On any given day, it's not uncommon for a company to generate 2.5 quintillion bytes of data, pushing the amount of data that must be processed and managed to unimaginable levels. Because of the requirements for power and low-latency connections that such data growth entails, many companies have become more inclined to outsource their Big Data needs to colocation data center facilities. In turn, this has created a huge demand for colocation space as additional processing grounds for Big Data. According to analyst firm Nemertes, colocation providers will not have the available space to capitalize on approximately $869 million of market demand by 2015. This is with good reason, though, as colocation data centers offer huge benefits for Big Data, including high-density power, opportunities to decrease latency and a community of like-minded companies with which to cross-connect.

First Things First: Powering Big Data
A lot has been made of Big Data analytics and the tools most capable of cataloging and valuing Big Data. However, before analyzing data, companies must first ensure that they are able to meet the power demands of such in-depth data analysis. With growing levels of data that must be processed, compute speeds must similarly increase to keep up. As a result, a higher level of power is required. Traditionally, power demands for data computation have been well below 1 kilo Volt Ampere (kVA) per square meter; with the rise of Big Data, however, this is now pushing power levels greater than 15 kVA per square meter. Without the ability to meet these power demands, it doesn't quite matter which analytics tool a company uses, as they won't have the resources to make it effective.

Colocation data centers are invaluable in this regard, as they typically purchase bundles of power, giving them the ability to offer power to customers at a much lower cost than customers would be charged in a private data center. In addition, data center facilities are required to provide backup power supplies, so that if they were to incur a power outage or electric shortage, there would still be enough power on reserve to negate any impact on Big Data value or performance for companies hosted in the facility. Finally, colocation data centers charge only for the power used by each customer, so companies can extract significant energy savings rather than miscalculating power usage and consistently overpaying for such an expensive resource.

Slow and Steady Doesn't Always Win the Race
Beyond the immense power requirements, Big Data requires faster compute speeds than the historic norm. This is not easily achieved, but without it, companies may face significant latency for Big Data. This likely won't make a difference in some industries, such as scientific research, where quality is prioritized over speed, but for many other industries, including online and mobile advertising, even a few seconds of latency has the potential to prohibit good service or remove value from Big Data.

For example, a mobile advertising company that targets consumers based on their location will need to process GPS data, consumer demographics and preferences, and advertising platform data - all within the time it takes for a consumer to walk by a store front. Clearly, time is of the essence when the window of opportunity for a sale is just a few seconds long.

The time between the GPS data noting that a customer is approaching a storefront to the end result of an advertisement popping up on his phone will inevitably have a lag - something I like to call the "virtual hop." Simply put, the virtual hop is the time that is required to mine and manipulate data to create an end result. This is a widely known concept when it comes to website impressions, which require a virtual hop of less than two seconds. Currently, a virtual hop for the mobile advertising scenario described above is much longer than this, though it is expected to develop into a much quicker process, similar to the evolution of website response times. Rather than wait, though, companies are looking for solutions to implement today to reduce their virtual hop time - and are finding their concerns answered in colocation data center facilities.

Connect With the Server Next Door
Colocating infrastructure in a data center has the potential to remove a layer of latency and reduce virtual hop times dramatically by providing a common ground for companies that often work together to collaborate and exchange information. The ability to directly connect to the servers of two different companies housed within the same data center, a concept known as cross-connecting, has huge potential to expedite compute times and eliminate a significant portion of latency.

Cross connecting has already been used in many colocation facilities to provide near instantaneous collaboration results between data center tenants. Common examples of this include a speedy deployment to a cloud environment or seamless content aggregation across digital media platforms.

The communities of like-minded companies that are often developed within data centers offer an ideal environment for decreased latency and improved Big Data analysis. Communities typically cater to specific verticals, including cloud, financial services and digital media. For example, going back to the mobile advertising example, a cross-connect between the mobile provider's database of customer demographics and the advertiser's data would greatly improve this process, as the two servers would be able to work congruently and seamlessly from within the same digital media facility.

Despite their enormous benefits to customers, these communities are not yet a common feature across colocation facilities. A recent study from Infineta Systems found that data center-to-data center connectivity, as opposed to cross connecting, is a "silent killer" for Big Data deployments, as many data centers do not prioritize the development of communities within their facilities. Therefore, this is a distinguishing feature that companies must seek when profiling data centers to house their Big Data.

Back to Basics
Without the high-density power required to process data, or the ability to improve Big Data analysis through connectivity and cross-connections, companies are essentially collecting data for fun.