Cloud Computing / Server Farm

Service

Cloud computing refers to the practice of leveraging third-party computing resources, such as network grids and server farms, to extend IT capabilities and reduce the cost of ownership. This practice offers numerous potential benefits to organizations that want to centralize software and data storage management while eliminating the costly overhead of in-house hardware and software maintenance and the personnel required to build, support, and maintain enterprise computing solutions.

Cloud com puting has emerged as a new computing paradigm that gathers massive numbers of computers in centralized data centers to deliver Web-based applications, application platforms, and services via a utility model. The primary difference between the service models of cloud computing and previous software (e.g., outsourcing or data center consolidation) is scale. The premise is that as the scale of the cloud infrastructure increases, the incremental time and cost of application delivery trends toward zero.

Cloud computing allows users to dynamically and remotely control processing, memory, data storage, network bandwidth, and specialized business services from pools of resources, providing the ability to specify and deploy computing capacity on-demand. If there's a need to scale up to accommodate sudden demand, users can add the necessary resources using a Web browser. The large data center can provide similar services to multiple external customers (multi-tenancy), leveraging its shared resources to increase economies of scale and reducing service costs.

Although cloud computing is in its early stages and definitions vary greatly, the underlying technologies today are consistent. These technologies include the following:

Grid computing: A form of distributed parallel computing whereby processes are split up to leverage the available computing power of multiple CPUs acting in concert.
Utility computing: A model of purchasing computing capacity, such as CPU, storage, and bandwidth, from an IT service provider, billed based on consumption.
Virtualization technologies: Virtual servers and virtual private networks provide the ability to quickly reconfigure available resources on-demand and provide the necessary security assurance.

There are a number of service offerings and implementation models under the cloud computing umbrella, each with associated pros and cons. These models can be grouped into the following three categories: Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS), and Software-as-a-Service (SaaS). These models target varying levels of services, ranging from general infrastructure services, such as operating systems or database services provided by IaaS vendors to targeted functional services provided by SaaS vendors (e.g., customer relationship management from Salesforce.com).

The various players in the current market can be differentiated into the following two categories:

Cloud Providers: Offer one or more of the cloud models (i.e., IaaS, PaaS, or SaaS) as a service. Examples include Amazon and Google.
Cloud Enablers: Provide technology or have adapted existing technology to run on or support cloud computing. A recent example is Oracle's partnership with Amazon to add Oracle 11g database support (technology and licensing) to Amazon's existing EC2 services offering.

We recognize that the transition to a cloud computing paradigm presents a number of challenges. Issues associated with information security, reliability, and service level agreements challenge mission-critical systems. Furthermore, we've identified what we consider the key characteristics of a cloud computing environment:

Minimized capital expenditure - infrastructure is provider-owned
Device and location independence
Multi-tenancy - enables resource and cost sharing among a large pool of users
Monitored and consistent performance - can be affected by high network load
Reliability via redundant sites - allows for business continuity and disaster recovery
Scalability to ever-changing user demands - results in lower costs
Improved security from centralized data and increased security-focused resources

Our experience has emphasized the importance of "architecting for the cloud" versus simply deploying system components to the cloud to ensure that business requirements are met. Typical software and systems that are not designed to take advantage of the scalability and parallelism of the cloud will likely not achieve the full benefit that is provided by a cloud computing environment. It's also highlighted the need to transition the role of IT managers to brokers and negotiators of IT services rather than the day-to-day management of the operating platform.

Our analysis of the benefits and challenges presented by the cloud computing paradigm has resulted in the identification of the following three cloud variations:

Commercial Cloud: It consists of deployment to one or more of the commercial cloud providers (e.g., Amazon or Google). It could be a simple integration with an existing SaaS service to support a subset of application functionality or could consist of a complete migration to the cloud. This may be appropriate for non-mission critical systems (e.g., < 99.99% availability) that do not process sensitive data or where sensitive data won't traverse system boundaries to the cloud.
On-Premises (Private) Cloud: An on-premises cloud could be created to provide some of the benefits of cloud computing. Booz Allen selected a similar option in our implementation for the FBI to address the security concerns associated with a classified environment; however, the multi-tenancy aspect is then limited to a single agency. Consequently, this option doesn't provide the massive scalability that's characteristic of a true cloud.
Government Cloud: The creation of one or more government cloud computing environments. These environments would be designed specifically to address the concerns that are unique to the government. For civilian agencies, this cloud could be an extension of the current eGovernment lines of business (LoBs).

The Cloud Strategy and Planning phase (Phase 1) consists of three steps designed to ensure that all aspects of moving to a cloud environment have been appropriately evaluated and agreed upon. The three steps are:

Conduct a Strategic Diagnostic
Define a Cloud Strategy
Create an Implementation Plan

After completing Phase 1, the organization will have fully analyzed its options, identified all requirements, thoroughly assessed short-term and long-term costs and benefits, gained executive governance approval, and socialized the solution with stakeholders (including oversight entities). This phase ensures that the organization will have a high degree of confidence in successfully moving to the cloud environment, reap the expected benefits, not constrain future functionality, and avoid hidden future costs.

The Cloud Deployment phase (Phase 2) focuses on implementing the strategy developed in the planning phase. Leveraging the various cloud models helps identify the most effective solution(s) based on the existing organization architecture. Some of the criteria used in recommending a vendor are the vendor's primary service model (i.e., infrastructure, platform, or software), business model, how much existing technology can they leverage, end-user experience, and risks involved in porting to the cloud. Deploying to the cloud involves taking the decision analysis from Phase 1 as input and proceeding with the following four steps: