Blog #1 of a 3 part series on The Economics of the Cloud, by Robert Darrow – Vice President of Operations at cloud provider, Peak (getcloud@ powered by peak .com) March 27, 2014

Buyer beware: what they say isn’t always what you get when it comes to the cloud. Sure, it sounds good when a cloud provider boasts about their virtual machine compute capacity or the fact that their cloud node offers 10 PB of storage resource capacity. Those levels of supply are fantastic if only one or two customers are sharing it; less fantastic when 100 or 200 customers are sharing it; and downright scary when it’s shared among tens of thousands of customers. When choosing a cloud, it all boils down to supply and demand.

The only way to be sure you will achieve the performance you’ve been promised (and therefore realize the value of the cloud) is by understanding the cloud provider’s underlying infrastructure, and the associated demand placed on it. In other words, you need to look under the hood. In this three-part blog I will share my rules of supply and demand for mission-critical apps in the cloud.

Much like an airline overbooking its flights, many public cloud providers continue supplying seats to their cloud long after they should. They’re making a gamble that when demand peaks for some it will fall for others. Unfortunately, that formula doesn’t always pan out—not for airplane passengers and not for cloud users. Someone is bound to be inconvenienced. When it comes to mission-critical applications, that “inconvenience” can translate into lost revenues and damaged reputations. The perception that cloud computing isn’t safe is the result of cloud providers that oversubscribe, negatively impacting performance.


  • Public Clouds – co-mingled environment, shared compute, high risk of inconsistent performance.
  • Private Clouds – dedicated compute, assurances of consistent performance and availability. You may or may not be given tools to analyze the infrastructure. Some cloud providers just don’t want you looking that closely.
  • Enterprise-class Clouds – private clouds with virtualization tools and hardware level visibility. In order to manage to SLAs required by business-critical, high availability apps, enterprise-class cloud providers supply dedicated compute and native tools to control the cloud infrastructure.

In public clouds, you don’t get to choose with whom you’re sharing the underlying CPU. If your neighbor is computationally heavy (ie., the “noisy neighbor”), your processing time till be impacted. If the cloud is oversubscribed, you will likely experience degraded performance, and that flies in the face of the scalability and convenience that cloud computing is supposed to deliver.

With private clouds, on the other hand, you can be assured of consistent performance, because you prescribe the size and quantity of the servers you want on hand in the infrastructure. Private clouds are more expensive than public clouds for good reason: like renting a car instead of taking the bus, you’re in control. You still need to understand the underlying infrastructure, however, in order to be sure you’re getting the full value of what you’ve paid for.