List:General Discussion« Previous MessageNext Message »
From:Michael Widenius Date:May 10 1999 10:51pm
Subject:Benchmark question
View as plain text  
>>>>> "Suresh" == Suresh Rajagopalan <sraja@stripped> writes:

Suresh> Hi;
Suresh> Comparing the benchmark results I have got from a P450/Linux 2.2.5 box vs
Suresh> the one supplied by Monty, I find that the insert test is much slower.  
Suresh> All other tests compare well.

Suresh> I've tried both my compiled version (using the Pentium optimized settings)
Suresh> and also the binary RPM from the distribution. 

Suresh> I'm not sure if the slowness can be attributed to 1 less CPU and the
Suresh> lesser amount of RAM.      The machine I used had a SCSI disk.

Suresh> In both cases, logging was turned off and debug was *not* compiled in.
Suresh> key_buffer was set to 16m in both tests.

Suresh> Thanks for any pointers.

Suresh> Suresh

Suresh> ------------------------------------------------------------

Suresh> The result logs which where found and the options:
Suresh>  1 mysql-Linux-2.2.5_450Mhz,                : MySQL3.22.21  
Suresh>   PIII 450 1 CPU 128M key_buffer=16m static compiled, egcs 1.1.2, SCSI

Suresh>  2 mysql-Linux-2.2.5_450Mhz_binary_RPM     : MySQL 3.22.21  
Suresh>   PIII 450 1 CPU  128M key_buffer=16m binary RPM, SCSI

Suresh>  3 mysql-Linux_2.2.1_i686                  : MySQL 3.22.18 
Suresh>   pentiumpro 400mz x2, 256M, SCSI, gcc 2.9 compiled, key_buffer=16M


Suresh> insert_key (100000)          |   1948|   1317|    156|


The insert_key() tests insert 100000 rows in a table with 16 simply
keys and one key 16 key parts.  This creates a table which uses about
50M disk space.  In this case the extra 128M memory in my machine is
probably the reason for the speed difference.

WHERE clauseGraeme B. Davis4 May
  • Re: WHERE clausePaul Wolstenholme4 May
    • Benchmark questionSuresh Rajagopalan4 May
      • Benchmark questionMichael Widenius11 May
  • Re: WHERE clauseChristian Mack4 May