
Big data is all the rage these days, and the hype meter is revving at an all-time high. But getting caught up in the hype can be a bit dangerous for your big data applications if you don’t consider how to get the highest disk I/O. Without the right infrastructure in place, you may be disappointed in your application performance.
Many people assume that cloud means virtualization, but we don’t think that’s necessarily true. At Internap, we believe cloud can also be bare-metal. This means you do not run a hypervisor, and it’s not virtualized – it’s just a bare-metal server with an operating system, but delivered on a cloud-like service model.
This means you can spin a server, or hundreds of them, up in a matter of minutes and throw them away once you’re done with your application or after your workload is complete. We’ve basically taken the service delivery model of the cloud and applied it to non-virtualized servers.
Why is this interesting for big data? The answer is disk I/O. When you look at solutions like Hadoop or any sort of big data application, disk I/O is the number one enemy. If you can solve that problem and give your application good disk I/O, your big data application is going to run circles around those deployed on a virtualized cloud. Bare-metal cloud, meaning no hypervisor, operating system only is a great match for big data, and Internap is one of the few providers who can offer a non-virtualized cloud environment.