High Performance MySQL, 3rd Edition

Chapter 12. High Availability

This chapter covers the third of our little trio of topics: replication, scalability, and high availability. At the end of the day, high availability really means “less downtime.” Unfortunately, high availability is frequently conflated with related concepts such as redundancy, protection against data loss, and load balancing. We hope that the preceding two chapters have set the stage for a clear understanding of high availability. However, this chapter can’t be singularly focused; like the others in the trio, it must synthesize a few related topics.

What Is High Availability?

High availability is actually a bit of a mythical beast. It’s usually expressed as a percentage, which is a hint in itself: there is no absolute high availability, only relatively higher availability. 100% availability is simply impossible. The “nines” rule of availability is the most common way to express an availability goal. As you probably know, “five nines” means 99.999% uptime, which is just over five minutes of downtime per year. That’s pretty impressive for most applications, although some achieve even more nines than that.

Applications have vastly different availability needs. Before you set your heart on a certain uptime goal, ask yourself what you really need to achieve. Each increment of availability usually costs far more than the previous one; the ratio of availability to effort and cost is nonlinear. How much uptime you need usually depends on how much you can afford. ...