Book Image

Ceph Cookbook - Second Edition

By : Karan Singh, Hackett, Umrao
Book Image

Ceph Cookbook - Second Edition

By: Karan Singh, Hackett, Umrao

Overview of this book

Ceph is a unified distributed storage system designed for reliability and scalability. This technology has been transforming the software-defined storage industry and is evolving rapidly as a leader with its wide range of support for popular cloud platforms such as OpenStack, and CloudStack, and also for virtualized platforms. Ceph is backed by Red Hat and has been developed by community of developers which has gained immense traction in recent years. This book will guide you right from the basics of Ceph , such as creating blocks, object storage, and filesystem access, to advanced concepts such as cloud integration solutions. The book will also cover practical and easy to implement recipes on CephFS, RGW, and RBD with respect to the major stable release of Ceph Jewel. Towards the end of the book, recipes based on troubleshooting and best practices will help you get to grips with managing Ceph storage in a production environment. By the end of this book, you will have practical, hands-on experience of using Ceph efficiently for your storage requirements.
Table of Contents (15 chapters)

Introduction

In the previous chapters, we covered different ways to deploy, provision, and administer Ceph. In this chapter, we will cover benchmarking the Ceph cluster, which is a must-do thing before moving to production. We will also cover advanced methods of Ceph administration and troubleshooting using the admin socket, REST API, and the ceph-objectstore-tool. Finally, we will learn about Ceph memory profiling.

Benchmarking your Ceph cluster before using it for the production workload should be a priority. Benchmarking gives you approximate results on how your cluster will perform during read, write, latency, and other workloads.

Before doing the real benchmarking, it's a good idea to establish a baseline for the expected maximum performance by measuring the performance of the hardware connected to the cluster node, such as the disk and network.

...