Re: Benchmarking SQL drivers

Well done!

-- 
Peter Bortas

On Thu, Nov 20, 2014 at 3:36 PM, Stephen R. van den Berg <[email protected]> wrote:
> Well, now that the dust has settled, it was/is time to compare the results
> with other drivers:
>
> I ran a benchmark over a table with 1163 entries, 10 columns.
> Where a (very straightforward) SELECT pulls 166 of those entries out,
> returning two columns.  The intent is to benchmark the driver, not the
> quality of the SQL engine.
> I open a single connection to the database.
> I run 10 queries in rapid succession (big_query), fetch
> all results using fetch_row().
> And run those 10 queries in a loop for 394 times.
>
> What I measure is throughput to/from the database, the queries probably
> are IO-bound (they fit in the database cache) by the network(latencies).  The
> database is on a different machine on the same ethernet.
>
> 11.3s  pgsql    Pike 8.1 (new driver, multithreaded)
> 17.3s  MySQL    Pike 8.1
> 18.4s  pgsql    Pike 7.8 (old driver, single threaded)
> 32.7s  postgres Pike 8.1 (C-libpq)
>
> MySQL daemon runs 5.5.39.
> PostgreSQL server runs 9.3.4.
>
> Running the tests multiple times shows a variation of the timings
> of less than 1%.  The database server was idle most of the time.
>
> Running multiple queries at the same time over the same connection will
> likely reveal that you can run (up to a certain amount) multiple queries
> simultaneously without consuming extra wallclock time.
> However since the other three drivers do not support that, a comparison
> benchmark will be pointless.
> --
> Stephen.
>