❑ mysqlbench 설치 및 사용

1. 설치
# rpm -ivh MySQL-devel-xxxxxx.rpm
# rpm -ivh MySQL-shared-xxxxxx.rpm
# rpm -ivh MySQL-shared-compat-xxxxxx.rpm
# yum -y install perl-DBD-MySQL

# yum -y install perl-DBI

# yum -y install mysql-bench

2. 사용
# cd /usr/share/sql-bench

# perl run-all-tests --dir=/download/bench_output --server=mysql --host=localhost --machine=MySQLBench_CentOS5.6 --user=root --password=root --threads=10 --connect-option=mysql_read_default_file=/etc/my.cnf --log

【옵션 설명】

All benchmarks takes the following options:

--comments

  Add a comment to the benchmark output.  Comments should contain

  extra information that 'uname -a' doesn't give and if the database was

  stared with some specific, non default, options.

--cmp=server[,server...]

  Run the test with limits from the given servers.  If you run all servers

  with the same --cmp, you will get a test that is comparable between

  the different sql servers.

--create-options=#

  Extra argument to all create statements.  If you for example want to

  create all MySQL tables as BDB tables use:

  --create-options=ENGINE=BDB

--database (Default test)

  In which database the test tables are created.

--debug

  This is a test specific option that is only used when debugging a test.

  Print out debugging information.

--dir (Default output)

  Option to 'run-all-tests' to where the test results should be stored.

--fast

  Allow the database to use non standard ANSI SQL commands to make the

  test go faster.

--fast-insert

  Use "insert into table_name values(...)" instead of

  "insert into table_name (....) values(...)"

  If the database supports it, some tests uses multiple value lists.

--field-count

  This is a test specific option that is only used when debugging a test.

  This usually means how many fields there should be in the test table.

--force

  This is a test specific option that is only used when debugging a test.

  Continue the test even if there is some error.

  Delete tables before creating new ones.

--groups (Default )

  This is a test specific option that is only used when debugging a test.

  This usually means how many different groups there should be in the test.

--lock-tables

  Allow the database to use table locking to get more speed.

--log

  Option to 'run-all-tests' to save the result to the '--dir' directory.

--loop-count (Default )

  This is a test specific option that is only used when debugging a test.

  This usually means how many times times each test loop is executed.

--help

  Shows this help

--host='host name' (Default localhost)

  Host name where the database server is located.

--machine="machine or os_name"

  The machine/os name that is added to the benchmark output filename.

  The default is the OS name + version.

--odbc

  Use the ODBC DBI driver to connect to the database.

--password='password'

  Password for the current user.

--socket='socket'

  If the database supports connecting through a Unix socket,

  then use this socket to connect

--regions

  This is a test specific option that is only used when debugging a test.

  This usually means how AND levels should be tested.

--old-headers

  Get the old benchmark headers from the old RUN- file.

--server='server name'  (Default mysql)

  Run the test on the given SQL server.

  Known servers names are: Access, Adabas, AdabasD, Empress, Oracle,

  Informix, DB2, mSQL, MS-SQL, MySQL, Pg, Solid and Sybase

--silent

  Don't print info about the server when starting test.

--skip-delete

  This is a test specific option that is only used when debugging a test.

  This will keep the test tables after the test is run.

--skip-test=test1[,test2,...]

  For run-all-programs;  Don't execute the named tests.

--small-test

  This runs some tests with smaller limits to get a faster test.

  Can be used if you just want to verify that the database works, but

  don't have time to run a full test.

--small-tables

  This runs some tests that generate big tables with fewer rows.

  This can be used with databases that can't handle that many rows

  because of pre-sized partitions.

--suffix (Default )

  The suffix that is added to the database name in the benchmark output

  filename.  This can be used to run the benchmark multiple times with

  different server options without overwritten old files.

  When using --fast the suffix is automaticly set to '_fast'.

--random

  Inform test suite that we are generate random inital values for sequence of

  test executions. It should be used for imitation of real conditions.

--threads=# (Default 5)

  Number of threads for multi-user benchmarks.

--tcpip

  Inform test suite that we are using TCP/IP to connect to the server. In

  this case we can       do many new connections in a row as we in this case may

  fill the TCP/IP stack

--time-limit (Default 600)

  How long a test loop is allowed to take, in seconds, before the end result

  is 'estimated'.

--use-old-results

  Option to 'run-all-tests' to use the old results from the  '--dir' directory

  instead of running the tests.

--user='user_name'

  User name to log into the SQL server.

--verbose

  This is a test specific option that is only used when debugging a test.

  Print more information about what is going on.

--optimization='some comments'

 Add coments about optimization of DBMS, which was done before the test.

--hw='some comments'

 Add coments about hardware used for this test.

--connect-options='some connect options'

  Add options, which uses at DBI connect.

  For example --connect-options=mysql_read_default_file=/etc/my.cnf.