Overview

As a quick test of Mirror’s write performance, we ran a pipeline of each size backfilling Ethereum blocks data (~18.5M rows as of October 23, 2023) with an unmodified schema.

This test was performed on a 720 GiB RAM, 180 vCPU ClickHouse Cloud instance to ensure that non-pipeline factors (eg. disk IO, available memory, CPU cores) do not act as a bottleneck.

Each test was run on a completely clean instance (ie. no existing tables) with no other queries or commands running on the database.

Results

Pipeline sizePeak (rows/m)Time to backfill
S350,00053min
M950,00020min
L2,100,0009min
XL3,500,000~5min
XXL6,300,0003min