Why is MySQL nearly twice as fast on Linux?

[David Xu]

Daniel Eischen wrote:

>On Sat, 22 May 2004, JG wrote:
>  
>
>>>Do you have any suggestions or ideas as to what the problem is?
>>>      
>>>
>
>[...]
>
>  
>
>>(Compiled with libpthread)
>>
>>Also w/ http://people.freebsd.org/~deischen/mysql40-server.diffs patch.
>>-------------------------------------------------------------------------
>>
>>BENCHMARK RESULTS:
>>
>>Local test:
>>
>>- Using AMD64-4BSD SMP kernel w/SCHED_4BSD
>>- Using default mysql config (no config file)
>>
>>amd64f# super-smack update-select.smack 30 10000
>>Query Barrel Report for client smacker
>>connect: max=25ms  min=4ms avg= 15ms from 30 clients
>>Query_type      num_queries     max_time        min_time        q_per_s
>>select_index    300000  11      0       2819.41
>>update_index    300000  7       0       2819.41
>>
>>
>>- Using AMD64-4BSD-NOSMP SINGLE CPU kernel w/SCHED_4BSD
>>- Using default mysql config (no config file)
>>
>>amd64f# super-smack update-select.smack 30 10000
>>Query Barrel Report for client smacker
>>connect: max=21ms  min=9ms avg= 16ms from 30 clients
>>Query_type      num_queries     max_time        min_time        q_per_s
>>select_index    300000  5       0       3015.96
>>update_index    300000  23      5       3015.96
>>    
>>
>
>Along with the other benchmarks it's still a little apples
>and oranges.  I've seen different versions of mysql run
>with different versions of FreeBSD.  I'd like to see the
>same mysql version run on the same hardware with libpthread
>and linuxthreads.  What I've seen so far is close WRT
>OS and mysql versions, but still not the same:
>
>  mysql-4.1.1-alpha on FreeBSD 5.2-RELEASE with linuxthreads,
>   SCHED4_BSD, amd32
>  ----------------------------------------------------------
>  Local:
>
>  - Using default mysql config settings (empty/missing my.cnf)
>  - Using SCHED_4BSD kernel
>
>  amd32f# super-smack update-select.smack 30 10000
>  Query Barrel Report for client smacker
>  connect: max=19ms  min=12ms avg= 14ms from 30 clients
>  Query_type      num_queries     max_time        min_time        q_per_s
>  select_index    300000          9               0               2589.28
>  update_index    300000          7               0               2589.28
>
>
>  mysql-4.0.20 (with process-scope patch) on FreeBSD 5.2-current
>    with libpthread, SCHED4_BSD, amd64
>  --------------------------------------------------------------
>  2819.41 (SMP), 3015.96 (UP), see quoted email above
>
>This shows that -current (both SMP and UP) with libpthread is
>faster than linuxthreads (different versions of FreeBSD and
>mysql, though).
>
>Also, if you use the default mysql config settings, then
>FreeBSD remote vs Linuxthreads remote also wins (2685.53
>vs. 2091.10 from earlier post).
>
>What modifications are you making to /etc/my.cnf and why
>do they turn the results around the other way?
>
>  
>
I have done a test. The machine is Athlon XP 3000+
with KT600 chipsets.
MySQL with libpthread using scope process thread has
best peformance, it gets 3200 q_per_s while scope system
only gets 2900 q_per_s. Linuxthreads gets 3100 q_per_s,
and this is same with libthr. I don't have /etc/my.cnf installed.

David Xu