At long last, I get to smash some hardware! Ever since we started the recovery-oriented computing (ROC) project at Stanford and Berkeley in 2001, I’ve been dreaming of demo-ing ROC by taking a sledgehammer to a running computer and have the software continue running despite the damage. I never quite had the funds for it!
The Aster nCluster analytic database embodies so much of ROC (microrebooting, undo, fault-injection-based testing, and so on), that fulfilling this “childhood dream” within the context of nCluster is a perfect match. It’s hilarious to watch, but for me it was a great experience; check out Recovery-Oriented Computing for Databases (the actual demonstration) and DBA’s Gone Wild (just for fun). The only thing I’d do differently is get a bigger sledgehammer, because some of that hardware was really built to last (hats off to HP )!
Beyond all the fun involved, there is also a broader message. A lot of data warehouses are way too fragile, and too many people believe that investing in more solid hardware is the way to go. Frontline business applications have to be able to withstand a wide range of failures, and what I do in these videos is really just scratching the surface.
At Aster, “always on” availability is much more than a key marketing message – it’s a core database innovation founded in recovery-oriented computing, which minimizes both planned and unplanned downtime for our customers. Whether it’s an analytic application, or an analyst requiring 24/7 queries for their modeling, data mining, or business intelligence (BI) report – having a database they can depend on is critical.