Generally speaking, data centers are custom built, manually managed and lack uniformity in terms of physical resources, making it difficult to forecast and plan for future requirements. Consequently, these factors contribute to over-purchasing resources simply to be on “the safe side”. Although enterprise IT leaders are aware of the need to increase their data center efficiency, according to past studies, the average CPU utilization level of typical on-premises physical servers was low. A McKinsey & Company research study once estimated an average server utilization rate of 6% to 12%. Moreover, the separation between the typical data center subsystems – compute, storage and network – has left data center environments rigid and not scalable. Data centers contain multiple bottlenecks, which include leaving overall scalability dependent on the least scalable subsystem (generally storage or network).
https://www.cloudyn.com/blog/software-defined-data-center-sddc-private-cloud-economy/
