After a power failure hit one of Amazon’s data centers, several popular Web Sites went down which hosted their data in the Amazon cloud. This is an perfect example for the vulnerability of centralized systems: they have a single point of failure – in this case Amazon’s data center. If such a system fails, everything goes down. This physical aspect of cloud based systems is something that is worth investigating. There is a book that discusses the physical nature of the Internet and the book sheds light in how the “physical” data (more specifically the electric current) moves through the wires of the Internet.

Now, in contrast to centralized systems, (physically) distributed systems are more resilient to failures. Granted, there will be always critical nodes in distributed systems which, if compromised, will have negative impact on the system. But, and this is classical undergraduate textbook knowledge, even if failures occur, distributed systems can overcome the problems and continue to work. This is exactly what made the Internet so great.

Cloud systems and their advocates tend to forget this and the recent outage can be seen as wake-up call to think of alternative approaches, before putting the data of millions into one (physical) location.

your ikangai science team

Tags: Amazon, Cloud