Chapter 12. Using Formal Theorems to Understand Trade-Offs
Essentially, all models are wrong, but some are useful.
George Box and Norman Draper
When things go wrong, we need to make sure that as many things as possible go right. Typically this means that we need to gracefully degrade our functionality in some way. We’ve looked at a number of examples of this approach throughout this book, including things like load shedding to get rid of excess load or using bulkheads to protect us from parts of the system that are malfunctioning. But we can make some fundamental trade-offs in the design of the system itself to ensure that when things go wrong, our system handles it in the way that is best for the given context.
In this chapter, we’ll look at two formal theorems, CAP and PACELC, which describe the trade-offs that have to be considered when parts of a distributed system are unable to communicate. Put another way, these two models help us understand what options we have to gracefully degrade our application when something goes wrong. We are also going to look at a concept known as Little’s Law, which will help you reason about how much load your system can handle and about how to keep load to manageable levels.
While all these concepts can appear at first to feel quite academic and removed from the software you build, I’ll share some examples of these models in action to help you see that in reality they are just formal definitions of fairly straightforward ideas. I also say upfront ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access