You are in Los Angeles reading a website for a company in Australia. Should you care about the weather in Virginia? It turns out you probably should. Consolidation of online services means it’s hard to tell where the information you are accessing is stored and how global events affect your internet activities.
The Structure Of The Internet
As many of you probably know, the internet was designed to be a redundant communication network. When you send or receive information, the data can travel along many paths just as you could take different routes to the grocery store. If there’s construction on Main Street, you can take Oak Avenue instead.
However, like road networks, there are preferred routes through the internet. These major nodes are key data centers in the global network, but if they go down for some reason your data still gets through by taking another route. You’d prefer taking the freeway but you can take surface streets if you have to. The problem facing internet users today isn’t the route; it’s the destination. Cloud computing is creating a growing consolidation of online services into a relatively small number of data centers and that could create problems.
Have You Been To Amazon Today? Are You Sure?
A study by DeepField Networks estimated that one third of daily internet users visit a site on Amazon’s network at least once. Flip that around and that means you probably visit one of Amazon’s sites every three days. Before you say, “But I never go to Amazon”, we aren’t talking about the online retailer. We are talking about their IT infrastructure service.
Amazon realized a few years back that they were using a small fraction of their IT infrastructure so decided to start leasing out their data center capacity to other companies through Amazon Web Services. Nobody knows how many organizations have their resources stored on Amazon’s network–well, nobody but Amazon and they won’t say–but it is known that some of their clients include Netflix, Reddit and Instagram. When Amazon has an outage, such as the one they had in July of this year, they take a big chunk of the internet with them.
A Hurricane In Virginia Could Devastate The Internet
Amazon Web Services’ largest data center is located in Northern Virginia. Rackspace, another major provider of online infrastructure, has two data centers in the area as well. The region has seen a dozen major hurricanes in the last century and a half, and meteorological studies have shown hurricane frequency and intensity are steadily rising worldwide. A major storm could cut power or even physically destroy the data centers.
A similar case was seen recently in New York. Hurricane Sandy left large areas of the city without power. Data centers typically have diesel generator backups, but the massive flooding destroyed the generators or their fuel pumps and left the facilities dark. Not to mention employees couldn’t get to the sites anyhow. Online providers scrambled to restore service to popular sites such as the Huffington Post, Gawker and Gizmodo.
How Does This Affect Data Center Design?
Redundancy is a common theme in data center management: redundant storage, redundant servers, redundant power. In the most extreme examples, a company could maintain two completely identical data centers in widely separated geographical areas. If the main site gets taken out by disaster, then the other can come online with barely a hiccup. These backup systems allow the facility to keep operating even in the face of problems, but managers have to balance the cost of maintaining duplicate equipment with the cost of downtime.
On the other hand, cloud computing means fewer companies maintain their own data centers anymore. Not only that, but employees don’t even have files on their individual computers since it’s all stored online. As IT information becomes more centralized it is going to become increasingly important to plan for disaster. Data center managers will have to answer to anywhere from dozens to thousands of companies who demand 100% uptime no matter what.
Consolidated data centers are more accessible, more efficient and more reliable but there are still concerns. The “single point of failure” can’t be allowed to exist. That’s not the way to run an internet. As online services continue to merge, data centers will have to become increasingly robust as they shoulder larger online burdens.