This week's cloud computing discussion covers Service Oriented Architecture (SOA) and Hadoop-style massively parallel data processing systems. SOA is interesting because this is how new Internet services are being developed. It is also a huge engineering challenge.
An epic blog post that helped me understand the importance of this was written by software engineer Steve Yegge [1], it is known as Stevey's Google Platforms Rant [2]. Yegge used to work for Amazon and now works for Google. Apparently, it was meant to be internal to Google, but it was accidentally published to great acclaim. Please read it for his passionate advocacy of a service oriented architecture and developer tools, and for his rather humorous, if somewhat salty and irreverent, description of life while working at these software companies.
Hadoop is an amazing system that was started by Doug Cutting, who wanted to provide the means to be able to index the entire Internet overnight, which at the time, only Google was doing effectively. Please read the Wikipedia entry on Apache Hadoop [3] for background. Hadoop is quite powerful, but also notoriously tricky to get working. Amazon has an interesting service called Elastic MapReduce [4] which claims to take a way a lot of the pain of setting up and maintaining such systems.