How to gauge cloud computing performance

There are three models of cloud computing, and the one you use determines the kind of performance you get

Does cloud computing perform well? That depends on whom you ask. Those using SaaS systems and dealing with standard Web latency can't tell you much about performance. However, those using advanced "big data" systems have a much different story to relate.

[ Get the no-nonsense explanations and advice you need to take real advantage of cloud computing in InfoWorld editors' 21-page Cloud Computing Deep Dive PDF special report. | Stay up on the cloud with InfoWorld's Cloud Computing Report newsletter. ]

You need to consider the performance models, which you can break into three very basic categories:

• Client-oriented (performance trade-off)
• Cloud-oriented (performance advantage)
• Hybrid (depends on the implementation)

Client-oriented cloud computing architectures are those systems where the cloud computing providers, typically SaaS (software as a service), interact with users constantly over the Internet. The issue here is not that the cloud provider is slow, but that there is latency with the constant back-end machine-to-machine conversation that occurs between the SaaS provider and the browser.

There's not much you can do about this, other than create your own expensive private link between your company and the SaaS provider, but that dilutes the value of SaaS quickly. Client-oriented platforms are clearly not as fast as applications running on the local network, but in most instances, the user won't notice the latency unless there is network saturation.

Cloud-oriented cloud computing architectures are those systems where the processing occurs within the cloud. Most infrastucture-as-a-service providers, and some platform-as-a-service providers, fit into this category. Typically, these systems can provide better performance than their on-premise counterparts because they have access to many more virtualized resources and can allocate those resources dynamically.