I'm using Pebble as the storage engine for one of my projects. My approach involves sequential reading from the Pebble DB. It's noticed that this process is typically constrained by the IOPS of the disk. Specifically, I've provisioned AWS EBS for 8k IOPS.
However, I made an observation: by adjusting the block size to 128k (as opposed to its default 16k), I managed to eliminate the IOPS constraint. As a result, the overall performance experienced a significant enhancement.
My understanding was that Pebble incorporates read-ahead mechanisms and consistently retrieves a fixed-size portion from the filesystem, subsequently interacting with the underlying disk. The correlation between increased block size and improved performance is unclear based on this premise.