Chapter 1 - Reliable, Scalable and maintainable applications

Data systems don't fit into clear categories now, redis is used for queues, kafka has database like durability guarantees.
Application code is responsible for making sure data remains consistent between any number of data systems the application might not be using.
Each application made from smaller data systems is a new data system. You're also a data system designer.
Lots of tricky questions: ensure correctness, design a good API, reliability, performance.
Many factors influence design: skills and experience of people, timescale for delivery etc.
Three concerns that are most important in systems:
- Reliability
- Scalability
- Maintainability

Application should performs the function that the user expected
Should tolerate the user making mistakes.
Performance should be good enough
Should prevent unauthorized access and abuse
Things that can go wrong are called faults.
Failures are when the system as a whole stops providing the required service to the user.
Fault tolerance doesn't mean being tolerant of ALL faults.
It's best to try to design a system that prevents faults from causing failures, i.e it tolerates faults.
You also want to prevent certain types of faults like security issues.

Hardware failures are generally not related, but correlated software bugs can cause the entire system to go down.
Systematic errors can be avoided by thorough testing, carefully thinking about assumptions and interactions in the system.