osdldbt/dbt3

Database Test 3: Fair Use TPC Benchmark(TM) H kit

dbt3's Issues

How to build TPC-H Tools for MySQL

Hello, I met some problems while building TPC-H Tools for MySQL. I refered to the INSTALL.rst and user's guide. The user's guide says dbt3 supports MySQL but it didn't work when I want to use dbt3-build-dbgen.

The command line is following:
./dbt3-build-dbgen --patch-dir=/home/czhang/mysql-dbt3/share/dbt3/patches --query-dir=/home/czhang/mysql-dbt3/share/dbt3/queries mysql /home/czhang/mysql-tpch

But the ERROR is reported: ERROR: Unhandled DBMS: mysql

dbt3 doesn't seem to natively support MySQL, so I have two questions I'd like to get help:

1. How do I modify the dbt3's scripts to make it work with MySQL? Is there a more detailed user's guide?
2. Where should the extracted TPC-H-tool folder be placed? Does it have to be somewhere in the dbt3 folder? Or is it anywhere?

Thank you for your help very much!

Result statistics broken in new version of dbt3-post-process

Hi,

Thanks for maintaining dbt3!

I've been using it to benchmark PostgreSQL over the past few weeks and I just noticed that the generated statistics in summary.rst don't make sense. As an example:

========  ========  ========  ========  ========  ========  ========  ========
 Stream      Q1        Q2        Q3        Q4        Q5        Q6        Q7
========  ========  ========  ========  ========  ========  ========  ========
       1      1.73      0.26      0.41      0.13      0.79      0.35      1.00
       2      1.74      0.25      0.40      0.14      0.79      0.36      0.97
     Min      1.74      0.25      0.40      0.14      0.79      0.36      0.97
     Max      1.74      0.25      0.40      0.14      0.79      0.36      0.97
     Avg      0.87      0.12      0.20      0.07      0.39      0.18      0.48
========  ========  ========  ========  ========  ========  ========  ========

As you can see, both Min and Max are simply the value of the last stream, whilst Avg is the last stream divided by the number of streams. As an example, Q1 here is reported to have an average of 0.87 instead of the correct 1.74. (The same type of errors can be seen in the example in doc/dbt3.rst.)

If we take a look at what the script does, the error is evident:

dbt3/scripts/dbt3-post-process.in

Lines 199 to 218 in a668b1d

	printf " Avg"
	for Q in $(seq "${Q0}" "${Q1}"); do
	if [ "${Q}" -eq 23 ]; then
	STRING="^PERF\.THRUPUT\.RFST.*\.RF1,"
	elif [ "${Q}" -eq 24 ]; then
	STRING="^PERF\.THRUPUT\.RFST.*\.RF2,"
	else
	STRING="^PERF\.THRUPUT\.QS${STREAM}\.Q${Q},"
	fi
	if [ $HTML -eq 1 ]; then
	SIZE=80
	else
	SIZE=8
	fi
	SUM=$(grep "${STRING}" "${QTIME}" \
	| cut -d "," -f 5 | xargs | tr ' ' + | bc -l)
	printf " %${SIZE}.2f" "$(echo "${SUM} / ${STREAMS}" | bc -l)"
	done

STREAM is set in a for loop further up in the script, but this section is not contained within it. Thus, STREAM just retains its latest value (in this case 2) and the basis of the calculation becomes just the final stream instead of all streams.

Looking at the history for scripts/dbt3-post-process.in, I believe this error was introduced when the post-processing was converted from R to shell in a1719d7.

I think a solution would be to not use the STREAM variable in the STRING regex for the aggregations, e.g.

STRING="^PERF\.THRUPUT\.QS[[:digit:]]\+\.Q${Q},"

Happy to make a PR if you agree!

All the best,
Ole