Large web applications need a database that can scale across continents, that offers operational simplicity...
There is a tide in the affairs of databases,
Which taken at the flood leads on to fortune.
On such a full sea we are now afloat.
For almost three years we have been looking for a company that has the combination of technology depth and management leadership to tackle Oracle's database incumbency. Today, we are excited to announce our investment in DataStax, the enterprise distribution of Apache Cassandra.
There is a fundamental performance throttle in every large web application: the relational database. Whether back-ended by Postgres, MySQL or OracleDB, most web applications are based on a three-tier architecture that scales suitably in both the web server and application server tier, but miserably in the database tier.
Over the past few years we have consistently asked companies how they plan to scale their database layer once they outgrow MySQL. The answer has invariably been a mix of fear at leaving the relational database world, confusion around the possibilities offered by NoSQL, and an appeal to 'sharded MySQL' as a last resort.
As it has become clear that structured data growth will eventually overwhelm the ability of relational databases to fit on a single server, we have closely followed the various efforts in the NoSQL world that are attempting to solve this problem. The need is easily defined: large web applications need a clustered database solution that can scale across data centers and continents, that offers long-term operational simplicity, exceptional fault tolerance, and low latency in Silicon Valley, Sydney and Surakarta.
As the world flattens, the data behind modern web applications has to be physically closer to users. The speed of light is fast but data can take 1/4 second to get halfway around the world and in 2013 that's too slow. Latency added by a distant database can seriously impair user experience. Akamai's content delivery network has solved this problem for static content but web architects need a different solution for the dynamic data usually trapped in an Oracle relational database. DataStax's database allows a single, unified database to span the globe so that users can access the data closest to them, while still maintaining eventual consistency in the data.
The second tremendous benefit that DataStax offers is the ability to easily scale the database tier without resorting to a sharded relational database. Many of our portfolio companies have been forced to shard their MySQL database and the pain has been obvious. Splitting database tables across multiple servers is not for the faint of heart and removes many of the benefits inherent to relational databases. Scaling Cassandra is a simple matter of adding new nodes to a database ring. Performance scales linearly and predictably with no change to the data model.
Our conclusion is that only Cassandra offers a complete solution and that DataStax's Enterprise distribution adds the necessary security and management features that large web-facing enterprises need. 25% of the Fortune 100 are already using DataStax and more are deploying it in production environments every quarter.
As the tide begins to shift, we believe that DataStax has its sail set to capitalize on the trends we have outlined above. Most important, though, we believe that they have a team with both the operational and technical depth to win the race.