When I.T. professionals start talking about resilience seriously inevitably they mention redundancy. Now, in this day-and-age the ‘R’ word would initiate furtive whisperings around the coffee machine and even more furtive glances towards management but for a techie it has a different meaning from being RIF’d, as our American cousins would say, or being let go (which always makes it sound as if you wanted to leave in the first place!).

In this context redundancy simply means having more components in the solution than you require so that if one fails another is already there to take its place. In techie speak we say n+1, n being the number you have to have plus an extra for luck or, perhaps, more appropriately for bad luck!

Critical systems, such as servers, are built all the time using this methodology. Disks are configured in RAID configurations, you have more than one power supply and memory has an extra parity bit. It all helps build in resilience.

If you’re really serious you have two servers or more in a cluster, should one machine fail then the clustered functions will continue on the remaining hardware. This is great until the room in which they are in burns down so to get around that you spread all of the hardware that make up the cluster between different rooms often in different buildings and even in different cities.

This all may sound rather a lot of trouble to go too but system downtime can cost a business tens or hundreds of thousands and generally be quite stressful for the resident I.T. fraternity so these measures are considered appropriate. Besides, they are also much bigger toys!

Google take it one step further. They have 36 data centres around the world. These are huge buildings housing many thousands of machines so to lose the services and data in one data centre would be a major problem. Losing a data centre would also be an unusual event as not only the computers built for resilience but also the building itself with more than one source of power, on-site generating capacity, all the building systems such as air conditioning would be fully redundant. This is one reason they cost millions to build.

However, losing a data centre is also not unknown.

Google appear to have work around this by developing the technology to transfer all data and services between data centres quickly so they can essentially lose one data centre and still function on the remaining 35. N+1 with attitude!