Would mounting a file system 'tree' to /mnt create a bottleneck?
 

A little embarassed to even ask this, but...

In the last few evenings, a typical Linux session has been creating a root file system on an SD card, mounted on /mnt/sd2 (partition2 of card). At the same time watching a movie from the HD, mounted at /mnt/HD.

I was wondering if everything going through /mnt might be slowing things down at all. The answer should be 'no' since everything is mounted to 'root' and that doesn't seem to be a problem, but maybe /mnt is handled differently than root.

Just ran an experiment with 'top' and seeing what is happening with mplayer and debian bootrap:

Can't really tell much from this, except that 'mmcqd/0' must be the bootstrap program unpacking things in the sd card. It kind of begs dumb question number 2: the unpacking process is crawling down the terminal screen, why is it not using more cpu to speed itself up? Must be the sd card can only accept writing to it at that speed, average about 1 second per package? Can't pull the card out at the moment, but pretty shure it's class 10, although suspiciously cheap from Eb.

AFAIK, /mnt is a mount point, just like /, /var, /home, whatever else. It should not affect speed drastically.

What do you mean by crawling down the terminal screen? Text is slow to print? What terminal? I found that certain terminals are just slower than others. An example: hen I was working on a PSP game project a while back if I used GNOME Terminal, printing debugging messages would slow the terminal and PSP to a crawl; using xterm would be smooth as silk on both.

Hi goumba, apologies, by terminal I meant a remote ssh connection, I didn't make that clear.

Well mounting two block devices in a separate subdirectory of /mnt doesn't slows it down.

Speed of SD card class 10 is usually 10MB/sec read or write. It's not that fast to slow down system of nowadays.

Compare with SATA Hard drive transfering at 100-200MB/sec.

Printing in a terminal actually slow downs whatever be the process. You can try a large file unpacking and observe time with terminal output and without.

Ah. Well, that's limited by other factors as well. Bandwidth, maybe a limitation on the sending end. I don't know your setup.

As veerain stated, printing output slows down the process as well. It has to send the text to the console, and in this case over ssh as well. Don't know what you're using, but if you don't need the output then disable it. Tools like gnu tar are actually silent except for an error, or unless you tell it to be verbose.

If you need the output, maybe piping to a file then check after the process is done will speed it up (not sure to the extent, or if that will even make a difference, never tried).

Hi goumba, I'm just going to accept the following:

Making several sub-mounts on /mnt is not really a problem.
Until someone figures out how to unpack things in parallel, a 1 Ghz processor is going to be slow, even if it has 2 ARMs (my first Linux joke!:)). Even slower if you are watching it remotely.

Another interesting thing has cropped up today, but will leave that for another question. Rhetorically speaking though, why in the heck are there 3 places to set environment variables for booting: default settings in kernel, boot.scr, and uEnv.txt. If it were windows I'd have given up by now, but since it's Linux, I'm sure there's an elegant complexity behind it all, just to make a simple task more interesting: 'this' is my monitor, my root partition is 'there', post your progress 'over here'. At the moment it feels like trying to coordinate three drunks. But that's a question for another post. Just a rant here.