helpful: useful little script for benchmarking...

tommytomthms5 · 07-27-2008, 03:07 PM

this is my benchmarking script it runs completely clean as in it removes all temporary files after they are used.

Code:

#!/bin/ksh
IFS=$'\n'
echo "benchmaring..."
start=$SECONDS
for i in $(ls ./);
do a=0;
while [ "$a" -ne "1000" ];
do echo "$a " > .testfile;
let a++;
done;
stop=$SECONDS
((runtime=stop-start))
rm .testfile
echo "file with 1k lines created and destroyed in $runtime seconds" > .time
done;
cat .time
rm .time
echo "wow that was fast!"

just copy and paste into benchmark.sh then run

sh benchmark.sh

now i know ive posted stupid scripts before but if you read through it its completely clean and removes all the temporary files it makes... on top of that i used a realistic way of creating a process then ending it... so no jokes no pranks no kidding just a useful little script

if you feel like posting your outputs please do!

XavierP · 07-27-2008, 03:18 PM

I get this:

Code:

ray@spinelli:~$ sh benchmarking.sh
benchmarking...
benchmarking.sh: line 14: syntax error near unexpected token `newline'
benchmarking.sh: line 14: `echo "file with 1k lines created and destroyed in $runtime seconds" > '

Running Debian Lenny

Mr. C. · 07-27-2008, 03:18 PM

Benchmarks are designed to test the performance of something. What is the something here?

tommytomthms5 · 07-27-2008, 03:28 PM

@ Mr. C >> it was supposed test how fast you can make and destroy a file that has 1000 lines yet i found an error in it... it just rewrites one line 1000 times then deletes it... messed up on the "echo %a > .testfile" line i should first create the file then do "echo %a >> .testfile"

XavierP · 07-27-2008, 03:28 PM

For some reason .time was being interpreted as being on it's own line. Slightly revised code is:

Code:

#!/bin/bash
	IFS=$'\n'
	echo "benchmarking..."
	start=$SECONDS
	for i in $(ls ./);
	do a=0;
		while [ "$a" -ne "1000" ];
	do echo "$a " > .testfile;
		let a++;
	done;
		stop=$SECONDS
		((runtime=stop-start))
		rm .testfile
		echo "file with 1k lines created and destroyed in $runtime seconds" > .time
	done;
	cat .time
	rm .time
	echo "wow that was fast!"

I changed to bash from ksh, no reason why it couldn't be changed back.
The response on my laptop is

Code:

benchmarking...
file with 1k lines created and destroyed in 2 seconds
wow that was fast!

I don't know if that is fast as I have nothing to compare it with.....?

colucix · 07-27-2008, 03:29 PM

I got:

Code:

benchmaring...
file with 1k lines created and destroyed in 1 seconds
wow that was fast!

But fast what? I tested in a directory with 2 dummy subdirectories. If I had tested from a directory containing 50000 subdirectories, the response would have been "wow that was fast!"...

Anyway, as Mr.C pointed out it is not really clear what is the aim of this "benchmarking" tool. If it is disk I/O... have you ever heard about the good ol' bonnie or Iozone filesystem benchmark?

Also a good code would at least use indentation!

Is my critique too destroying?

tommytomthms5 · 07-27-2008, 03:34 PM

well... i see all your points and yea i guess you do sorta need a comparison for this to work right....

oh and as said b4 i found that i messed up it doesnt write 1000 separate lines it rewrites one line 1000 times

Mr. C. · 07-27-2008, 03:38 PM

Quote:

Originally Posted by tommytomthms5

@ Mr. C >> it was supposed test how fast you can make and destroy a file that has 1000 lines yet i found an error in it... it just rewrites one line 1000 times then deletes it... messed up on the "echo %a > .testfile" line i should first create the file then do "echo %a >> .testfile"

Colucix understood my point. What your test actually benchmarks is how fast the shell can run a loop. This is very different from how fast the system can append a line to the end of a file.

Benchmarks require very careful analysis and construction of the proper procedures to ensure you are actually testing what you think you are testing. The shell is an interpreter, and is a fairly slow one. Because of that, your benchmark is limited by the speed of the slow shell. File I/O is very fast, and its impact is insignificant relative to the shell loop speed - so it is hidden in the noise.

tommytomthms5 · 07-27-2008, 03:44 PM

tell me more of this "file i/o"

Mr. C. · 07-27-2008, 03:46 PM

Download, build, and run bonnie or bonnie++ and see how fast your I/O really occurs on your system. That will show you the speed difference, and start you off towards understanding File input and output (I/O).