[prev in list] [next in list] [prev in thread] [next in thread] 

List:       yaffs
Subject:    [Yaffs-archive] yaffs performance
From:       Charles Manning <manningc2 () actrix ! gen ! nz>
Date:       2002-07-03 7:19:51
Message-ID: 20020703072642.EE6C913FB0 () dragon ! actrix ! co ! nz
[Download RAW message or body]

I'll transfer this trhead over onto the yaffs list....

On Wed, 03 Jul 2002 01:14, you wrote:
> Setting lcp-echo-interval to 5 and lcp-echo-failure toi 12 (ie 1 minute
> total) got the same scp to 58% before it stalled and then the ppp was
> dropped with a Peer not responding message.
> Restarting and it got to 72%. Restarting with lcp-echo-failure at 5 and
> lcp-echo-interval of 60 (ie 5 mins timeout) got to 86%. Restarting and it
> finished without ppp dropping.
>
> I wonder if a schedule() is needed somewhere?
>
> scp to ramfs was much faster (so maybe not a fair comparison) but without
> incident.
>
> ram-bash was rm'd ok.
> The first and second yaffs-bash took an age to rm.
>
> Nick

yaffs itself does not do any schedule() or similar calls and I don't think it 
needs to. Instead, these will be happening in the nand mtd layer when calling 
writes & erases etc. To get optimal performance might need some tuning of the 
NAND mtd layer. To get a handle on this loading could you try a test? Run up 
yaffs on the nand mtd emulation. This uses the full mtd interface in yaffs, 
but emulates NAND with RAM. This emulation does no schedules etc. The 
difference in speed is thus attributed to the NAND code.

A snippet of kernel logging from yaffs would be interesting to note the size 
of the writes. If these don't fit the page size, then performance will 
degrade (more page rewrites, more erases...).

Also, rm'ing a file will be resulting in erases. As a file is erased, the 
pages it occupies are being freed up as dirty. Once a block holds only dirty 
pages it will be erased and reclaimed (inline with the page erase).

Hmmm, maybe it's time for me to add some more detail to procfs etry for yaffs 
to improve performance tracking.


-- Charles






---------------------------------------------------------------------------------------
This mailing list is hosted by Toby Churchill open software (www.toby-churchill.org).
If mailing list membership is no longer wanted you can remove yourself from the list by 
sending an email to yaffs-request@toby-churchill.org with the text "unsubscribe" 
(without the quotes) as the subject. 



[prev in list] [next in list] [prev in thread] [next in thread]