Boston Linux & UNIX was originally founded in 1994 as part of The Boston Computer Society. We meet on the third Wednesday of each month at the Massachusetts Institute of Technology, in Building E51.

BLU Discuss list archive


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

[Discuss] linux on usb stick



Stephen Adler wrote:
> So I then used VMware on my laptop to boot off the linux usb memory
> stick configuring the VMware guest to use the memory stick as a raw disk
> device. And after some screwing around I got it to work!

Glad that worked out.


> Now here's an interesting test point. When I run linux on the usb memory
> stick through VMware on my desktop, the response is phenomianl.
> ...
> When I boot from the usb sick on my laptop, the performance sucks.

So you're saying the Linux environment performs better in the VM than it
does running directly on the hardware? Obvious that's opposite what
you'd expect.

With other OSs one would be inclined to blame drivers, but with Linux
most things get probed and dynamically loaded to match the discover
hardware on every boot. However, you still might have some hardware that
requires a proprietary driver that isn't included in your default
install, like an Nvidia driver needed for hardware accelerated video
decoding.

The kernel might also not be optimized for your CPU, though I wouldn't
expect to see the difference you described from running a 32-bit kernel
on a 64-bit machine, and it is unlikely you have a kernel built for old
386 hardware or some such.

I'm sure someone else will have a better suggestion for what to check.


Richard Pieri wrote:
> That's because VMware heavily caches I/O. Direct USB flash I/O
> performance is somewhere between poor and abysmal.

You think disk I/O is relevant to the test case Stephen described? I
wouldn't think so once the browser is loaded and the video starts
playing, unless the browser is spooling the video stream to disk. It
might be.

VMware caching I/O wouldn't really explain it, if the underlying file
system is still on a Flash device, which in Stephen's scenario it was.
If this was true then tuning the kernel to boost file buffers would have
the same effect.

I've tuned Linux systems for running off of Flash drives, but haven't
benchmarked to see whether the changes made a significant difference.
The general idea is that you turn off file access time updating on the
file system, avoid using swap, and put /tmp and other cache file systems
onto a RAM disk. This approach does chew up RAM.

Some articles on the topic:

http://www.cyrius.com/debian/nslu2/linux-on-flash/
http://askubuntu.com/questions/2099/is-it-worth-to-tune-ext4-with-noatime
http://superuser.com/questions/173283/which-linux-distro-to-run-on-a-slow-ssd

 -Tom

-- 
Tom Metro
The Perl Shop, Newton, MA, USA
"Predictable On-demand Perl Consulting."
http://www.theperlshop.com/



BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org