Congratz on a FAST Distro (Speed comparison inside)

[Artthou]

Hello all,

I am a recent Vector dabbler, and just wanted to leave the results of a comparison I did just so the Devs can know how well their goals are being attained for us average joe tinkerers.

I am dual booting an older laptop (Pentium III with 256MB RAM) with PCLinuxOS and Vector 5.8 Standard, and I did a load time comparison for Firefox between the two.

So on the same machine, same (roughly) version of Firefox: (drumroll please...)

PCLOS:  12 seconds from click to browse
Vector:  5 seconds from click to browse

I like PCLOS and appreciate the work the team is doing.  They receive kudos for how fast their distro is and justly so.  And even though the KDE vs xcfe might account for much of the difference, I am just ecstatic to have an OS that can give me 7 seconds of my life back everytime I want to open a browser (not to mention all the other speed enhancements).

Congratz again on a job well done, and thank you for your hard work.

Satisfied Vectorite

[tomh38]

Artthou,

As a fellow tinkerer, I want to say welcome to Vector!.  It is amazing what the devs here have done with this distribution.  I've installed VL on a lot of machines (older hardware and new), and I've never found a Linux that's faster.  I've also found that KDE runs much faster than on any other distro I've tried.  Now I'm just waiting for KDE 4 to come out, since it's supposed to be faster than 3.x KDE, and see what the VL team does with that!

By the way, you may not know that Vector is developed by a fairly small number of people, especially compared to, for example, Ubuntu or Fedora.  To me that makes it all the more amazing what they've done.

Tom

[lagagnon]

...  And even though the KDE vs xcfe might account for much of the difference

Yes, undoubtedly. Especially as you only have 256MB of RAM with PCLOS and KDE I am absolutely positive you will have had to use swap memory to get Firefox up and running and hence the difference in boot time. Still, VL is yes, a fast distro but a more valid comparison would be needed before getting too excited 

[Artthou]

Especially as you only have 256MB of RAM with PCLOS and KDE I am absolutely positive you will have had to use swap memory to get Firefox up and running

This is likely the case, but discouraging from a memory usage perspective in KDE since I had no other Applications running when I launched Firefox.  With 256 RAM I am already swapping memory the first web browser I open?  If so, xfce is definetly where this laptop needs to be.

A more valid comparison would be SOHO, (or I could try installing KDE on Standard), but I would wait till the Live CD version was more reliable.  Or, I could try the xfce desktop (SAM Linux) on the PCLOS partition!

All in good fun

[lagagnon]

With 256 RAM I am already swapping memory the first web browser I open?

I've done this test myself with other boxes running KDE with 256MB RAM. Boot the system and open a terminal and type "free" and check out the "used" column before you run any apps. Then run Firefox and run "free" and you will probably see that you have used some swap.

[Artthou]

A little more info on Memory usage in Vector compared to PCLOS...

I installed XFCE on my PCLOS partition and learned a bit about configuring the panels enough to get it to look like Vector 

When I checked memory usage just after bootup (no apps running), I noticed I was around 210MB, compared to about 250MB when booted to KDE on the same partition.

Vector on the otherhand was at around 172MB after booting, so I know I am getting back about 40MB if I use vector instead. 

I am still having a few configuration problems in vector (zd1211rw not grabbing my usb wireless (a zd1211b device) that I don't have in a problem with in PCLOS (which is odd since I am using a newer kernel in my vector install), but the performance potential is encouraging me to stick with it and get vector figured out.

[lagagnon]

Vector on the otherhand was at around 172MB after booting, so I know I am getting back about 40MB if I use vector instead. 

The next step to see what is going on is to run "ps aux" on both systems to show you why the difference. That command will show you every process running on the system and its CPU/memory footprint. I suspect PCLOS will have a whole mess of extra daemons running that VL does not. VL attempts to run only those necessary to give you a fast capable workstation. You can even reduce the VL memory footprint even more with a few little tricks!

[Joe1962]

You are stating full memory used, not the real amount used by OS and applications. The full amount includes buffers and disk cache. For example, here in a 160 MB virtual machine, VL 5.8 boots to XFCE with: used=52MB, buffers=6MB and cache=89MB, for a total of 152MB (I know, the sum gives 148MB, but that's the total given in htop and free, for some reason). The used amount (or maybe used + buffers) is what you need to compare, since disk cache is variable. As I got these figures in htop, this means I already had to run a console + htop inside XFCE. BTW, this is practically a vanilla install, and sshd was running.

[Artthou]

So does disk cache shrink then as more apps are openned and the RAM is needed?  Is the size of the disk cache variable in that it expands as more disk reads happen as long as RAM is available?

I wonder how the availability of RAM for disk cache impacts end user performance appearance.  Could be a good consideration if one finds that while no virtual memory swapping is happening, there is no disk cache space available and so performance is less than it otherwise would be.

Thanks for the input!  I will check out the numbers with this in mind and report back.

[Triarius Fidelis]

So does disk cache shrink then as more apps are openned and the RAM is needed?  Is the size of the disk cache variable in that it expands as more disk reads happen as long as RAM is available?

I wonder how the availability of RAM for disk cache impacts end user performance appearance.  Could be a good consideration if one finds that while no virtual memory swapping is happening, there is no disk cache space available and so performance is less than it otherwise would be.

Thanks for the input!  I will check out the numbers with this in mind and report back.

I don't know about RAM cache, but you can tweak 'swappiness' (tendency of kernel to use swap) with /proc/sys/vm/swappiness

http://lwn.net/Articles/83588/

[Joe1962]

Also, I have often read that Linux is, by default, better than Windows at using available RAM. No sense keeping it idle, eh?

[uelsk8s]

I am still having a few configuration problems in vector (zd1211rw not grabbing my usb wireless (a zd1211b device) that I don't have in a problem with in PCLOS (which is odd since I am using a newer kernel in my vector install), but the performance potential is encouraging me to stick with it and get vector figured out.

you may need to blacklist the zd1211rw module so you can use the zd1211b module both should be present

[Artthou]

you may need to blacklist the zd1211rw module so you can use the zd1211b module both should be present

Ohh nice!  I didn't even bother to look for zd1211b as almost everyone is just depending on the kernel based zd1211rw.  I will try it when I get back to the laptop tonight.  I am assuming it was built against the 2.6.18 kernel that shipped with Standard?

Thanks for the heads up.

[Artthou]

I found my problem.  The zd1211 directory under firmware did not have the zd1211b firmware files, so zd1211rw was not finding the firmware.  I copied them over from my PCLOS partition and I am using it now successfully, but still may go back to the zd1211b driver since it has better features for coming up automatically.

Anyway, thanks for the help, and perhaps the zd1211b firmware files should be added to your package or base install.

[metvas]

Hello:
I certainly hope that those Digg-It and distro watch folks who slammed this forum are reading this.
What a team...Very good work.
regards
Darrell