云计算的基本特征

When the United States National Institute of Standards and Technology (NIST) published the cloud computing definition, they also defined the essential characteristics of this new model. These have come to be more important than the definition in that the characteristics have helped to define and protect the marketplace against all the marketing hype that has accompanied the cloud.

The first characteristic of cloud computing is that it is an on-demand, typically self service model. On-demand, meaning that it can be purchased when needed, for as long as needed, and given back when finished. Self service refers to the consumer's ability to buy, deploy and shut down services without any assistance from the service provider. This speeds up the process controls cost and moves control to the consumer. (Refer back to earlier paragraphs where we discussed the de-centralization and continual innovation of pushing compute and control closer to the edge of the consumer's control and consumer-controlled devices. The same applies here.)

From a security perspective, this has introduced governance challenges about the acquisition, provisioning, use, and operation of cloud-based services. Interestingly, these new services may violate existing organizational policies. By its nature, cloud computing may not require procurement, provisioning, or approval from finance due to its low initial cost, self-service nature, and immediate deployment options. Cloud infrastructure and services can be provisioned by almost anyone with a credit card, also known as shadow IT. For enterprise customers, this low-entry cost, quickly deployed on-demand model may become one of the most important characteristics as it instantly wreaks havoc on governance, security, long-term cost, strategy, internal politics, and collaboration.

The second characteristic, broad network access, is required. Ever heard the phrase the network is the cloud? Anything referred to as-a-service requires a network connection. How would it be accessed, managed, operated, or utilized without some network connection, typically to the internet using standard protocols that promote use by disparate client platforms? Because the cloud is an always-on and always-accessible offering, users have immediate access to all available resources, and assets. Think convenient access to what you want, when you need it, from any location. In theory, all that is required is internet access and relevant credentials. The mobile device and smart device revolution have introduced an interesting dynamic into the cloud conversation within many organizations. These devices are often able to access relevant resources that users require; however, compatibility issues, ineffective security controls and non-standardization of platforms and software systems have made the first adoption climb more difficult for some enterprises.

The third characteristic, resource pooling, is the characteristic that, in essence, lies at the heart of all that is good about cloud computing. Combining many smaller compute resources into farms or pools that can serve many consumers simultaneously enables dynamic resource allocation and re-allocation, cost predictability, IT resource control, and higher rates of infrastructure utilization. Utilization and consumption patterns directly affect cost. Resource pooling enables different physical and virtual resources to be allocated and re-allocated according to consumer demand. As mentioned earlier, the true cloud innovation was economic, allowing us to stop billing and give back the resource when finished. More often than not, traditional, non-cloud traditional deployments see low- utilization rates for their resources, typically between 10 and 20%. Cloud deployments from pools used across multiple clients or customer groups can see as high as 80 to 90% utilization (100% is not ideal in most cases). Resources can automatically scale and adjust to dynamic needs, workload or resource requirements. Cloud service providers or cloud solution providers (CSPs) typically have scores of resources available, from hundreds to thousands of servers, network devices, and applications, enabling them to quickly and economically accommodate, prioritize and implement the varied size, and complexities each client presents.

The fourth essential characteristic of cloud computing centers on elasticity, the ability to dynamically match the need. Product and service capabilities are developed, acquired, priced, and provisioned elastically, enabling rapid response to continuously changing user demand. To the consumer, capabilities often appear unlimited and easily deployed in any quantity at any time. Because cloud services utilize a consumption-based pay-per-use model, you only pay for what you use. As mentioned earlier, cloud innovation and adoption are being driven mainly by economics that affects strategy. For cyclical loads, applications with intermittent use, seasonal or event-type business cloud eliminates the need to pay for 100% of a physical server (CAPEX) when only 5% is used 2% of the time (OPEX). Think of selling thousands of tickets to an Olympic event. Leading up to the ticket release date, little to no computing resources are needed; however, when the tickets go on sale, they may need to accommodate 100,000 users in the space of 30 minutes-40 minutes. This is where rapid elasticity and cloud computing can be beneficial. Enterprises no longer require traditional IT deployments with substantial capital expenditure up front (CAPEX) to support the temporary project load.

The final key characteristic mentioned here is that the cloud is a constantly measured service. Cloud computing natively offers a unique and important component that traditional IT deployments have struggled to providemeasurement and control of resource consumption and utilization. As mentioned often, billing was the big innovation. Cloud resource consumption needed to be measured and billed for accurately. Once that was possible, the true power of the cloud, which included the ability to shut it off, was realized. The capability enabled automated reporting, monitoring, and alerting which provided much-needed transparency between the provider and the client. Like a metered electricity service or cell phone data usage, consumers have transparent and immediate access to usage data enabling immediate behavior change if needed. Itemized billing provides transparent trendable data providing insight that may lead to needed change. Proactive organizations can now utilize this well measured, transparent, granular, trendable data to charge departments or business units for their actual consumption. IT, product development, and finance can now move toward operating collaboratively as a revenue-driving team that can quantify, qualify, and justify exact usage and costs per department, by business function, per leader, and so onsomething that was incredibly difficult to achieve in traditional IT environments. The following diagram is a graphical representation of the five essential characteristics of cloud computing:

As a side note: people have been utilizing the cloud for years without realizing it. It is not a new thing, but it has just started to port over into more popular arenas. Let's look at internet access as an example.

Characteristic 1: Based on the first characteristic mentioned earlier, how many people dig up the street to put in connectivity when they want to access the internet? None. We pay for it as a service that allows us to use it when we choose. Characteristic 2: Do we need a network to access the internet? Of course. Characteristic 3: Do we have dedicated switches, routers, SONET ring, and so on in our living space? No, those resources are pooled by the service provider and shared with all the clients in the area or region. Characteristic 4: Can we utilize more if we need it? Absolutely. We only use what we need with the ability to scale all the way up to the maximum performance for which we are willing to pay. If more is needed, we call and change what performance level we pay for to match up with the changing need. Characteristic 5: Are we paying as we go? Is our service metered and measured? Definitely yes. If we choose to stop paying for the service, the service is shut off. We get a bill every month and often have a portal that we can log in to that details what we pay for, what we use, performance details, uptime, downtime, and so on.