sahilseth / benchm-r-mysql

Benchmarking R MySQL Data Exchange

Time

Action	RMySQL	RMySQL-dev	mysql CL	RJDBC	RODBC	dplyr
Insert 1M rows [s]	12	800*	6	900*	800*	10
Read 1M rows [s]	2.4	1	2.5	3.4	2.7	2.2
Insert 1 row [ms]	5	4		12**	6	13**
Read 1 row [ms]	0.4	0.4		3.5	0.5	0.6

[*] CRAN RMySQL uses LOAD DATA for inserts and for 1M (million) rows is 100x faster than RJDBC/RODBC and RMySQL-dev (the later ones probably inserting 1 row at a time)

[**] RJDBC bug? cannot do append = TRUE, while dplyr will create new (temporary) table by design

RAM usage

While reading the 1M rows dataset RAM goes up to a max value (RAM_max), then goes down and finally after a gc() call reaches RAM_after. Here is RAM/object.size:

Tool	RAM_max	RAM_after
RMySQL	2	1
RJDBC	8	7.5
RODBC	5	3.5

Versions

2015-07-04
R 3.2.1
mysql-server 5.5.43
DBI 0.3.1
RMySQL 0.10.3
RMySQL-dev 9fb0f8685a3 (with corresponding DBI)
RJBC 0.2-5 / mysql-connector-java-5.1.36
RODBC 1.3-12 / libmyodbc 5.1.10-3 / unixodbc-dev 2.2.14p2
dplyr 0.4.2