SalixOS vs. Debian performance

[GJones]

Today, for reasons mostly involving security, I replaced SalixOS with Debian Squeeze on my ex-laptop.

This is an extremely minimal Debian setup; because I'm not using the computer for normal desktop stuff, I don't even have DBUS and HAL running. So it should be fast. But even so, I'm absolutely shocked by the difference in performance.

On Salix for instance, I can count three seconds before Geany starts. On Debian, on the same computer, it starts immediately (in < 1 second). Abiword also starts immediately; and starting Iceweasel after a cold boot takes 4 seconds, as opposed to 8 on Salix.

I'm kind of stumped! As far as I can tell, the only major differences are in hdparm settings and kernel config. On Salix IO_support defaults to 1 (32 bit), on Debian it defaults to 0 (16 bit) and cannot be set to anything else... And Salix uses a 1000 Hz kernel, while Debian uses a 250 Hz one. That seems to be it.

Could either of the things I mentioned be responsible? Or could it be something else? Any ideas?

[Shador]

Are we talking about first time launches here? Because a second, third, ... launch is always much faster due to memory cache and you need to be careful when comparing the start times. I.e. you shouldn't mix first-time and later launches.

Maybe some sort of prefetching is running on Debian? That means some application/software module that possibly analyzes your behaviour or loads data into the cache by some other rules. That would mean no slow disk accesses are needed anymore once you launch the application as the data has been read into memory already earlier.

As far as I can tell, the only major differences are in hdparm settings

It should be easy to change that one. But I doubt it matters.

kernel config

You could try to boot Salix with the debian kernel, if you put the modules in the expected location on the Salix fs.

[GJones]

Are we talking about first time launches here? Because a second, third, ... launch is always much faster due to memory cache and you need to be careful when comparing the start times. I.e. you shouldn't mix first-time and later launches.

Yes, I was referring to launch times following a cold boot. Wall-clock-based and not exactly scientific, but better than just "this feels faster/slower."

Maybe some sort of prefetching is running on Debian? That means some application/software module that possibly analyzes your behaviour or loads data into the cache by some other rules. That would mean no slow disk accesses are needed anymore once you launch the application as the data has been read into memory already earlier.

Shouldn't be, on a minimal system like that. I definitely did not install prefetch, readahead, etc., and Debian Stable wouldn't be using experimental adaptive readahead kernel patches.

You could try to boot Salix with the debian kernel, if you put the modules in the expected location on the Salix fs.

I think that'll be the next step. Will report back with the results.

[GJones]

Yep, it's the kernel. Applications launch faster after a cold boot with Debian kernel 2.6.32-5-amd64. Still no particular idea why though.

I think I'll try some other binary kernels and see which perform the best...

[deja69]

There is Shorewall in the Salix repositories.

Plus has anyone tried the Liquorix kernel with Salix?
If the jigsaw pieces fit,it just takes a bit of shuffling

Edit PS..oops GJones ,just noticed your Liquorix posts from last year

[GJones]

Firewalls can only do so much when you have ports open to the internet. I *believe* (could be wrong) that Debian is a bit more stringent than Salix/Slackware about security patches, being a more server-oriented distro.

Liquorix kernel isn't worth it IMO. Seems to perform worse than the stock Debian kernel. Maybe a lower kernel tick rate is better for low spec machines.

OTOH, I realized I'd forgotten to enable zram on the ex-laptop. That usually makes a perceptible difference.