View Single Post
Old 29th November 2011
Lives for gear

Originally Posted by AwwDeOhh View Post
That's what i would figure.
The cost of implementation is a fixed cost. The cost per usage goes down the more people you multiply it by.
That's too simplistic.
A solution simple enough to be economic at a small scale often does not scale up. I'm very familiar with that scenario, many of the products I work on are large scale.

For a small ISP, you might wrap a few lines of SQL around an Access database on a Windows box and link it to the AAA server. For a large ISP, it's a major project - you have the cost of the redundant (usually Sun) multicore servers and SAN, the licensing of the (usually Oracle) Enterprise database, the planning and provisioning of the necessary bandwidth to the AAA servers and the extra load thereon, the cost of writing and testing the code, and one or more extra people to administer the system, do the regulatory paperwork and handle queries / complaints from subscribers.

For a familiar example,think about a car diesel engine. That's a small scale product, with certain design decisions that make sense at that scale. Now think about a ship's diesel engine. Same basic principle, many different design decisions because of the scale difference.
Wärtsilä-Sulzer RTA96-C - Wikipedia, the free encyclopedia
Most powerful diesel engine in the world

Regardless of the size of the system, you can't make it fully automatic. It will make mistakes. You either handle the mistakes manually before sending the notices, or you manually handle the calls from the subscribers later. Any such manual process does not scale - the man-hours involved are directly proportional to the number of subscribers / notices.