In moving to the cloud, Enterprise IT needs to get their arms around the cause/effect and limitations that existing datacenter infrastructure design has on their agility to adopt and exploit cloud-computing models. Previous datacenter design choices have resulted in not meeting the needs of the business.

In particular, these design choices have resulted in complexity, waste, performance barriers, and cost models that don't work for the business. Lack of understanding and transparency of what has been done in the past will continue to create misalignment with business needs if not addressed. Moving to an enterprise cloud model without understanding the datacenter infrastructure mistakes is like automating and extending an already bad process.

It is important that the reader does not take this to be an attack on any work products they may have done in the past. All of us in IT have made some if not all the mistakes below - usually due to influences or drivers out of our control.

To fix this and enable enterprise cloud models, we must not employ the following datacenter design limitations that include:

Supply-driven management: Most datacenter infrastructure teams design and manage from the bottom up. The typical approach is to standardize, partition, allocate, and implement a "vanilla" solution of compute and storage that is attached to the network based on the topology of the datacenter floor. Provisioning is then designed for peak workloads leveraging a bottom-up-designed platform that has been selected for typical reasons of vendor rationalization, price, etc. The business workload and service requirements incorporating factors of performance, price, or efficiency are not incorporated and misalignment of needs and inconsistent service delivery begin to occur.

One size fits all: Most datacenter infrastructures and typical vendor strategies are built around a perceived "standardized" footprint. The problem is that this is designed typically from the bottom up with little or no correlation to the workflows, workloads, information, content, and connectivity requirements of the business and its competitive needs. Such disconnects result in poor performance, unnecessary costs, waste, and in agility issues for both the business and IT.

Spaghetti transaction flow: Transaction flow across traditional datacenter infrastructures must deal with a design that does not consider proximity of the various devices that comprise a service unit that delivers processing to users. This results in significant performance impacts (user experience of the business suffers) whereby compute, memory, I/O fabric, disk, storage, and connectivity to external feeds are provided in terms of layout, not in terms of service delivery. Performance can be impacted by 30-fold due to this approach. Moreover, this creates waste in terms of unnecessary network traffic congestion and bandwidth usage (ROE suffers).

Definition of "insanity": The continuous use of a typical datacenter layout incorporates homogenous pooling of various classes of resources. Servers by multiple classes are typically in multiple pools; storage by file or block are in different pools in their own area of the datacenter; network load balancers, network switches, and network routers are pooled/deployed across various areas of the datacenter. This approach is not designed for business impact, business needs, optimal workload throughput, or time-to-provision or optimal space/power usage. The average provisioning cycle in datacenters with this type of layout are measured in weeks or months versus the minutes or days needed to provision, troubleshoot, or perform to meet the needs of the business.

Bottom line: Enterprise Cloud models will enable IT to take business to the next level of competitive advantage. If the supply chain of IT does not provide the delivery paradigm that enables a cloud utility model, it will fail the business in realizing its potential. The journey to the enterprise cloud begins with understanding the limitations of current design approaches.