After months of hard-work, we are proud to release the first official version of Sensei (v1.0.0).

On the day after its release, we already received a pull request, 150 points on Hacker News and claimed #1 spot for a while, which drove 23000 page views to our project page from 8500 unique visitors.

As part of this release, we are dedicating this blog to Sensei technology.

Architecture Diagram:

What is it?

One-liner:

SenseiDB is a distributed data system optimized for scenarios with rapid updates and complex queries, including full-text search capabilities.

Sensei is used in many places to power LinkedIn.com, powering search pages to analytics products. Most notably the LinkedIn Homepage.

What is it NOT?

To clear up some confusions from some comments on Hacker News, Sensei is not meant to replace every features of a RDBMS. As with any systems, design trade-offs were made for the primary problem the system tries to solve:

• Sensei is NOT relational - JOIN operations are not supported. We adopt the philosophy of NOSQL data systems of data to be de-normalized.
• Sensei is NOT transactional - although we provide durability and eventual consistency guarantees, we don't support a full transactional insert model, e.g. roll-back etc.
• Eventual Consistency - Sensei does NOT have the same consistency guarantees that a traditional RDBMS supports. Instead, we guarantee the data to be eventual consistent, e.g. at All nodes will guarantee to converge to a consistent state.

Trying it out

Best way to try out Sensei is start from getting started page and setup all examples. This way, you can see some of the functionalities Sensei provides.

Join our community

We invite you to join us in making Sensei better:

• Twitter: @sensei_db
• LinkedIn group
• Mailing list
• IRC: irc.webchat.org, channel #senseidb