Nine Design Principles of an Enterprise Cloud-oriented Datacenter Blueprint
The following design principles are proven to enable firms to create an enterprise cloud-oriented datacenter
Aug. 22, 2010 05:00 AM
An Enterprise Cloud-oriented datacenter is a top-down, demand driven datacenter design that maximizes efficiency and minimizes traditional IT waste of power, cooling, space, and capacity under-utilization while providing enhanced levels of service and control. The following design principles are proven to enable firms to create an enterprise cloud-oriented datacenter.
#1: Dynamic provisioning
A package-once, deploy-many application facility that incorporates application, infrastructure, and IP dependencies into a simple policy library is critical. This affords the capability to deploy in minutes versus days.
#2: Dynamic execution management
Dynamic execution management performs allocation that is managed in real time based on monitored workloads compared against service contracts adjusting allocation of applications, infrastructure services, and infrastructure resources as demand changes throughout the operational day.
In order to achieve dynamic allocation of resources, applications and other types of demand cannot be tied to specific infrastructure. The traditional configuration constraints of forced binding of infrastructure and software must be broken in order to meet unanticipated demand without a significant amount of over provisioning (usually 2‐6 times projected demand).
The design of each infrastructure component and layer needs to have sufficient abstraction so that the operational details are hidden from all other collaborating components and layers. This maximizes the opportunity to create a virtualized infrastructure, which aids in rapid deployment and simplicity of operation.
#5: Real time
Service & resource allocation, reprovisioning, demand monitoring, and resource utilization must react as demand and needs change, which translates into less manual allocation. Instead, a higher degree of automation is required, driving business rules about allocation priorities in different scenarios to the decision‐making policies of the infrastructure, which can then react in sub‐ seconds to changes in the environment.
#6: Service-oriented infrastructure
Service-oriented infrastructure is infrastructure (middleware, connectivity, hardware, network) as a defined service, subject to service level agreements (SLAs) that can be enforced through policies in real time, and that can be measured, reported on, and recalibrated as the business demands.
#7: Demand-based infrastructure footprints
Because demand can be categorized in terms of operational qualities, efficient, tailored infrastructure component ensembles can be created to meet different defined types. These ensembles are tailored to fit into operational footprints (or processing execution destinations) that are tailored based on demand.
#8: Infrastructure as a service-oriented utility
The key operating feature of a utility is that a service is provided for when needed, so that spikes are processed effectively. It reflects a demand‐based model where services are used as needed and pools of service can be reallocated or shut down as required. A real- time, measured, demand‐based utility promotes efficiency, because consumers will be charged for what they use.
#9: Simplified engineering
The utility will only be viewed as a success if it is easy to use, reduces time to deploy, provision is highly reliable, and it is responsive to demand. Applying the principles stated above will simplify engineering and meet the objectives that will make the utility a success. Simplified engineering will promote greater efficiency, enable the datacenter to migrate to a "green" status, enable greater agility in the deployment of new services, all while reducing cost, waste, and deployment time, and providing a better quality service.