Chapter 10. Ceph Performance Tuning and Benchmarking

In this chapter, we will cover the following topics:

  • Ceph performance overview
  • General Ceph performance consideration – hardware level
  • Ceph performance tuning – software level
  • Ceph erasure coding
  • Ceph cache tiering
  • Ceph benchmarking using RADOS bench

Ceph performance overview

Data and information always keep on changing and increasing. Every organization, whether small or large, faces data growth challenges with time, and most of the time, these data growth challenges bring performance problems with them. In this era, where the world is generating enormous amount of data, organizations have to look for a storage solution that is highly scalable, distributed, extremely reliable, and on top of that, well performing for all of their workload needs.

One of the several biggest advantages of Ceph is that it possesses scalable, distributed architecture. Due to its distributed nature, all the workload that comes to Ceph is evenly distributed and allocated to the entire cluster for storage, which makes Ceph a good performing storage system. You can imagine storing a data file on a traditional storage system is quite limited in scalability, almost nondistributed, and has only a limited numbers of data storage nodes and disks. In such a kind of traditional setup, performance problems might occur with data growth and increasing user requests.

These days, when applications serve a huge number of clients, they require a better-performing backend storage system. A typical Ceph setup contains numerous nodes, and each node contains several OSDs. Since Ceph is distributed, when data arrives to Ceph for storage, it's distributed across several nodes and OSDs beneath, delivering an accumulative performance of several nodes. Moreover, the traditional systems are limited to a certain performance when you add more capacity to them, that is, you will not get added performance with increasing capacity. In some scenarios, the performance degrades on increasing the capacity. With Ceph, you will never see performance degradation with increase in the capacity. When you increase the capacity of a Ceph storage system, that is, when you add new nodes full of OSDs, the performance of the entire storage cluster increases linearly since we now get more OSD workhorses and added CPU, memory, and network per new node. This is something that makes Ceph unique and differentiates Ceph from other storage systems.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
3.135.187.210