On Mon, May 20, 2013 at 4:42 PM, Andrey <[email protected]> wrote: > Ben, > > 2M result I mentioned is probably not very accurate (but real) - it's 500k > rows query with single numerical field, and parseInt(buffer.toString()) > seems to be much slower than manual int parsing ( > https://github.com/sidorares/node-mysql2/blob/master/lib/packets/packet.js#L183 > ) Here is simple benchmark manual vs builtin parseInt from buffer: > https://gist.github.com/sidorares/5612620 - there is 10x times difference. > > Inner loop row parser is compiled dynamically based on field types - > https://github.com/sidorares/node-mysql2/blob/master/lib/compile_text_parser.js > Removing Function.bind() also helped to save 10-20% couple of times. > De-framer stream to packets does not buffer incoming data and in most cases > executes parser on incoming chunk without copying. > > Results on 2 DigitalOcean small servers (512M Ubuntu 12.10 server 32bit, 1 > CPU) using https://github.com/mscdex/node-mysql-benchmarks > > 150k insert+select > Results (init time in seconds, other values in ops/s): > module,init,escapes,inserts,selects > =================================== > C,1.009,6790961,1728,299671 > PHP,0.014,1501147,1632,209946 > mysql,0.022,674157,1369,89074 > mysql2,0.02,691882,1467,340909 > mysql-libmysqlclient,0.017,1570681,1721,142315 > mysql-native,0.02,0,1503,199734 > mariasql,0.013,1662971,1594,282486 > > 500k inserts + select > Results (init time in seconds, other values in ops/s): > module,init,escapes,inserts,selects > =================================== > C,0.021,6809424,1696,340689 > PHP,0.021,1499440,1655,206863 > mysql,0.035,705799,0,0 > mysql2,0.024,690131,1512,384911 > mysql-libmysqlclient,0.031,1860465,1569,161290 > mysql-native,0.028,0,1445,187829 > mariasql,0.029,1392434,1594,257467 > > I definitely interested to see more real life benchmarks.
Thanks, that makes sense. So basically node-mysql is leaving some obvious optimizations on the table? By the way, I noticed that you're using vm.runInNewContext(). Is there a reason for that? You could probably reduce memory usage a little by running the generated code in the same context. -- -- Job Board: http://jobs.nodejs.org/ Posting guidelines: https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines You received this message because you are subscribed to the Google Groups "nodejs" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/nodejs?hl=en?hl=en --- You received this message because you are subscribed to the Google Groups "nodejs" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
