Perso j'ai téléchargé le script de Mike Benoit, fs_bench.py, pour voir un peu de quoi il s'agit.
L'auteur t'explique très bien dans quel condition sont réalisés les tests, dans quel but et pourquoi :
"The reason I created this little script was because I was tired of searching high and low
for non-complete file system benchmarks. It seemed I could never find benchmarks that were
ever close to the "real-world" situation I was looking for. The benchmarks I could find
always seemed to be biased in some form, or ommiting "certain" information.
Like reiserfs v3 against ext3 with really small files. But never medium/large files. And
never all the main file systems pitted against one another on the same page.
One file system is never faster in absolutely every situation, so a multitude of tests
need to be run with many different options to get the full picture. Therefore the
purpose of this script is to make it really easy, and completely automated to benchmark
any number of file systems, or even file system options against one another with any
number of different benchmarks and benchmark settings, in hopes of seeing much more
complete benchmarks.
Yes, running this many tests takes days and days, but thats why this script
should automate it all for you, and at least you get enough data to see where the
strengths/weaknesses are of each different file system. Thats the idea anyways."
Donc, si tu te retrouves dans les mêmes cas d'utilisation que lui, je pense que tu dois pouvoir te fier à ses tests puisqu'ils sont assez proche d'une situation réelle.
[^] # Re: Benchmarks 'compréhensibles' des systèmes de fichiers sous Linux
Posté par Mickael_83 . En réponse à la dépêche Benchmarks 'compréhensibles' des systèmes de fichiers sous Linux. Évalué à 3.
L'auteur t'explique très bien dans quel condition sont réalisés les tests, dans quel but et pourquoi :
"The reason I created this little script was because I was tired of searching high and low
for non-complete file system benchmarks. It seemed I could never find benchmarks that were
ever close to the "real-world" situation I was looking for. The benchmarks I could find
always seemed to be biased in some form, or ommiting "certain" information.
Like reiserfs v3 against ext3 with really small files. But never medium/large files. And
never all the main file systems pitted against one another on the same page.
One file system is never faster in absolutely every situation, so a multitude of tests
need to be run with many different options to get the full picture. Therefore the
purpose of this script is to make it really easy, and completely automated to benchmark
any number of file systems, or even file system options against one another with any
number of different benchmarks and benchmark settings, in hopes of seeing much more
complete benchmarks.
Yes, running this many tests takes days and days, but thats why this script
should automate it all for you, and at least you get enough data to see where the
strengths/weaknesses are of each different file system. Thats the idea anyways."
Donc, si tu te retrouves dans les mêmes cas d'utilisation que lui, je pense que tu dois pouvoir te fier à ses tests puisqu'ils sont assez proche d'une situation réelle.