List:General Discussion« Previous MessageNext Message »
From:Michael Widenius Date:September 22 1999 12:54pm
Subject:MySQL tuning on SPARC
View as plain text  
>>>>> "Orlando" == Orlando Andico <orly@stripped> writes:

Orlando> Hello,

Orlando> I've installed MySQL (3.22.26) on a Sun E250 Ultrasparc (Solaris 7). I
Orlando> just ran top(1) and get the following information:

Orlando> load averages:  1.18,  1.14,  0.88  16:02:51
Orlando> 33 processes:  31 sleeping, 1 running, 1 on cpu
Orlando> CPU states:  0.8% idle, 51.3% user, 47.9% kernel,  0.0% iowait,  0.0% swap
Orlando> Memory: 1152M real, 812M free, 87M swap in use, 987M swap free

Orlando>   PID USERNAME THR PRI NICE  SIZE   RES STATE   TIME    CPU COMMAND
Orlando> 15138 mysql      6  10    0   68M   42M run    14:09 85.25% mysqld

Orlando> I've already set up mysql with the following options (contrary to
Orlando> instructions, I modified the mysql.server file directly):

Orlando> $bindir/safe_mysqld --user=$mysql_daemon_user --pid-file=$pid_file
Orlando> 	--datadir=$datadir \
Orlando> 	-O key_buffer=64M -O table_cache=128 -O sort_buffer=2M \
Orlando> 	-O record_buffer=2M -O tmp_table_size=8M -O join_buffer=2M &

Orlando> However this is STILL not using the machine's resources optimally. I'd
Orlando> like to see at most 200-300MB of free memory (not the 800-plus MB listed
Orlando> above!) but I don't know what parameters need to be set. I recall a post
Orlando> on the list where Monty said improperly setting these parameters will only
Orlando> worsen performance.

Hi!

As always everything depends on your queries.  Note that any free
memory will be used by our operating system to cache rows.

At any rate, the safe buffers to increase are key_buffer and
table_cache. To increase any other may help or make things worse...

We are working on a select result cache;  This will make it possible
to put aside a big cache to hold result for old queries...

With MySQL 3.23 you can also copy some read-only tables to heap tables
(RAM based tables) on startup.

Regards,
Monty
Thread
MySQL tuning on SPARCOrlando Andico22 Sep
  • MySQL tuning on SPARCMichael Widenius22 Sep