Principles for Scalable Service Design
Posted by Igor Moochnick on 03/10/2009
I’m preparing a presentation for the upcoming CodeCamp about the best practices in building enterprise system that can scale to the cloud and stumbled on a real gem – the Dr. Werner Vogels Availability & Consistency presentation during QCon 2007. For those who doesn’t know – Dr. Werner Vogels is VP & CTO at Amazon.com.
In this presentation Dr. Vogels in crystal clear way lays out all the principals that I was a big promoter for the last 10 years. I agree with him on 100,000% and really think that the future is with the BASE scalable systems (and not with ACID) simply because ACID propagates the error to the end user whereas BASE system deals with errors locally. I do agree that in some rare cases ACID is the easier and cheaper way to go, but still sure that even those cases is possible to implement as BASE system.
The Dr. Vogels recipie for success is build system following this menu:
- Controlled concurrency
- Controlled parallelism
- Decompose into small well-understood building blocks
- Failure tolerant
- Local responsibility
- Recovery Built-in