A distributed shared-data database

Get the code on Github

A feature rich platform to run all your data workloads.


Tell has proven to be fast. It can execute online transactions within milliseconds while running analytical workloads in parallel. All data is read from and written to main memory. The whole implementation is lock-free and inter-node communication is kept at a minimum. Tell can outperform state-of the art competitors. It is also optimized to run on fast low-latency networks (Infiniband).


Tell is not just another database. You can use its modules in isolation and just use the parts you need. Its modular design makes developement easy - for you and for us. Tell consists of a fast storage with support for shared scans and simple relations, a supporting library called Crossbow, a library that supports full transactions on top of your data and much more. And even more will come!


Tell decouples storage and processing. Since processing nodes are almost stateless, scaling them is straight-forward. Our storage is optimized for the worst-case: we do not partition your data according to some workload requirements. This has the big benefit, that distributing the data gets a trivial task. Your system will behave well no matter how your workload looks like.

Mixed workloads

While most databases are optimized to run either online transactions or analytical workloads, Tell is designed from ground up with mixed workloads in mind. You can run your analytical queries while your customers are searching and updating data. We use shared scans to make sure that resources are divided accordingly and that your analytical queries finish in a predictable time.