MySQL on Ulitzer

What if all that tap dancing, posturing, bargaining and political in-fighting Oracle had to do to get clearance to pick up MySQL along with Sun Microsystems was an utter waste of time and money. What if MySQL was suddenly shown to be one of Mr. Darwin's fabled evolutionary dead ends.

Clustrix, a 30-man San Francisco start-up that just broke cover, thinks it was born and bred to replace MySQL with a more viable, genetically superior built-from-the-ground-up product that can do what MySQL's not very good at and that's scale into the billions of the entries.

There's no MySQL DNA in the Clustrix product, only MySQL emulation. But, in a move that could impede MySQL's longevity with the checkbook set, it interoperates with the MySQL protocol so applications don't have to be ported.

The widgetry is cast as the first clustered database system for Internet-scale applications and is said to follow the evolutionary trail of the application servers and storage systems that mutated into scaleable and clustered species.

It's a cross between the massive scalability and high performance of NoSQL's key value stores and the reassuring ACID-tested (update-confirmed) relational functionality of SQL wrapped in a three-node Intel-based server appliance called the CLX 4010 that's supposed to be awfully good at heavy read/write workloads.

Each of the three or more rack-mounted appliances needed runs a so-called Sierra Clustered Database Engine. According to Clustrix how scalable you want or need to be simply depends on how many of the things you lash together in a rack.

The Sierra Clustered Database Engine is a shared-nothing execution environment that includes a Sierra Parallel Planner and a Sierra Distributed Execution Engine.

It brings the query to the distributed data, rather than the data to the query like an RDBMS by breaking the queries apart and figuring out how to spread the bits around in parallel.

That means the Clustrix Clustered Database is supposed to execute the query with maximum parallelism and many simultaneous queries with maximum concurrency.

The result is supposed to be high scalability, high read/write workload performance, high availability, online schema changes, self-healing and self-management.

The guys behind Clustrix cut their teeth at Isilon Systems doing similar massively parallel and distributed kinds of thing to storage and have picked up $18 million from Isilon-backer Sequoia Capital as well as US Venture Partners and ATA Ventures to try their hand at scaling databases.

The Clustrix Clustered Database is targeted at transaction-oriented cloud computing service providers, enterprises and Internet companies like the social networking set frustrated with the kind of application-level partitioning or sharding they have to go through to get the scale needed to deal with the mind-boggling minutia of daily life on the Internet.

That's the kind of "Daddy, does the data scale yet" problem open source Hadoop and Cassandra as well as Google's proprietary Big Table are supposed to address.

To scale MySQL usually requires a lot of painful custom coding, a pricey, time-consuming job, and even then Internet-scale, relational database functionality in a single-instance database can be elusive. Clustrix promises no more of that kind of coding.

Its comparison of six nodes of its own stuff covering six databases to sharding MySQL works out to be an eighth of the cost over five years or $253,496 versus $3,446,250.

Clustrix' Clustered Database System is supposed to scale incrementally and seamlessly to hundreds of nodes as a single-instance database, have fully relational functionality and immediate transactional consistency.

The widgetry is supposed to deploy transparently and non-disruptively into unsharded, sharded and replicated MySQL environments. As customers need to add more CLX 4010 nodes, the distributed and parallel architecture is supposed to automatically distributes data to the new nodes and scale performance linearly even under heavy write loads. It provides high availability via automated load balancing, failover, recovery and self-healing.

The Clustrix dual quad-core appliance includes two 1Gbps Ethernet front-end ports and two 20Gbps InfiniBand back-end ports, along with 32GB of RAM and seven 160GB solid state drives. The list price of the base three-node appliance is $109,995. Underscore list.

Clustrix says it's already selling the things and got its first deal in Q1. Apparently some brand name accounts were in production with the gismos before then. The company's got evals in the field now and claims they're oversubscribed. Apparently some are paying MySQL clients.