Finding a HA Database Solution
Wednesday, January 25, 2012 20:37Finding a highly available datastore has proved a tough topic in todays highly available zero downtime technological universe. With MySQL being the gold standard for startups and sites that don’t need to scale quickly on demand, it becomes a bigger problem when you finally hit the limits of MySQL as it does not scale well due to its active-passive nature. Although nearly impossible to maintain, to get an active-active MySQL cluster it would require tedious maintenance of auto incrementing keys and endless replication issues that are not suited for the faint of heart. Finding an open source solution that would guarantee eventual consistency and was active-active was not the hard part, deciding on a solution proved to be the challenge as each solution has its pros and cons as HA has still not been made turn-key and adopted the keep it stupid simple method. The Apache Project has come out with quite a few new tools to address this deficit, The Hadoop project has changed the world of offline processing with it’s Map Reduce functionality as the new NoSQL key/value datastore Cassandra has come close to solving the HA challenge with its eventual consistency but thats not good enough. Oracle offers the only enterprise solution designed to scale but at a price no one can or wants to afford. We can cherry pick pieces of open-source to solve our issues if done methodically and though careful consideration of the pros and cons of each piece. They may require more work to get off the ground and maintenance but the cost savings alone can put a workforce of 200 people together in no time when we save an IT organization from its highest expense, licensing and software. If North America was split in half tomorrow and one side fell into the ocean would you still be serving people online is the million dollar question in todays complex every penny / second counts IT environments.