public inbox for linux-mtd@lists.infradead.org
 help / color / mirror / Atom feed
* Samsung RFS Filesystem
@ 2005-05-31 11:42 Ludovic Guilhamat
  2005-06-01  4:41 ` Charles Manning
  0 siblings, 1 reply; 4+ messages in thread
From: Ludovic Guilhamat @ 2005-05-31 11:42 UTC (permalink / raw)
  To: uClinux development list, linux-mtd

Hi,

I would maybe interested in testing the Samsung RFS File System...

Does someone, here, already used it, and what are the conclusions ?

Thank you very much.

Ludovic.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Samsung RFS Filesystem
  2005-05-31 11:42 Samsung RFS Filesystem Ludovic Guilhamat
@ 2005-06-01  4:41 ` Charles Manning
  2005-06-01  6:28   ` Ludovic Guilhamat
  0 siblings, 1 reply; 4+ messages in thread
From: Charles Manning @ 2005-06-01  4:41 UTC (permalink / raw)
  To: Ludovic Guilhamat, uClinux development list, linux-mtd

On Tuesday 31 May 2005 23:42, Ludovic Guilhamat wrote:
> Hi,
>
> I would maybe interested in testing the Samsung RFS File System...
>
> Does someone, here, already used it, and what are the conclusions ?


Please do do some testing. It would be good to see the results, positive or 
negative.

I have not used it, but from a description I can draw some immediate 
conclusions.

RFS = Robust FAT file system

>From the blurb at 
http://www.samsung.com/Products/Semiconductor/Flash/TechnicalInfo/rfs.htm

this looks a bit like FATFS running on top of DOC. It might have some 
transactioning which would make it look a bit more like Microsoft's TFAT.

Neither of these are a proven reliable system. DOC can still get FAT 
corruptions if you don't umount before power loss (== potentially a 
completely scrambled fs). TFAT is completely unprovedn and is slower and not 
always robust (not robust with typical mount options). Perhaps RFS gets the 
robustness right.


Running FAT on NAND costs some performance due to the FTL etc.  To make a 
journaling system on FAT, as RFS claims, costs even more performance.

IMHO: If you want robustnes on NAND use a log structured fs designed for 
flash: YAFFS or JFFS2.

For completeness, I will state that I wrote YAFFS, but I don't think this 
biases my answer.

-- CHarles

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Samsung RFS Filesystem
  2005-06-01  4:41 ` Charles Manning
@ 2005-06-01  6:28   ` Ludovic Guilhamat
  2005-06-01 23:49     ` Charles Manning
  0 siblings, 1 reply; 4+ messages in thread
From: Ludovic Guilhamat @ 2005-06-01  6:28 UTC (permalink / raw)
  To: manningc2; +Cc: linux-mtd, uClinux development list

Charles Manning a écrit :

>On Tuesday 31 May 2005 23:42, Ludovic Guilhamat wrote:
>  
>
>>Hi,
>>
>>I would maybe interested in testing the Samsung RFS File System...
>>
>>Does someone, here, already used it, and what are the conclusions ?
>>    
>>
>
>
>Please do do some testing. It would be good to see the results, positive or 
>negative.
>
>I have not used it, but from a description I can draw some immediate 
>conclusions.
>
>RFS = Robust FAT file system
>
>>From the blurb at 
>http://www.samsung.com/Products/Semiconductor/Flash/TechnicalInfo/rfs.htm
>
>this looks a bit like FATFS running on top of DOC. It might have some 
>transactioning which would make it look a bit more like Microsoft's TFAT.
>
>Neither of these are a proven reliable system. DOC can still get FAT 
>corruptions if you don't umount before power loss (== potentially a 
>completely scrambled fs). TFAT is completely unprovedn and is slower and not 
>always robust (not robust with typical mount options). Perhaps RFS gets the 
>robustness right.
>
>
>Running FAT on NAND costs some performance due to the FTL etc.  To make a 
>journaling system on FAT, as RFS claims, costs even more performance.
>
>IMHO: If you want robustnes on NAND use a log structured fs designed for 
>flash: YAFFS or JFFS2.
>
>For completeness, I will state that I wrote YAFFS, but I don't think this 
>biases my answer.
>
>-- CHarles
>
>  
>
Thanks.

Actually, the system I work on uses Jffs2 (Coldfire with uClinux). But, 
mount times are very long (the memory is a Flash Nand 16Mb). So, I 
planned to test Jffs2 patches (from the Jffs2 Improvement Project), then 
Yaffs. And as RFS is not freely distributed, and I don't know yet its 
cost, I wanted to know if someone had tested it.

Thanks for your response.

Ludovic.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Samsung RFS Filesystem
  2005-06-01  6:28   ` Ludovic Guilhamat
@ 2005-06-01 23:49     ` Charles Manning
  0 siblings, 0 replies; 4+ messages in thread
From: Charles Manning @ 2005-06-01 23:49 UTC (permalink / raw)
  To: Ludovic Guilhamat; +Cc: linux-mtd, uClinux development list


> Actually, the system I work on uses Jffs2 (Coldfire with uClinux). But,
> mount times are very long (the memory is a Flash Nand 16Mb). So, I
> planned to test Jffs2 patches (from the Jffs2 Improvement Project), then
> Yaffs. And as RFS is not freely distributed, and I don't know yet its
> cost, I wanted to know if someone had tested it.

Pick up a more recent JFFS2, I believe that there have been some speedups.

I get a lot of feedback that YAFFS is typically a few times faster than JFFS2

FAT is indeed "instant mounting" and uses less RAM because it is not log 
structured. FAT however does have a bad record as far as reliability and 
write speed on raw NAND devices.

In the end you need to look at pros and cons and fit them to your system 
needs.

-- Charles

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2005-06-02  0:16 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-05-31 11:42 Samsung RFS Filesystem Ludovic Guilhamat
2005-06-01  4:41 ` Charles Manning
2005-06-01  6:28   ` Ludovic Guilhamat
2005-06-01 23:49     ` Charles Manning

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox