disappointed result on enabling EBS on Samsung small page NAND chip

disappointed result on enabling EBS on Samsung small page NAND chip

[Peter J Zhu]

hello,

Recently I enanbled EBS with 2.6.15 on Samsung small
page(oobblock = 512B)NAND chip. The results is as
following:

            Without EBS       With EBS

Imagesize  31378292B         42379624B

Mouttime   17s ~             14s ~

The overhead space is too much while providing NOT
high mout time improvement But from the thread
http://mhonarc.axis.se/jffs-dev/msg01763.html,my
results seems too bad compared with what's claimed. 

The only diff I can imagine is that he might use big
page NAND. Any hints are very much appreciated. 

Thanks
Peter


		
___________________________________________________________ 
无限容量雅虎相册，原图等大下载，超快速度，赶快抢注！ 
http://cn.photos.yahoo.com

EBS effect

[Peter J Zhu]

hello,
 
  Recently I enanbled EBS with 2.6.15 on Samsung small
  page(oobblock = 512B)NAND chip. The results is as
  following:
  
             Without EBS       With EBS

  Imagesize  31378292B         42379624B

  Mouttime   17s ~             14s ~

  The overhead space is too much while providing NOT
  high mout time improvement But from the thread
  http://mhonarc.axis.se/jffs-dev/msg01763.html, 
  results seems too bad compared with what's claimed. 

  The only diff I can imagine is that he might use big
  page NAND. Any hints are very much appreciated. 

Thanks
Peter



	

	
		
___________________________________________________________ 
雅虎1G免费邮箱百分百防垃圾信 
http://cn.mail.yahoo.com/

Re: EBS effect

[Artem B. Bityutskiy]

Peter J Zhu wrote:
> hello,
>  
>   Recently I enanbled EBS with 2.6.15 on Samsung small
>   page(oobblock = 512B)NAND chip. The results is as
>   following:
>   
>              Without EBS       With EBS
> 
>   Imagesize  31378292B         42379624B
> 
>   Mouttime   17s ~             14s ~
> 
>   The overhead space is too much while providing NOT
>   high mout time improvement But from the thread
>   http://mhonarc.axis.se/jffs-dev/msg01763.html, 
>   results seems too bad compared with what's claimed. 
> 
>   The only diff I can imagine is that he might use big
>   page NAND. Any hints are very much appreciated. 
> 
What's the size of your MTD device?

-- 
Best Regards,
Artem B. Bityutskiy,
St.-Petersburg, Russia.

回复： Re: EBS effect

[Peter J Zhu]

The size of my MTD device is 48M. And I used the small
block NAND with erase block size being 16KB.

--- "Artem B. Bityutskiy" <dedekind@yandex.ru>写道:

> Peter J Zhu wrote:
> > hello,
> >  
> >   Recently I enanbled EBS with 2.6.15 on Samsung
> small
> >   page(oobblock = 512B)NAND chip. The results is
> as
> >   following:
> >   
> >              Without EBS       With EBS
> > 
> >   Imagesize  31378292B         42379624B
> > 
> >   Mouttime   17s ~             14s ~
> > 
> >   The overhead space is too much while providing
> NOT
> >   high mout time improvement But from the thread
> >   http://mhonarc.axis.se/jffs-dev/msg01763.html, 
> >   results seems too bad compared with what's
> claimed. 
> > 
> >   The only diff I can imagine is that he might use
> big
> >   page NAND. Any hints are very much appreciated. 
> > 
> What's the size of your MTD device?
> 
> -- 
> Best Regards,
> Artem B. Bityutskiy,
> St.-Petersburg, Russia.
> 
>
______________________________________________________
> Linux MTD discussion mailing list
>
http://lists.infradead.org/mailman/listinfo/linux-mtd/
> 



		
____________________________________________________

Re: EBS effect

[Ferenc Havasi]

Hi Peter,

>  The overhead space is too much while providing NOT
>  high mout time improvement But from the thread
>  http://mhonarc.axis.se/jffs-dev/msg01763.html, 
>  results seems too bad compared with what's claimed. 
>
>  The only diff I can imagine is that he might use big
>  page NAND. Any hints are very much appreciated. 
>  
>
The size of the overhead space and the rate of speedup depend on the
size of the erase blocks, not the page size.

What is the size of your erase block? If it is small, unfortunatelly EBS
will not be too effective.

BR,
Ferenc