I’m not really understanding the difference between cloud computing and regular server hosting. Doesn’t a cloud still have to be hosted on a somewhere? I understand that a cloud is just something where people put a website, or some data that someone can access anywhere and we can increase decrease the amount of clouds based on the amount of traffic.
But how is the accessibility any different than regular server hosting and where is a cloud hosted?
1. Firstly, “cloud computing” is a vague term created by marketing as a set of features, and diluted by sales people pushing services when applications aren’t obvious to their customers. I will assume we’re mainly discussing elastic computing and any technologies necessary to implement that, like hypervisors and distrubuted storage.
Elastic computing is a tool to scale your computer power up and down as needed. It’s related to time-share, but instead of one large mainframe to rent server time on, you’re given a large cloud of servers to rent or share. You can script the start and closing of additional nodes, to match your use of the cloud to demand for the services those nodes offer.
The important distinction between elastic compute clouds and normal hosting is provisioning. Imagine you run a website that publishes football scores, and you’re very popular. To make a profit you need to keep the website responsive under heavy load. We’re talking Superbowl heavy load. Constant refreshes and sustained traffic for hours. In order to meet that goal, you could buy a massive server farm that can handle Superbowl traffic, and let them sit mostly idle during the off-season. Or you could buy server time from an elastic compute cloud to make up the difference. Normal hosting services may choose to simply fail during high load, with catastrophic effects on your Superbowl revenue. They may even kick you off for too much CPU use or network traffic.
Economically, cloud computing allows for full employment of servers. Rather than have everyone buy lots of beefy hardware in case of Slashdot, the hardware that would serve Slashdotters can migrate to the sites that need it (and pay for it). Combined with economies of scale, we can expect that large compute farms may become cheaper than hosted or colocated solutions. If APIs are created to migrate servers between clouds, additional competitive forces may help drive prices towards marginal costs; hence the chasm between Amazon and the Cloud Computing Bill of Rights. Some are proposing a cloud marketplace, where cloud computing is bought and sold by principles of supply and demand. This would encourage people to shift compute power to off peak hours, as we see with cell phone plans and industrial use of electricity.
The reasons to stay away from cloud computing are twofold: price, and privacy. None of the above guarantees cloud computing will be cheaper than your current solution. You may be fine with failure during Superbowl events. Or it may be cheaper for you to build and buy your own servers and datacenter. Alternatively, you may have data you would prefer not reside in the hands of anonymous cloud vendors whose security and technology may leak information about your service or your customers. The last part means you may in fact be legally impaired from implementing cloud computing, as the cloud vendor has access to your disk and RAM.
on: 24th March 11