Types of Cloud Deployments
Cloud computing is a big force in IT today, and it isn't going away. In fact, cloud adoption is going up geometrically, both for end users (think apps on your phone or tablet) as well as for organizations of all sizes. In fact, many smaller organizations may not have any on-premises infrastructure at all, other than networking infrastructure to get connected to the cloud. With this transformation in IT, it behooves all of us in the industry to understand it and adapt or risk being out of a job, like punch card operators.
This white paper will discuss the basics of cloud computing, including a brief discussion on the location of the resources, followed by a review of the characteristics of cloud computing and the types (models) available. We will also briefly compare and contrast the various models.
This document is the first in a series of white papers that will discuss each of these cloud-computing models in further detail, with a separate document for each type. If you are already familiar with cloud computing, you may wish to skip this white paper and jump directly to the particular type(s) you are interested in.
Cloud Computing Locations
While not very relevant to the cloud-computing models available (as each model is available at any of the possible locations), the locations nevertheless will be mentioned in this and future white papers and thus will be briefly defined here.
The National Institute of Standards and Technology (NIST), an arm of the US federal government), has defined much of what cloud computing is (at least to them, but as they are a standards organization, many others have followed their definitions). We will refer to them throughout this series of white papers for consistency.
Note that NIST doesn't call them cloud locations, but rather "Deployment Models."
The public location means that the resources (servers, storage, networking, and/or applications) you will be accessing are usually located on the Internet (hence publicly available and the name of this type). This is not always true as there are some specialized networks (such as those used by the government) that may have restricted access, but for the most part, the resources you want to access are reached via the Internet.
The broader definition is that the resources are owned by a third party (the cloud provider) which is rented (either by directly paying or via ads you are shown) in some fashion from them. The resources are located at one or more datacenters of the provider.
The private location means that the resources are (usually) owned by and accessible through a private network, but in any case always for the exclusive use by a single entity or company. Typically, the idea is that an IT department at an organization owns the resources and makes them available to employees of the company in the various ways that cloud-computing offers. This doesn't have to be the case, as a third party could own them and make them accessible to just that organization.
One of the biggest advantages is that the company owns, or at least controls, all the resources and can optimize them any way they wish and deploy them much more quickly than traditional methods provided; the (potential) downside is that the company must purchase all the resources. Another advantage of this model is that the company has complete control over all the security aspects of the deployment.
Hybrid is simply some combination of the previous two locations, where some resources are located within the organization's datacenters and some are accessed publicly. This doesn't have to be the case, as it is possible to federate several private clouds or public clouds as well, but this is a far less common scenario.
Use cases for this model include the following:
- Development and testing, where resources can be quickly provisioned as needed and just as quickly deprovisioned when the project is complete. Sensitive company data may be stored in the private cloud onsite.
- Cloud bursting, where the normal load is handled by the company's own resources, but during period of peak demand (such as during the holidays for an e-commerce site), when the company's own resources are fully utilized, additional capacity is rented as needed to maintain desired performance levels. The advantage is that it is generally cheaper to own something than to rent or lease it if it will be used most or all of the time, but if needed for a short duration, renting is cheaper. This provides the best of both worlds, minimizing the total cost required to meet required performance levels.
- Backup/Disaster Recovery, where data may be kept onsite, but backed up offsite somewhere, similar to the way that tapes used to be shipped offsite. It can also be used for companies that need a disaster recovery location, but only have a single datacenter for all their resources and need someplace they can run temporarily in the event of an emergency, much like companies like Sun Guard (now Sun Guard Availability Services) provided in physical datacenters in the past. In other words, they kept servers in a datacenter that could be powered up in the event of a disaster. These servers were available to multiple customers.