Jump to content

Database benchmark

From Meta, a Wikimedia project coordination wiki

Currently doing some benchmarks of MySQL/MyIsam vs. BerkleyDB(direct access) using these python scripts:

Berkley:

import bsddb
db = bsddb.btopen('db')
for i in xrange(0,5736615):
    t = str(i)+'testx'
    db['test'] = t
db.close()

result:

real    3m29.724s
user    3m19.539s
sys     0m0.279s

MySQL:

import MySQLdb
db = MySQLdb.connect(db='wikidb2', user='root')
c = db.cursor()
for i in xrange(0,200000):
    t = str(i)
    c.execute("update links set l_to = '"+t+"' where l_from = '999999';")
db.close()

result:

real    3m17.025s
user    3m4.017s
sys     0m0.802s

So mysql is about 43 times slower for this simple query.