Introduction

This is a modified version based on the TPC-E version from Percona https://www.percona.com/blog/2010/02/08/introducing-tpce-like-workload-for-mysql/. The package is based on the TPC-E workload described on http://www.tpc.org/tpce/default.asp and uses EGen software provided by TPC.

In this package SAP HANA Support was added to the TPC-E Percona version.

Installation

Before you can build the tool you need to install following dependencies:

unixodbc driver (>= 2.3.4)
ODBC driver for HANA (Linux) (Tested with 64 bit version of driver)
LLVM
CLANG

Install unixodbc from Source (>= 2.3.4)

You have to install unixodbc from source instead of using package manager installed on your system (e.g. homebrew), since the package homebrew is providing is outdated and missing some crucial fixes introduced in 2.3.3.

Get source from http://www.unixodbc.org/ and untar it into an directory
Run ./configure on your system
Run make
Run make install (sudo maybe required here to copy libraries into system path)

Compile TPC-E

Adjust the include path in the makefile in the directory prj to the directory with the unixodbc library / include files (if not located in the usual system directories) on your system
Execute the makefile by calling make

Configure unixodbc

To be able to connect via ODBC to SAP HANA you have to configure unixodbc. The default location of the odbcinst.ini and odbc.ini configuration files are the directory /etc. Add following lines to odbcinst.ini:

[HDBODBC]
Description="HDBODBC Driver"
Driver=<path to HANA ODBC>/libodbcHDB.so
Setup=<path to HANA ODBC>/libodbcHDB.so
Threading = 2

Add following lines to odbc.ini:

[hana]
Description = "HANA"
Driver = HDBODBC
servernode = <IP of HANA Instance>:3<Instance Number>15

Usage

Generate Test Data

First you have to generate Test Data using EGenLoader in the bin directory supplying the path to the flat_in and flat_out directory. You can adjust the generator by supplying additional arguments to control the size of the generated dataset.

./EGenLoader -i flat_in -o flat_out -c <number_of_customers> -t <active_customers> -f <number_of_customers_for_1TRTPS> -w <days_of_trade>

Concrete Example:

./EGenLoader -i flat_in -o flat_out -c 2000 -t 2000 -f 200 -w 50

After successfull generation of the benchmark data the generated data is located in the flat_out directory.

Prepare Database for Loading of the Flat Files

First you have to create the tbales for the TPC-E like benchmark by executing the SQL script 1_create_table.sql located in the directory scripts/<your_database>.

Load Flatfiles into SAP HANA with hdbsql

Before you execute can import the generated flat files you have to generate the SQL import script in the directory scripts/<your_database> (only SAP HANA). To generate the load script execute the python file python 2_load_data.py <path_to_flat_files> <schema_name> supplying the path to the direcotry containing the flat files and the schema name.

We use the hdbsql command line tool supplied with the HANA ODBC driver to load the generated data into the SAP HANA database. We load the flatfiles into the SAP HANA database by executing following command:

<path_to_hdbsql_executable>/hdbsql -n <IP_HANA_INSTANCE>:3<INSTANCE_ID>15 -u <USERNAME> -p <PASSWORD> -I <PATH_TO_IMPORT_SCRIPT>/hana_csv_import.sql -c ';'

Please confirm that the User used for the import has the user rights to IMPORT / INSERT data into the SCHEMA/TABLE!

Create FK / Indicies / ... in Database

After successfully importing the previously generated flat_files you have to execute the remaining SQL scripts (3-5) in ascending order.

The last required SQL Script has to be generated executing the python file python 6_create_sequence.py <path_to_Trade.txt> supplying the path to the direcotry containing Trade.txt that has been generated by ./EGenLoader as parameter.

The SQL script 6_create_sequence.sql has to be executed with the user account which is used for the TPC-E benchmark run!

Execute TPC-E like Benchmark

To execute the Benchmark run the following command:

./EGenSimpleTest -c <number_of_customers> -a <active_customers> -f <number_of_customers_for_1TRTPS> -d <days_of_trade> -l <number_of_customer_load_unit> -e <path_to>/flat_in -D <name_of_data_source> -U <username> -P <password> -t <duration> -r <ramp_up> -u <number_of_users>

Example

./EGenSimpleTest -c 2000 -a 2000 -f 200 -d 50 -l 200 -e ../flat_in -D hana -U <user> -P <password> -t 30 -r 10 -u 2

Usage Example for Percona TPC-E like benchmark: https://www.percona.com/blog/2010/02/09/introducing-percona-patches-for-5-1/

The values <number_of_customers>,<active_customers>,<number_of_customers_for_1TRTPS>,<days_of_trade> have to be the same as used at the ./EGenLoader data generator !*

Known Issues

This project can be only compiled with CLANG / LLVM due to the usage of some deprecated C functions GCC seems to have an issue with.

Further Improvements

Adjust the Benchmark using the TPC-E Changelog from 1.10 to 1.14 (Currently 1.9) and supplied Source
Fix Compile Error when using GCC

Miscellaneous

Debugging

Compile the program with the argument -DDEBUG and execute the benchmark with the argument -o <directory_for_error_log> for additional debugging output.

The compiler flag -DPRINT_DEADLOCK can also help to identify locking conflicts in the program when several transactions fail.

Running using not Prepared Statements

When compiling the TPC-E benchmark without -DUSE_PREPARE set the ramp-up period to at least 60 seconds -r 60 to get measurement results.

License

This package based on tpc-e workload described on http://www.tpc.org/tpce/default.asp and uses EGen software provided by TPC.

The package is fully compatible with TPC license http://www.tpc.org/tpce/egen/TPC-E%20License%20Agreement.pdf

The results you get with this package can't be named "TPC Benchmark Results" and are not compatible or comparable with TPC-E Benchmark results

VincentS/tpcemysql