* Single Instance - You can use ArangoDB as a specialized database for key/values, documents or graphs and even mix up those data models in one query. * Cluster - Synchronous replication. Automatic failover. Self-organizing cluster state management. Fast cluster startup. Easy maintenance. Multiple data models. All with an ArangoDB Cluster. * Mixed Services - Modern Architectures demand scalability, polyglot persistence and low maintenance. Modern data center operating systems put your data cluster on autopilot. * Graph - ArangoDB provides also the full feature set of a mature graph store (e.g. pattern matching, shortest path, full blown traversals) and due to a little magic it can perform graph queries very fast compared to many other leading graph solutions. * Document - You can store as many attributes in a document as you like (default size of a document is 32MB, but configurable to your needs). ArangoDB has a wide feature-set for querying and working with documents like joins, secondary indexes or acid transactions. You can horizontally scale join usage to the technically possible.

* Clusterpoint database is available on-premises or run on Cloud. Using our Cloud you can scale instantly and get computational resources at millisecond granularity on-demand. How so? Because Clusterpoint Cloud distributes your database amongst very many servers in our cluster. This gives every user access to whatever computational resources you need. * Clusterpoint uses JavaScript to retrieve documents. This really makes Clusterpoint a computing engine that can not only store, but process data and run arbitrary computation. * Clusterpoint can help computationally intense queries to complete 100x faster due to massive parallelism, without huge investment in hardware and upfront provisioning. * Clusterpoint system offers flexible pay-per-use model, where computational resources are available instantly when necessary, but users have to pay only for what has been used. In such a setting a total cost of ownership question boils down to what computation is necessary and excludes the hard to predict provisioning, benchmarking and hardware sizing questions.