ts-7000
[Top] [All Lists]

[ts-7000] Re: TS7300 optimization (on debian, on filesystem, on power)

To:
Subject: [ts-7000] Re: TS7300 optimization (on debian, on filesystem, on power)
From: "suptouch" <>
Date: Thu, 22 Jun 2006 16:31:27 -0000
many thanks for your remarks !!

First, about the journaling file system, I think is not a good idea to
use it on flash (except if these is done to do it), the journal will
use always the same blocs and it'll reduce the flash life ...

Also, I can anticipate when power be turned off , but I can't accept a
long power-off time to the root filesystem unmounting (killing
processes, etc.)  and I've to unmount not to corrupt the filesystem !!
so I've a new idea ...

The linux root filesystem should not be change during my project life
except the /var/log ... Only some executable will be upgrade. So,
perhaps I can mount the linux on a ramdisk, and have only one flash
ext2 part with the upgradable files. With this solution, when I've to
power-off, I've only to umount the ext2 part, but always have the root
filesystem !! The root filesystem can be next power-off without
corrupt risk. 
To do this, to load the linux in RAM, the linux have to be a small
embedded linux like TS-linux and not the Debian linux ... 

Have you got others remarks ? ideas ? 



--- In  "Don W. Carr" <> wrote:
>
> Yes, I think for many applications, you can close files when they do not
> need to be open, and anticipate when power would/could be turned
off. For
> one of my applications, if the train is not moving, I can close the
files
> and buffer any events in memory. When the train IS moving, we should NOT
> lose power, and if we do, we have a lot bigger problems than losing
a block
> of flash memory on the SD card. Heck in my application, I might even run
> directly off of the batteries, and power myself down when the train
is shut
> off.  Don.
> 
> On 6/22/06, Jesse Off <> wrote:
> >
> >    BSD uses a technique dubbed "soft-deps" on its filesystem which
does
> > not require a journal. It acheives about the same benefits of a
> > meta-data journaled filesystem without the extra writes.
> > http://www.mckusick.com/softdep/
> >
> > I have some history with another filesystem called LFS. (log-
> > structured filesystem) This FS is "pure-journaled" and not
> > just "metadata journaled" like (I presume) EXT3. Pure-journaled
> > filesystem designs (like YAFFS2) do not need extra block writes to
> > commit the journal and would represent about the closest thing to
> > perfection as far as general purpose flash filesystems go.
> > Unfortunately, you cannot use YAFFS2 with a traditional block device
> > like SD or a hard drive -- it needs flash, and more specifically,
> > commodity NAND flash chips in their raw interface.
> >
> > If there are any generic Linux filesystem types that are "pure-
> > journaled", that would be IMHO preferred choice #1. #2 would be
> > something with metadata ordering such as softdep. #3 would be
> > hybrid meta-data journaled fs's like EXT3 or traditional filesystems
> > such as FAT/EXT2 mounted with sync & noatime.
> >
> > Another thing you must consider if you're trying to accomodate
> > someone yanking power in the middle of operation is that if there
> > just happens to be a flash write going on when power collapses, you
> > will likely loose 1 block of flash permanently irrespective of FS.
> > I believe this is also the case with SD. You usually don't have
> > anything to worry about since SD will then remap a sector from its
> > spare pool and YAFFS2 will probably just never use that block again
> > (thereby decreasing free space).
> >
> > //Jesse Off
> >
> >
> > --- In  <ts-7000%40yahoogroups.com>, "Don
W. Carr"
> > <doncarr@> wrote:
> > >
> > > This brings us back to using a journaling file system on SD flash
> > cards. I
> > > am thinking if you have a big enough flash that distributes the
> > writes, it
> > > should not be a problem. I really would like to know how many
> > extra writes
> > > there are with a journaling file system. Don.
> > >
> > > On 6/21/06, suptouch <suptouch@> wrote:
> > > >
> > > >
> > > >
> > > > > We should release these modifications within a few days...
> > We're
> > > > > still trying to squeeze a few more tenths of a second. Below
> > is the
> > > > > output of the fastboot dump-- as you can see we don't print
> > kernel
> > > > > messages any more either (we got a couple tenths of a second
> > from
> > > > > that)
> > > > > //Jesse Off
> > > >
> > > > Good news ! so I'm waiting fo this modification with a new TS-
> > BOOTROM
> > > > and a new INITRD. Perhaps you can create two INITRD, one to
> > start the
> > > > Debian and the second to start the TS-linux (lighter than the
> > debian).
> > > >
> > > > About the linux root filesystem, which robust filesystem can I
> > use
> > > > instead of the ext2 (sync+noatime) on the sdcard (yaffs2, jffs2,
> > etc.) ?
> > > > The goal is to allow power switch-off the power without the
> > command
> > > > "shutdown -h now" ?
> > > >
> > > > And to finish, about the power consumption, Can I use the 2mA
> > STOP
> > > > Mode of the EP9302 under linux (cpu stop) ? If yes, have you got
> > a
> > > > idea of the method ?
> > > >
> > > >
> > > >
> > >
> > >
> > >
> > > --
> > > Dr. Don W. Carr
> > > J. G. Montenegro 2258
> > > Guadalajara, Mexico
> > > +52-333-630-0704
> > > +52-333-836-4500 ext 2930
> > >
> >
> >  
> >
> 
> 
> 
> -- 
> Dr. Don W. Carr
> J. G. Montenegro 2258
> Guadalajara, Mexico
> +52-333-630-0704
> +52-333-836-4500 ext 2930
>






------------------------ Yahoo! Groups Sponsor --------------------~--> 
Check out the new improvements in Yahoo! Groups email.
http://us.click.yahoo.com/6pRQfA/fOaOAA/yQLSAA/CFFolB/TM
--------------------------------------------------------------------~-> 

 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/ts-7000/

<*> To unsubscribe from this group, send an email to:
    

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 


<Prev in Thread] Current Thread [Next in Thread>
Admin

Disclaimer: Neither Andrew Taylor nor the University of NSW School of Computer and Engineering take any responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the birding-aus mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU