Supporting Hardware - Expert Consolidation in Oracle Database 12c - page 100

db_files = 300
processes = 1000
sga_target=10G
filesystemio_options=setall
Note that the above initialization file will only be used to create the database. The initialization file for the actual
benchmark test runs will be different and is found in the top-level SLOB directory named simple.ora. Nevertheless,
once the database is created, keep a note of the control_files parameter to ensure you do not have to dig around
looking for them later.
Now create the database by executing cr_db.sql . A few minutes later, a minimalistic database is ready. It is not in
archivelog mode, which could be important for your testing.
In the next step the wait kit needs to be compiled. This is not normally a problem with Linux-based systems
but can be a bit of a challenge on Solaris and AIX. If you followed the pre-requisites for the database installation on
Linux, you should already have a compiler, linker and interpreter for Makefiles. On other platforms you should have
everything except for the C-compiler “cc.” Regardless of platform, once you have the compiler suite in place, you use
the make command in ~/SLOB/wait_kit to compile the code, as shown here on a Linux system:
[oracle@server1 SLOB]$ cd wait_kit/
[oracle@server1 wait_kit]$ make
rm -fr *.o mywait trigger create_sem
cc -c -o mywait.o mywait.c
cc -o mywait mywait.o
cc -c -o trigger.o trigger.c
cc -o trigger trigger.o
cc -c -o create_sem.o create_sem.c
cc -o create_sem create_sem.o
cp mywait trigger create_sem ../
rm -fr *.o
[oracle@server1 wait_kit]$
The empty database has to be populated with test data next. SLOB works by creating a number of schemas
(users) on the tablespace created as part of the cr_db.sql script named IOPS: 128 by default. A new feature of SLOB 2
allows you to scale the data volume per user. Unless defined otherwise, the setup script used to create the test data will
create just a single table named CF1 with a unique index in each schema. The table structure is especially prepared for
the benchmark so that a block contains a single row only. By default, 10,000 rows are created based on a seed table but
in completely random order so that no two users have the same dataset.
If you would like to scale the dataset per user, you need to modify the SLOB configuration script slob.conf
and modify the SCALE variable. The default of 10,000 rows equates to approximately 80 MB per user on the lab test
system:
SQL> select sum(bytes/power(1024,2)) mb,segment_type
2 from dba_segments
3 where owner = 'USER1'
4 group by segment_type
5 /
MB SEGMENT_TYPE
---------- ------------------
.1875 INDEX
80 TABLE