What is cloud computing? Really? It doesn’t have any official, legally accepted definition. In the broadest sense, it just means “someone else’s computer.” When you look for cloud computing, you expect something more than that, but you need to make sure that’s what you’re getting. Otherwise it’s just a remote data center at best. If it doesn’t live up to certain standards, it’s a cloud service in name only.
The NIST definition
NIST defines cloud computing by five characteristics:
- On-demand self-service.
- Broad network access.
- Resource pooling.
- Rapid elasticity.
- Measured service.
This isn’t a legally binding definition, but it matches what most people expect. If a service doesn’t offer these things, it’s fair to call it a “cloud impersonator.”
The self-service feature means that you can set your resource requirements interactively, without any human intervention. If your needs go up, you can switch to a higher service level. With cloud impersonators, you’re typically stuck with a fixed piece of hardware or a share of a machine, and you have to go to the support department to get a change, or even re-negotiate your contract.
Network access on a NIST-compliant cloud doesn’t tie you to a dedicated client. In most cases, you can access the services through a Web browser or a published API. An impersonator might tie you to specific client hardware or software.
Resource pooling is a big one. Without it, you’ve got just a data center. With a real cloud system, you aren’t tied to a specific piece of hardware. You generally don’t even know which machine is running your services at any moment. Proximity can be useful when you want to keep latency to a minimum, so you may be able to request resources in a specific city or data center, but even in that case you might be using different computers at different times.
Elasticity means that you can get as little or as much service as you need without disruption. If your needs are small, you’ll use just part of a computer. Other people will pay for the part of the machine you aren’t using. If your needs go up, you can increase the amount of computer power you’re getting. The service might move you to a faster machine with more memory, and you don’t have to reinstall or even restart anything.
Measured service means that you pay for the amount of service that you use. “Amount of service” might have different kinds of measurement with different vendors; it might mean your CPU usage, storage, amount of data transferred, number of accounts, or some combination. You pay for what you get, rather than paying a fixed rate.
Data centers and cloud variants
Data centers where you lease specific hardware and pay a fixed rate have many legitimate uses. The only problem is that they aren’t cloud services, except in the broad “someone else’s computer” sense. When someone offers you one thing and calls it another, you need to be careful. What else might they not be telling you?
There are variations on the cloud model. A private cloud is a group of servers allocated to a single client. It provides greater security and may make performance more reliable, but doesn’t give as much flexibility as a public cloud. Resource pooling and elasticity are limited to the machines allocated to the client, with users or services trading off against each other.
A hybrid cloud setup keeps some services on the client’s on-premises computers, where they interact with cloud services. This is appropriate where security requirements don’t allow data to go outside the client’s own computers, or where it’s important to minimize latency. Interaction with the cloud services is usually through publicly defined APIs.
“Cloud” can mean anything or nothing. When looking for cloud services, look carefully at what a vendor offers, and make sure it satisfies the level of service you need.
At Applied Innovations, we provide managed cloud hosting to meet the most demanding needs. Please contact us to learn more.