Unlike unplanned outages of your key systems that cause staff to grumble, pace around the office or think of heading home, most of us endure planned outages of critical applications with a sense of inevitable endurance. Who has never seen the “Server down for planned maintenance 5-6PM Friday” email? Even for SAAS applications, it is not uncommon to see the cold shoulder of a “Site down for planned service, try again later” message glaring at you from your browser.
WANdisco’s CTO and VP Engineering of Big Data, Jagane Sundar, reports the weekly planned outages in the big data Hadoop infrastructure while at Yahoo! were one of the biggest pain points they faced. Big business is becoming dependent on big data.
Through the 20th century and into the 21st we’ve gritted our teeth against this inescapable cost. We’ve built massive failover servers, concocted elaborate master/slave replication schemes, and built businesses around High Availability and Disaster Recovery scenarios (HADR). We thought we were doing the best we can.
And we were, until recently.
WANdisco’s active:active replication blows the doors off the horse and carriage of master/slave. With the ability to give existing applications some cloud-like capabilities, DConE is a patented replication technology that slices the head off of a single point of failure, and with it many of the headaches traditionally associated with master/slave replication and HADR architectures.
Eliminating “planned outages” is one such desirable outcome. A node can be taken offline and upgraded or serviced. When it can communicate with the group again, it will transparently catch up with the queued transactions. The users of the replication should not even be aware changes are being made to the infrastructure.
So the next time you see a notice for a planned outage, consider asking yourself if that application could be DConE enabled. And then let us know!