CES last week was quite a gadget fest in Las Vegas; quite a few new portable devices got introduced, and more are to come. The only thing they have in common? A thirst for optimized video so that end users can receive the best looking video as quickly as possible…which is why I am very honored to welcome Encoding.com to the Rackspace Cloud ever-growing family.
These guys have built a cost effective, flexible, fast and user friendly transcoding platform that easily allows any company or person to integrate video transcoding within their workflow. To do so, Encoding.com decided from the get-go to build a scalable product and they therefore decided to free themselves from hardware costs by using cloud computing to meet their ever-changing needs, as discussed in this video:
Encoding/transcoding video is a very CPU intensive task and they found a perfect match for their needs when they started looking at our CPU bursting capabilities that are an inherent property of our Cloud Servers offering.
What do I mean by CPU bursting? Basically, we allocate CPU using a very simple fair-weighted algorithm to guarantee that any Cloud Server always receive a minimum CPU amount (based on the server size), furthermore, this algorithm allows any server to use, in a fair fashion, any unused CPU that is available on the physical gear, in essence, allowing any server CPU to burst on a as needed basis.
As demonstrated in the recent study produced by Bitsource, this makes Cloud Servers a very cost effective platform for compute intensive tasks, such as the transcoding that Encoding.com performs. There are a few interesting facts in the Bitsource study. For example, if time to compute is not the top priority, the smallest Cloud Server ends up being the cheapest option for compute intensive tasks. And then, if both time and cost are priorities, small Cloud servers are definitely the way to go as shown in the attached chart. This makes sense as the smaller a Cloud Server is, the greater is its opportunity to burst and grab extra CPU resources. And even then, the time to compute is still within the same order of magnitude as larger servers or competitors’ servers that have larger, but limited capacities.