Book Image

Mastering Ceph, - Second Edition

By : Nick Fisk
Book Image

Mastering Ceph, - Second Edition

By: Nick Fisk

Overview of this book

Ceph is an open source distributed storage system that is scalable to Exabyte deployments. This second edition of Mastering Ceph takes you a step closer to becoming an expert on Ceph. You’ll get started by understanding the design goals and planning steps that should be undertaken to ensure successful deployments. In the next sections, you’ll be guided through setting up and deploying the Ceph cluster with the help of orchestration tools. This will allow you to witness Ceph’s scalability, erasure coding (data protective) mechanism, and automated data backup features on multiple servers. You’ll then discover more about the key areas of Ceph including BlueStore, erasure coding and cache tiering with the help of examples. Next, you’ll also learn some of the ways to export Ceph into non-native environments and understand some of the pitfalls that you may encounter. The book features a section on tuning that will take you through the process of optimizing both Ceph and its supporting infrastructure. You’ll also learn to develop applications, which use Librados and distributed computations with shared object classes. Toward the concluding chapters, you’ll learn to troubleshoot issues and handle various scenarios where Ceph is not likely to recover on its own. By the end of this book, you’ll be able to master storage management with Ceph and generate solutions for managing your infrastructure.
Table of Contents (18 chapters)
Free Chapter
Section 1: Planning And Deployment
Section 2: Operating and Tuning
Section 3: Troubleshooting and Recovery

Tuning tiering

Unlike the majority of Ceph's features, which by default perform well for a large number of workloads, Ceph's tiering functionality requires careful configuration of its various parameters to ensure good performance. You should also have a basic understanding of your workload's I/O profile; tiering will only work well if your data has a small percentage of hot data. Workloads that are uniformly random or involve lots of sequential access patterns will either show no improvement or in some cases may actually be slower.

Flushing and eviction

The main tuning options that should be looked at first are the ones that define the size limit to the top tier, when it should flush and when it should evict...