Cephalocon 2022

Benchmarking Ceph has always been a complex task - there are lots of tools but many have drawbacks and are written for more general-purpose use. For Ceph we need to benchmark Librados, RBD, CephFS, and RGW and each of these protocols has unique challenges and typical deployment scenarios. Not only that, Ceph works better at scale and so we need to ensure that we can build a benchmarking system that will also scale and be able to generate an adequate load at large scale. When sizing clusters it’s useful not just to sweep over typical things like object size and count, but also being able to sweep over workers to identify the correct benchmarking cluster size, and over OSD count to identify a suitable cluster size for our performance requirements is also useful. Sibench is an in-house tool that we’ve written at SoftIron to help us benchmark our solution and do a lot of the heavy-lifting for us. In this talk we’ll go over our benchmarking journey, what prompted us to write Sibench, how it works, and maybe also show a demo of the project in action.