Step 1. Identify the most critical business applications and largest-consuming applications in terms of IT resources in the datacenter. These become the primary targets for a demand-driven optimization approach to transforming the datacenter. The rule of thumb is that 30 percent of applications typically consume or create need for 70 percent or greater of the datacenter infrastructure. Apply a decomposition process to these applications one at a time or in group of similar types. Measure and map the workload across the IT supply chain in terms of performance, consumption, and attributes of how the work execution is managed and what infrastructure it actually uses. This linkage approach of demand and supply creates a data-driven, objective view to affect change. By changing how you manage the application (dynamically at runtime, virtualized across all layers -- work, information, and infrastructure) and changing what it runs on (standardized and purpose-driven infrastructure that leverages network and computing physics), firms can consistently find they can do twice the work on half the infrastructure. That math alone will drive radical impacts quickly.
Step 3. Standardize your management strategy of the IT supply chain across the datacenter from a top-down perspective. Incorporate building blocks of runtime management and service orchestration with holistic virtualization (it is very important to note the "with" and "holistic"). Firms that implement just infrastructure virtualization without workload and information virtualization will negate the optimization they are trying to achieve, as they will create new bottlenecks for the delivery of IT from the datacenter. Firms that do not implement dynamic runtime management of workloads will not exploit the elasticity of their virtualization efforts. You have to bring demand to supply as it happens, not force demand into a predefined supply model.
The second building block is a purpose-built combination of datacenter footprints that provide optimized physics -- from energy draw to heat dissipation to quantity and types of cables required to connect, communicate, and run workloads. A unified footprint of network, compute, storage, appliances, and software runtime that matches the types of workloads the enterprise supports creates a simpler, leaner platform engineering model.