3 of the 4 cables that provide internet connections between Europe and North Africa (and thence the Middle East) are broken at the moment. Repairs won’t be completed for a few more days, because a ship has to find the broken ends, and drag them to the surface, where they will be stitched together fibre by fibre. The impact on business has been severe in some parts, with call centres unable to function, phone lines degraded, and slow or unavailable internet connections.
It was surprising to find that most of the connectivity between Europe and North Africa was provided by 4 cables that run along the same route across the floor of the Mediterranean, where they can all be taken out by 1 incident. The internet was conceived as a way of delivering electronic communications that could survive whole cities being obliterated by war, with the data always finding a way around the break to get to its destination. While that’s the theory, in practice there are lots of places where we find single points of failure. There are facilities in London’s Docklands that, if they broke for some reason, would take out a significant percentage of the UK’s internet connections. A fire in a tunnel in the North West of England 4 years ago impacted a very large number of people and businesses for several days, melting the only route many of them had to their favourite websites. The theft of some communications equipment from a Cable & Wireless facility took out Sainsburys online presence and many others. Single points of failure seem commonplace, in fact.
As the internet takes on a more crucial role for all of us every day, this seems like an increasingly serious risk. How is any small or medium size business supposed to keep its online operations going, if Sainsburys can’t? Even if it’s technically possible, costs would surely be prohibitive?
One interesting answer to this question lies in the increasing availability of fully managed hosting and connectivity in what’s referred to as Cloud Computing. Companies paying hundreds or thousands per month for dedicated servers in data centres may well find that it’s just as cheap (or cheaper) to transition to a service like Amazon’s Elastic Computing Cloud. This offers virtual servers on Amazon’s infrastructure (you don’t get a physical box: but a share of some of their kit) at interesting prices, and it’s possible to specify the set-up so that different servers are on different continents. If a box fails, it has no impact, and if you need another one, it’ll be there in minutes.
This is going to be a fast-moving area of development over the next few years. Amazon’s offering is attractive, but others are on their way, and it could mean more resilience, more performance, easier management for the same money. Soon many companies that have their own servers, racks and suites in data centres might find they don’t need them any more.
It might even be good for the environment, as the data centres these cloud businesses are building are among the most efficient that there are.
Post a Comment