November 24, 2014

tpcc-mysql: Simple usage steps and how to build graphs with gnuplot

Lots of times we could see different benchmarks performed by tpcc-mysql. So today I want to tell you about how to use tpcc-mysql and how to build graphs with gnuplot in a few easy steps.

As an example I’ll compare Percona Server 5.5 (latest version: 5.5.31) performance by changing InnoDB buffer pool size: innodb_buffer_pool_size = 256M / innodb_buffer_pool_size = 768M on my old test machine

System Info

  • CPU: Intel(R) Pentium(R) 4 CPU 1.80GHz
  • MemTotal: 1543732 kB
  • OS: LinuxMint 15 (based on Ubuntu 13.04)

You can find the source code of all files at the end of this post


sudo apt-get install bzr
bzr branch lp:~percona-dev/perconatools/tpcc-mysql
make all

In this case it’s installed to ~/tpcc-mysql/ directory

  • Install gnuplot

sudo apt-get install gnuplot

DB Config
First test will be running with innodb_buffer_pool_size = 256M option enabled and second one with innodb_buffer_pool_size = 768M

Test for innodb_buffer_pool_size = 256M

Create DB
Assuming that Percona Server 5.5.31 installed and configured

cd ~/tpcc-mysql
mysql -u root -p -e "CREATE DATABASE tpcc1000;"
mysql -u root -p tpcc1000 < create_table.sql
mysql -u root -p tpcc1000 < add_fkey_idx.sql

Load Data

./tpcc_load tpcc1000 root "root-password" 20


  • Host:
  • DB: tpcc1000
  • User: root
  • Password: root-password
  • Warehouse: 20


In this case DB size is 1.9GB

Run tpcc-mysql test

./tpcc_start -h127.0.0.1 -dtpcc1000 -uroot -p -w20 -c16 -r10 -l1200 > ~/tpcc-output-ps-55-bpool-256.log


  • Host:
  • DB: tpcc1000
  • User: root
  • Warehouse: 20
  • Connection: 16
  • Rampup time: 10 (sec)
  • Measure: 1200 (sec)

The most interesting part in the output is:


10, 25(17):9.005|9.221, 21(0):1.866|1.869, 3(0):0.647|0.840, 1(0):0.000|10.614, 2(2):19.999|29.490
20, 22(14):9.419|9.555, 26(0):1.591|1.593, 2(0):0.593|0.788, 4(0):10.453|10.688, 3(3):19.999|22.962
30, 41(32):8.703|9.057, 32(0):1.615|1.662, 3(0):0.588|0.777, 2(0):9.530|10.495, 3(2):19.999|22.983

The first two values are “time range” and “transactions”, so you can read it as:

0-10 sec, 25 transactions
10-20 sec, 22 transactions
20-30 sec, 41 transactions

Test for innodb_buffer_pool_size = 768M

Repeat following steps for innodb_buffer_pool_size = 768M (change it in my.cnf) and get results:

  • DB Config
  • Create DB
  • Load Data
  • Run tpcc-mysql test

./tpcc_start -h127.0.0.1 -dtpcc1000 -uroot -p -w20 -c16 -r10 -l1200 > ~/tpcc-output-ps-55-bpool-768.log

There are 2 files: tpcc-output-ps-55-bpool-256.log and tpcc-output-ps-55-bpool-768.log which have benchmarking results for both tests.

Generate data file for each test

./ ~/tpcc-output-ps-55-bpool-256.log > tpcc-256-data.txt
./ ~/tpcc-output-ps-55-bpool-768.log > tpcc-768-data.txt

Merge data files

paste tpcc-256-data.txt tpcc-768-data.txt > tpcc-graph-data.txt

Build graph

./ tpcc-graph-data.txt tpcc-graph.jpg

In this case tpcc-graph-data.txt is a filename of source datafile and tpcc-graph.jpg filename of graph which will be generated

Graph ready: tpcc-graph.jpg

Note: “using 3:4 … with lines axes x1y1″ in means that columns number 3 and 4 in datafile will be used for as axises x and y accordingly while building second line


File listing (I got it there and a bit modified)


if [ -n “$2″ ]
echo “Defined $2″

cat $1 | grep -v HY000 | grep -v payment | grep -v neword | awk -v timeslot=$TIMESLOT ‘ BEGIN { FS=”[,():]”; s=0; cntr=0; aggr=0 } /MEASURING START/ { s=1} /STOPPING THREADS/ {s=0} /0/ { if (s==1) { cntr++; aggr+=$2; } if ( cntr==timeslot ) { printf (“%d %3dn”,$1,(aggr/’$TIMESLOT’)) ; cntr=0; aggr=0 } } ‘


### goto user homedir and remove previous file
rm -f ‘$2′

gnuplot << EOP

### set data source file
datafile = ‘$1′

### set graph type and size
set terminal jpeg size 640,480

### set titles
set grid x y
set xlabel “Time (sec)”
set ylabel “Transactions”

### set output filename
set output ‘$2′

### build graph
# plot datafile with lines
plot datafile title “PS 5.5.1, buffer pool: 256M” with lines,
datafile using 3:4 title “PS 5.5.1, buffer pool: 768M” with lines axes x1y1


About Michael Rikmas

Michael joined Percona in October 2007. He serves in several roles, including Persona's 24x7 support coverage. He has an undergraduate degree in computer science, and in 2010 he started pursuing studies to earn an MBA.


  1. Great article Michael. Thank you for sharing.

  2. wula says:

    There is a little problem in analyze file.

    printf (“%d =
    ”,$1,(aggr/’$TIMESLOT’)) ;

    should be changed to:

    printf (“%d =
    ”,$0,(aggr/’$TIMESLOT’)) ;

  3. Michael Rikmas says:

    In case of awk with field separator (FS) $1 means a first field of input record so this is exactly what I was looking for and it works just like needed. Is that not a true?

  4. wula says:

    Oh, you were right, yesterday I had a careless mistake in the test. :(

  5. Michael Rikmas says:

    Sure, not problem ;)
    Thanks for your interest!

  6. Will tpcc-mysql be available in your Yum Repository sometime soon?

  7. Michael Rikmas says:

    Stefan, I’m not sure about these kind of plans, but I’ll check that with our team.

Speak Your Mind