It's a very strange benchmark. 1) Clickhouse allows to trivially setup a batch buffer on server side. It will always be faster on insert with low disk usage.

It's a very strange benchmark. 1) Clickhouse allows to trivially setup a batch buffer on server side. It will always be faster on insert with low disk usage. 2) Suspicious axis on query benchmark. It seems it represents diff in query latencies. So if one executes query in 1.00 and another in 1.34 on graph it will be whopping 34.