JavaZone 2009 - Redis, a new kind of key-value database
Redis is a key-value database that stands out from the crowd. It is
not just another key-value database as it provides atomic and
non-blocking operations on its keys and values. In Redis the values can be
either primitives, sets or lists. Most other key-value databases
allow only simple values like strings or bytes. There are commands to
perform complex operations on these data structures, most in constant
time. It is therefore not unreasonable to think of Redis as a data
structures server.
The dataset can be replicated using a master-slave setup and also be
distributed across a set of servers using consistent hashing.
The server provides blazing performance as most operations are
non-blocking and the entire dataset is kept in-memory. Performance
tests have shown results of 10000 SETs/second, 81000 GETs/second on an
entry level Linux box.
Redis is open source, and the database server and clients are being
hosted on Google Code. The project is aiming for a final 1.0 release
mid June 2009.
In the presentation we'll take a look at the characteristics of the
database server and the operations that it provides including demos of
how it works in practice. The Java client will also be reviewed.
Geir Ove Grønmo
Works at Bouvet as a consultant. He is one of the founders and main architects behind the Ontopia Topic Maps software (now an open source project). He has been working with innovative and exciting technologies for many years and never stops looking for new ones to explore.