however requires strong discipline: query-driven modelling (not relational normalisation), careful partition key choices, anticipation of table growth, rigorous operations (repairs, compactions, monitoring). Structured training drastically shortens the learning curve and avoids costly design errors in production.

The ITTA catalogue: NoSQL and Cassandra end to end

Start with the NoSQL panorama

Introduction to NoSQL databases: Cassandra, HBase, MongoDB, Couchbase positions the main NoSQL families (key-value, document, wide column, graph) and explains when each is relevant. It covers the fundamentals of Cassandra, HBase, MongoDB and Couchbase engines, their use cases and limits. This is the recommended entry for any profile choosing a NoSQL database or making a data architecture decision.

Go deeper into Cassandra hands-on

Cassandra – Fundamentals covers distributed architecture (ring, gossip, vnodes), the CQL language, query-driven modelling, tunable consistency, replication strategies, common operations (repairs, compactions, monitoring) and production best practices. It is calibrated for developers, data architects and DBAs starting a Cassandra project or maintaining an active cluster.

Featured Cassandra courses

Selected Apache Cassandra training courses listed in the ITTA catalogue:

• Introduction to NoSQL: Cassandra, HBase, MongoDB, Couchbase
• Cassandra – Fundamentals

Cassandra in the ITTA data ecosystem

Cassandra fits into a broader catalogue. The database design and development sub-domain covers modelling, relational and NoSQL databases. The data science sub-domain brings the analytical use cases on the data engineering side. For profiles combining distributed storage and machine learning, ITTA Artificial Intelligence regroups our AI training, including RAG pipelines that increasingly rely on vector databases, Cassandra included.

On the publisher side, you will also find Open Source regrouping our open technologies training, and Microsoft SQL Server for the relational dimension. This combination allows designing consistent hybrid data architectures rather than defaulting to a single engine.

Profiles training on Cassandra at ITTA

Our Cassandra audience is broad: backend developers integrating Cassandra in a microservices stack, data architects validating a NoSQL choice for a high-volume project, SQL Server or Oracle DBAs discovering a distributed model, IoT engineers collecting industrial telemetry at high frequency, media teams handling high-traffic content, or data AI teams exploring Cassandra 5 native vector search for their RAG pipelines.

Small session sizes let our trainer tailor examples to the contexts represented. Whether you work on a time-series workload, application logs, product recommendation or a specific AI case, your trainer adjusts exercises and demos accordingly.

The ITTA partnership with the open source data ecosystem

ITTA has been based in Geneva and Lausanne for more than twenty years. On open source data technologies, our positioning is to offer a coherent catalogue (Cassandra, PostgreSQL, MySQL, MongoDB, Hadoop ecosystem, Python data) rather than a single isolated engine. Our Cassandra trainers operate on real projects in Switzerland and internationally, which lets them bring current examples and address design and operations questions from your contexts.

Sessions are available in Geneva, Lausanne and interactive virtual classroom. For data teams seeking grouped upskilling, we organise in-house sessions calibrated on your cluster, use cases and monitoring. Corporate training generally includes a small thread project based on your target schema.

Cassandra FAQ at ITTA

When to prefer Cassandra over MongoDB or PostgreSQL?

Cassandra is relevant when you need massive writes, linear horizontal scalability and multi-datacenter availability by design. MongoDB suits a general-purpose document store with variable structure. PostgreSQL remains a widely used tool for SQL transactional workloads with a volume manageable by a relational cluster. Our trainers help you objectify this choice based on your real workload.

Do I need SQL knowledge before training on Cassandra?

SQL helps to understand CQL, but Cassandra is not modelled like a relational database. The classic trap is transposing a 3NF schema directly into Cassandra. Our Fundamentals course emphasises query-driven modelling, the opposite of the relational approach. Profiles without prior SQL can follow, provided they are comfortable with tables and keys.

Do Cassandra 5 and vector search change the game for AI pipelines?

Yes. Cassandra 5 introduces native vector indexing (SAI), allowing you to use the cluster as a vector base for RAG pipelines, alongside dedicated solutions like Pinecone, Weaviate or Milvus. Teams already running Cassandra in production can avoid introducing an extra database for their AI cases. We cover this in the Fundamentals course.

What is the main difficulty in production?

Operations. Managing repairs, compactions, token sizing, monitoring and multi-DC replication is more demanding than a classic relational database. Our courses cover these topics concretely, with examples drawn from real clusters.

Operational best practices addressed in class

Beyond concepts, our Cassandra Fundamentals course emphasises operational best practices that make the difference between a cluster that holds the load and one that saturates. Initial sizing is a topic in itself: node count, partition size, choice of cloud or bare metal instances, anticipation of expected growth over 12 and 24 months. A miscalibrated partition (too fine or too coarse partition key) creates hot spots or monster partitions that degrade performance across the whole cluster.

Monitoring is also central. Cassandra metrics (latency, read and write rate, sstable size, compaction duration, repair state) require tools such as Prometheus, Grafana or managed solutions (DataStax Astra DB, Instaclustr). Without monitoring, a cluster drifts slowly and symptoms appear too late. Our course covers these tools and provides reference dashboards. Backup and restore are discussed concretely, with the different strategies (snapshots, incremental, cloud externalisation). Finally, deployment options are numerous: self-managed on bare metal for maximum performance, Cassandra on Kubernetes via the K8ssandra operator, or managed solutions (Astra DB, Instaclustr, Aiven). Our course addresses each option honestly to choose based on context.