SQLite benchmarks are completely WRONG

[socketpair]

To have comparable benchmarks:

instead of CREATE TABLE test (key blob, value blob, PRIMARY KEY(key)) please use CREATE TABLE test (key blob not null, value blob not null, PRIMARY KEY(key)) WITHOUT ROWID

and possibly even

CREATE TABLE test (
  key blob not null,
  value blob not null
  PRIMARY KEY (key, value)
) WITHOUT ROWID

to enable index-only scan.

Also, use select value from ... instead of select * from.

use insert into instead of REPLACE INTO.

issue VACUUM before making read-test.

Use latest SQLite version (3.15.0). It has in changelog: Miscellaneous micro-optimizations reduce CPU usage by more than 7% on common workloads.

Also, mention in HTML 64-bit OS or 32-bit OS is used.

In order to benchmark, we MUST use fastest way to implement tasks for each technology.

After these fixes PLEASE update benchmarking page.