I’m about to reuse a CentOS 7.5 system with hardware RAID to turn it into an IPFS node, and was wondering whether anyone has any experiences or recommendations to share with respect to configuring and optimizing the file system parameters. The linux kernel has a number of parameters (e.g., the scheduler, fifo queue sizes and more), and I already have a little bit of experience tuning parameters using workloads that use the file system directly. However, I have no idea how IPFS will change the read/write behavior on the actual file system, so my past tuning experiences may not be relevant now. I’m curious what other people have done to maximize performance.
Related to that, does anyone have any kind of repeatable load/exercise suite that could be used to do parameter tuning for IPFS servers? What I’m looking for here is some kind of test I can run repeatedly in a loop, alternating between changing file system parameters and testing the results.
(Hardware is a Dell R730xd with Perc H730 Mini controller and 8 x 4 TB 7200 RPM drives, in case that’s relevant. Expected preliminary content to be made available over IPFS is scientific images and videos.)