PHP v. Perl v. Python ... Questions ...

Joe Stump joe at digg.com
Mon Jan 7 18:06:21 UTC 2008


Has anyone benchmarked Gearman's speed when running no-op tasks? I'm  
seeing absolutely bizarre things on Linux (2.6.12). I've set up  
workers and clients in Python, Perl and PHP and get odd behavior from  
all of them (though I'm most interested in getting my PHP code fixed  
at this point).

The test setup is 1 gearmand, 1 job that just returns true and a  
client that runs that task 25 times in a taskset. PHP and Python both  
run these tasksets in the 40ms range. Python is more stable sticking  
pretty true at about 41ms per run. PHP is a bit more unstable with  
values ranging from 15ms to 63ms (though the average is around 40ms).  
Perl is even more bizarre in that it ranges from 0.6ms to 39ms. What  
would account for these tasks running in 0.6ms?

An example test run for each is below ...

--Joe

Python:
0.0419089794159
0.0421991348267
0.0101318359375
0.0425651073456
0.041759967804
0.0417029857635
0.0421628952026
0.00922393798828
0.00925397872925
0.0469858646393

PHP:
T: 0.0181191
T: 0.0159421
T: 0.0288019
T: 0.0472040
T: 0.0555351
T: 0.0635679
T: 0.0452640
T: 0.0448749
T: 0.0447688
T: 0.0443289

Perl:
0.000616
0.0401
0.040389
0.040349
0.039609
0.040174
0.000623
0.040178
0.039257
0.03987


More information about the Gearman mailing list