So ok, I really like looking at new interesting NoSQL databases. Up until yesterday I felt Riak was the most interesting one (have also coauthored Phriak - a Riak binding for Pharo Smalltalk) but after an evening of discussing choices of key/value backends for the Oak project I decided to “google a bit” and stumbled upon HyperDex.
Without having had the pleasure to actually play with it yet - on paper - it sure looks like a killer NoSQL database. It seems to offer basically the same kind of robustness and horisontal “just add another box” kind of scalability that Riak offers including auto balancing etc. But at the same time it offers awesome raw speed beating both MongoDB (without cheating like Mongo does) and even Redis, although HyperDex is truly persistent. And to put sugar on top - it has atomic operations in Redis style and manages to somehow avoid the eventual consistency corner - not sure about the details around that.
In summary - an extremely interesting NoSQL database.
Building from source
NOTE: You don’t need to build anymore, as of release 1.0.rc5 there are binaries available for Ubuntu
The instructions on how to build HyperDex from source were a tad incomplete. So I offer the following procedure that seemed to work for me in a relatively clean Ubuntu 12.10.
First we need a bunch of tools (updated 2013-01-17 with some corrections from Robert Escriva):
1
|
|
Then we need a bunch of git clones:
1 2 3 4 5 |
|
Then, in the order listed above, do the following in the first four:
1
|
|
…and perhaps something like this in HyperDex:
1
|
|
If all is well you should be able to find the hyperdex
command in your path and if hyperdex daemon --help
provides helpful output then all is probably fine. For more details take a look in HyperDex/doc/02.installation.rst.
Happy hacking!