Lab 01 Setup and Introduction: IBM DB2 10.5 BLU Acceleration Proof of Technology Environment
Lab 01 Setup and Introduction: IBM DB2 10.5 BLU Acceleration Proof of Technology Environment
Lab 01 Setup and Introduction: IBM DB2 10.5 BLU Acceleration Proof of Technology Environment
Since we are running a VM from a laptop using a single hard drive, this proof of
technology is not a demonstration of the performance, but hands on exercises. So, please
be patient working through the lab exercises, as it may take a little longer for the commands to
run.
For example: When you type cd1 from anywhere, the directory will be changed to 01setup.
Linux Tips
If you are new to the Linux environment, the following tips will help.
As opposed to the Microsoft Windows, you have to click in a Linux GUI window to get its
focus, so that you can type the commands.
You can type the clear command from the command line window to clear the contents.
When you are typing in a command shell, after few characters, hit tab to auto complete
instead of typing the whole command. This will save you some typing.
To close a command line window, you can either type exit followed by Enter or press
CTRL-D.
The command that you need to type are bold faced such as:
$ ls –l
If the command shown starts with $, please assume that you are running that command
as db2psc user.
If the command shown starts with #, it is assumed that this command needs to run as
root.
It is assumed that you will hit Enter after typing the command.
__2. When the VM machine powers on, it does the automatic login of the db2psc user.
Note: Please note that cd1 is an alias pointing to the 01setup directory.
Explore Databases
__5. Run db2 list db directory to check databases that are already created in the VM.
$ db2 list db directory
__6. Please remember that the above 3 databases DB2DB, ROWDB and COLDB created in the Virtual
Machine will be used to show and compare DB2 BLU Acceleration throughout these labs.
__7. Please note that three databases ROWDB, COLDB and DB2DB are already created using scripts
given in the 01setup directory. Type ls -1 to list scripts.
$ ls –l
__8. The scripts shown in the following table have already been run. Please do not run any script
as these have already been run to save the time for creating databases.
setup03.sql Table definition for row organized uncompressed tables for DB2DB database
setup06.sql Table definition for row organized compressed tables for ROWDB database
setup09.sql Table definition for column organized tables for COLDB database
setup20 Create Optim Query Tuner objects in all 3 databases so that we could use
__10. Intel® Processor Identification Utility (download from www.intel.com) can be used to determine
the capabilities of the CPU. We ran this utility for this VMware machine and the output is as
shown.
__11. Please note that the Intel® Streaming SIMD Extensions (Intel® SSE) support is indicated in the
above output.
Note: You can download the Intel processor Identification Utility from
http://www.intel.com/p/en_US/support/highlights/processors/toolspiu and
check SSE support in your environment.
__14. Please notice that different variant of Intel® Streaming SIMD Extensions (Intel® SSE) are
mentioned in the flags field which confirms that the Intel CPU supports SIMD instructions.
__16. There are several examples given and we will run 2 of them to see SIMD processing cycles.
__20. From the above output, please notice that the C/C++ program generates an image file using
SIMD processing and serial processing (without using SIMD). Please note that the program uses
33 million cycles by using SIMD processing compared to 246 million cycles without using SIMD
processing. Both program generates an identical image file.
__21. Please also note by using SIMD processing, it uses 7.33 times less CPU cycles to generate the
same image.
__29. Please note that the SIMD processing using 1 core of CPU takes 12661 million cycles to
generate the image. It takes 3976 million cycles using multi-core SIMD processing. Without
using SIMD processing, it takes 72641 million cycles to generate the same image.
__30. Please also note that the single core SIMD processing speeds up by 5.74 times and multi-core
SIMD processing speeds up by 18.27 times.
__31. Run eog volume-serial.ppm to check the image which was generated by the program.
# eog volume-serial.ppm
__34. After running the above sample programs, we have seen the benefit of SIMD processing.
__36. DB2 10.5 BLU Acceleration uses encoding algorithm to compress columnar data and column
compression is turned on by default when using column organized tables.
__37. After running the second example, we noticed that SIMD processing achieves far better results
when using multiple CPU cores.
__38. DB2 10.5 BLU Acceleration benefits from more number of cores and more memory.
__39. Keep GNOME Terminal open as we will need it for the next lab.