Home

RubyHaze¶ ↑

RubyHaze is a little gem that wraps the Java Hazelcast library into a more comfortable Ruby package (in JRuby, of course).

Why Hazelcast?¶ ↑

Because Hazelcast is awesome. Plain and simple. Some of the features that interested me from projects like Maglev (e.g. shared objects) and Riak (e.g. fault tolerance) are possible in a fully distributed fashion:

• Queues

• Topics (PubSub)

• Maps (Hashes)

• MultiMaps

• Sets

• Lists

• Locks

• Events

And all that with an enterprise flavor and commercial support (for Java). Find out more in the links page.

Why RubyHaze?¶ ↑

For a long time I looked at JRuby as an amazing project but was put-off a bit by the perception that I would need to be a Java guru to create something worthwhile with it. That’s when I read the eye-opening article by Roja Buck and realized that JRuby was not so distant and that a nice little wrapper could be easily created.

RubyHaze at the moment is in its early infancy but it can provide you with a few niceties from the start:

• Quick access to the common distributed objects already mentioned.

• The beginnings of a simple distributed storage layer for JRuby objects (Stored) with SQL-like searching.

• A small start into skipping all the xml configuration for a builder style config.

• All packaged in a single gem (athough I might refactor Stored into another gem).

Todo¶ ↑

There is so much more we can do with Hazelcast. Here are some places we can go:

• Durability: we already have fault tolerance through multiple nodes and backups but there are mechanisms in place to create our own durability strategies. Why not use Riak/CouchDB/MySQL for simple, perdurable storage?

• Associations: adding has_many/belongs_to ala ActiveRecord or MongoMapper would be nice.

• Query Builders: we can create predicates using simple strings but something like Sequel’s Person.first(:name => “David”) could be useful.

• Better Serialization: to support predicate searching we need to create a shadow Java class to write/read our internal data in Hazelcast distributed maps. We need a conversion mapping layer for attributes other than strings, integers, boolean and doubles. DataMapper has a nice way of deling with custom mappings that we could use for inspiration.

Besides working on the RubyHaze library itself I think we can work on creating all kinds of examples from simple chat systems to recreating Resque with Hazelcast. Do you any other project ideas?

Feedback¶ ↑

Please let me know how you like this little project and feel free to ask questions. I will do my best to answer them. I promise!