[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Benchmarking tmpfs for /tmp

Hello Michael,

First of all, I want to seize this opportunity to congratulate you about
Phoronix and all the work you do helping all the Linux community with
the nice benchmarks that you publish.

In the last days there is a growing flame on debian-devel mailing list
about if "tmpfs for /tmp" is a good or bad idea, and about if this is a
sane default.

Some people argue that it gives significant performance gain while
others argue that is not the case because of the linux filesystem cache.

Also another distributions (Fedora/OpenSUSE) are making the jump to have
this option (a tmpfs for /tmp) enabled by default without actual data
(benchmarks) to back their choice.

In order to have actual and real data about if having a tmpfs for /tmp
really gives any performance gain (and how much) I will like to ask you
about performing some benchmarks about this.

In my opinion the benchmark should test typical desktop and server
workloads on a system using tmpfs for /tmp (with a size of the 20% of
the total ram) and on the same system without using tmpfs for /tmp.

It will be helpful also to have the same tests benchmarked on a 32-bit
and also on a 64-bit system, as also on a system with a HDD and on
another with a SSD.

Best regards!

Thanks a lot!

Carlos Alberto Lopez Perez                           http://neutrino.es
Igalia - Free Software Engineering                http://www.igalia.com

Attachment: signature.asc
Description: OpenPGP digital signature

Reply to: