To create a cost-effective infrastructure option for housing big data deployments, many IT organizations have been forced to use a combination of cloud, dedicated hosting and colocation from multiple providers. But the emergence of new colocation services that provide cloud-like flexibility along with secure physical infrastructure helps meet the demands of big data through data center hybridization.
Today, the largest hard drive you can buy provides 3 terabytes of storage. Online gamers playing Battlefield 3 will generate enough data to fill that entire drive in just three days. Twitter generates enough information to fill one up in a mere six hours. Facebook is even more impressive: they collect 500 terabytes of data per day, which means it would take them a scant 8 minutes to fill up the world’s largest hard drive.
So where do companies actually put all of that data? If you’re Facebook, Twitter, Google or one of a handful of other industry titans, you build your own data centers and fill them with servers as fast as you can build them. But if you’re not yet a multi-billion dollar corporation, building your own state-of-the-art data center is probably not a viable option. Cloud and dedicated hosting can provide flexible contracts and tools that make hosting and scaling traditional web applications easier, but these options make less sense when you start talking about storing petabytes of big data.
The evolution of colocation
Many companies find that the most cost-effective way of deploying a big data cluster is to put it in someone else’s state-of-the-art data center via colocation. But cost-effectiveness is only part of the equation. Traditionally, colocation hasn’t been able to offer the customizable options of dedicated hosting, or the agility and flexibility of cloud. Whether it’s trying to meet additional seasonal demand, running an intensive one-off report on your big data deployment, or deploying a new web application, companies can’t wait four weeks to deploy it in a collocated environment. They need access to on-demand resources.
The ability to connect your colocation and cloud resources can fill this void, and help bridge the gap between big data deployment and the cloud. Internap’s Platform Connect offers true hybridization and allows you to seamlessly link your colocation, Custom and Agile hosting, and AgileCLOUD environments across the same layer 2 or layer 3 network within any of our Agile-enabled facilities. Launching your new free-to-play game next month and need to make sure your MongoDB deployment can temporarily handle a 500% increase in the amount of data you’re collecting? Need to run a series of reports on your CouchDB deployment that’s already seeing 85% utilization? Want to replicate your 80TB Hadoop deployment for a few weeks so you can see firsthand what effects an hdfs changes will have on your queries? No problem.
Multiple services, one provider
Instead of searching for multiple providers – one partner for colocation and IP, one for dedicated hosting and one for cloud – Internap offers multiple services under one roof, including award-winning colocation for housing your big data, the industry’s fastest and most reliable IP , exceptional Custom and Agile hosting services for your websites and mission critical applications and the fastest and most cost-effective cloud for your on demand needs. Being able to reach out to a single provider who understands your IT needs at every layer of the OSI model is a small but important step.
Until now, traditional colocation has taken a backseat to dedicated hosting and cloud computing, but it’s emerging as an agile, cost-effective option for housing big data. The ability to hybridize your data center with Platform Connect and apply cloud-like flexibility to physical servers offers IT organizations a best of both worlds approach to big data deployments. (See these capabilities in action at Internap’s New York Metro data center.)
Learn more about how Platform Connect can hybridize your data center and provide a new approach to big data infrastructure challenges.