Is the Cloud New or a New Buzzword?

There are some who say that cloud computing is just old technology—mainframe time-sharing, outsourcing, virtualization—wrapped up in a new buzzword dreamed up by marketers.

I think it might be more appropriate to say that it is a classic business model, leveraging increasingly mature technologies, applied only recently in its entirety in a convenient to use and mature way to the domain of computing.

We’ve discussed what cloud computing is; let’s compare it to prior approaches with which it exhibits some similarities.

• Cloud is not classic time-sharing—but it certainly has similar characteristics. The cloud is based on a different scale-out architecture, whereas the time-sharing model was intended for access to a single mainframe computer. The cloud also introduces the additional dimension of geographic dispersion to support location-independent processing of highly interactive workloads.
• Cloud is not managed services. Managed services come in a variety of flavors, such as basic managed, performance managed, and remote management. Basic managed services are typically resources such as managed servers or managed storage. In this model, one or more physical servers, say, are provided to a customer with either basic monitoring and management (e.g., ensuring that the server is alive and responding to pings, i.e., requests to verify its health) or enhanced monitoring and management (e.g., ensuring that the server is processing at least 5,000 transactions per minute or meeting a response time objective). Cloud environments are managed for availability as well, but a key difference is that managed services traditionally offer a fixed quantity of resources under multiyear contracts as opposed to on-demand, scalable resources for extremely short durations.
• Cloud is not outsourcing. It’s true that outsourcing has similarities to cloud, in the sense that a customer, rather than owning assets, maintaining competencies, and running an organization, shifts an entire capability to a third party. However, in the outsourcing model, assets are transferred in their entirety as a fixed quantity. For example, 5,000 servers, two data centers, and 400 IT professionals may be moved over the outsourcer. However, outsourcing has not typically made use of shared, scalable capacity to drive benefits but rather focuses on leveraging centers of excellence to achieve economies of skill, cost management, and managing scope creep to drive costs out of the business. In short, although outsourcing appears to gain and lose favor over time but generally can drive benefits, these benefits are not the same as what the cloud offers.
• Cloud is not virtualization. It’s true that virtualization can be important in building clouds, enhancing security, and migrating workloads between physical servers. However, the concept of virtualization is orthogonal to “cloud.” A cloud service could be offered without virtualization, and virtualization can be used in many ways that have nothing to do with the cloud. From an economic perspective, virtualization generates enormous value for small workloads but has limited value for large ones. It is cheaper to send a postcard via a “virtual” or multitenant berth on a postal truck rather than renting out the entire truck to carry it, but when moving an entire household, there is not much difference.
• Cloud is not grid. The term “grid” has been used to describe desktop grids, such as SETI@home; “departmental grids,” which arguably are just clusters; and global grids, such as PlanetLab and the National Terascale Facility. The primary usage of the term is to describe a heterogeneous set of entities and resources that operate in a loosely coupled fashion. Moreover, traditionally the model is a peer-to-peer resource-sharing system between entities, without cross-charging. It is true that economic concepts related to large-scale distributed systems were being explored over two decades ago5 and that the grid computing community is increasingly moving its center of gravity toward concepts more frequently associated with cloud computing. Cloud computing is also assimilating concepts typically associated with grid: The Intercloud is an emerging heterogeneous federation of cloud providers, in the same way that the Internet is a heterogeneous agglomeration of data networks. However, a salient difference is the usage-based pricing of the cloud, as opposed to merely distributed resource allocation of grids, as is the notion that a cloud itself is a service offered by a single entity rather than a loose collection of entities.
• Cloud is not utility computing. Although usage-sensitive pricing grew out of utility computing, the utility computing concept did not incorporate the notion of location independence: A single server in a single location could be utility-priced.
• Cloud is not distributed computing. Although most clouds are distributed, the two concepts are not equivalent. Distributed computing can support location independence but doesn’t necessarily incorporate common resources across multiple customers, utility pricing, or on-demand resources.
• Cloud is not ASP. Application service providers are perhaps one of the closest models to one type of cloud: software as a service. However, as application service providers, they don’t expose infrastructure resources via an infrastructure as a service model.
• Cloud is not the Internet. The Internet is a useful network for enabling ubiquitous access to cloud services. However, a variety of other network technologies are better for certain cloud architectures, such as hybrid clouds, and for certain types of traffic, such as high-definition video.6
• Cloud is not the Web. Web 1.0 is usually viewed as the one-way display to Web surfers of static Web pages created by Web site designers, whereas Web 2.0, and its business-internal counterpart, Enterprise 2.0, is usually considered to encompass two-way interaction, for example, between blog authors and commenters. Web 3.0 is the mobile Web. Although the Web is one example of a cloud service, the cloud concept is much broader.
• Cloud doesn’t require self-service. Some believe that self-service is somehow important to the cloud. Although the ability to immediately provision resources is surely preferable to a multi-month delay, there doesn’t appear to be that much of a difference between using the Web to provision resources versus, say, calling up a customer service representative to do so. The fundamental economic value inherent in accelerated provisioning of resources is not necessarily related to the mechanism by which it occurs. Moreover, self-service is inferior to an application programming interface, where an application programmatically requests resources, or to autonomic control, where service provider logic realizes that additional resources are needed and supplies them either reactively or proactively in accordance with predefined policy or service level agreements.

In short, the cloud concept is related to and draws from other models of computing. However, it can be distinguished from them using the five C.L.O.U.D. criteria.