While the default configuration of Linux and Ceph will likely provide reasonable performance due to many years of research and tweaking by developers, it is likely that a Ceph administrator may want to try and squeeze more performance out of its hardware. By tuning both, the operating system and Ceph, performance gains maybe realized. In the Chapter 1, Planning for Ceph you learned about how to choose hardware for a Ceph cluster; now let's learn how to make the most of it.

In this chapter, you will learn about the following topics:

• Latency and why it matters
• When you should tune
• The dangers of blindly setting configuration options
• The importance of benchmarks and when you should learn to ignore them
• The importance of being able to observe the results of your tuning
• Key tuning options that you should look at

It's important to understand that by tuning, all you are doing is reducing bottlenecks. If you manage to reduce enough bottlenecks in one area, the bottleneck will simply shift to another area. You will always have a bottleneck somewhere, and eventually, you will reach a point where you are simply overt the limit of what a particular hardware can provide. Therefore, the goal should be to reduce bottlenecks in the software and operating system to unlock the entire potential of your hardware.