In a recent LinkedIn posting on the IaaS Infrastructure as a Service forum, a conversation developed around the issue of cloud providers using the ability to over commit resources to drive down their prices. Essentially, this means that while a provider may have the resources (CPU, RAM, storage, etc.) to support a specific volume of a computing resource, they know that the vast majority of cloud servers won’t be utilized their resources at full capacity. This leaves them with spare resources to sell additional cloud servers with without having to increase the actual amount of physical resources.
While this business model (very prevalent from traditional web hosting) is a valid concern on many fronts, it is only one part of the problem. Many cloud providers, especially those routed in the hosting industry, also bundle their resources together to build fixed server instances, thereby creating an inherent issue of over purchasing i.e. contention of resources. The nature of computing is such that the customer needs enough CPU, RAM, storage, etc. So, the customer must pick a bundle that covers the minimum necessary of each resource. With bundling of resources and fixed server sizes, the chances are that that means getting way too much of one or more resources in the process. We call that baked-in over purchasing.
The problem gets worse when you consider scaling. If you suddenly find that you need more RAM, you now need to upgrade your entire server instance package, regardless if you need more CPU or storage – you’ll have to buy more of those too. Now, you are over provisioned on all your resources and particularly on storage and CPU since all you really needed in the first place was more RAM! If you are unlucky enough to be running software based on resource size you will get hit with additional license fees on that too.
For example, you are running an Oracle 11G database that becomes RAM limited (which they often do!). You now need to increase your RAM by a few GBs. Instead you now have to upgrade your server to a new size which is 50% bigger and has additional storage and CPU cores which you are paying for. Bad news, the Oracle 11G license is based on cores in use, your license fees just went up for CPU you don’t even need.
In short, through the combined practices of resource over commitment and bundling, server size mismatch and poor scaling options, customers end up wasting a significant portion of their budget on resources they don’t need and don’t use and often are not actually receiving the resources that they do purchase.
Recognizing these customer issues and poor cloud resource management practices, we at CloudSigma have architected our IaaS offering in a completely unique way to avoid such costly over-provisioning. With completely unbundled resources, priced as a utility service, we offer customers the flexibility to choose exactly the right amount of RAM, CPU and storage for their business needs. That, combined with offering the shortest billing cycles on the market of just 5-minute, means customers are put back in the driver’s seat, regaining control over their cloud server provisioning, scaling and resource utilization. CloudSigma delivers purchasing efficiency.
What do you think about other cloud providers’ over commit and resource bundling practices? Will demand for such a flexible, transparent model as CloudSigma’s ultimately drive other providers to change their business models?