| From: | Jens Kirk |
| Date: | 14 Aug 2000 at 09:28:42 |
| Subject: | Re: SFS vs PFS |
Hello Neil
On 14-Aug-00, you wrote:
> Kisa 2o0o said,
>
>>> Here are Amiga Format's filesystem test results in seconds:
>>> Copy Rename Delete
>
>>> FFS 254.72 103.04 7.74
>>> PFS3 35.04 4.16 2.64
>>> SFS 54.36 5.32 8.46
>
>> This test gave WRONG results! Upon this test deleting file
>> on FFS is faster (!!!) then on SFS ?! Try to create a very
>> huge file (ISO image is ok) and then delete it! IT take about
>> 2-3 mins on FFS! and only few seconds on SFS !!!
>
> The test didn't use a single large file, while is atypical usage. It
> used a directory of average files. The tests I ran when reviewing PFS3
> for Amiga Active gave similar results, even though the tests were
> completely independent of AF's.
Well one thing is speed .. Another is features! I know that SFS have some
features you wont see on PFS3. The features missing in SFS at this point
I'm sure we'll see later, I'm not sure I can say the same about PFS3. I'm
also under the impression that SFS in under continued development in a
higher degree than PFS3. It's been very long since I last saw an update
from GREED.
PFS3 seems a bit odd sometimes, when you want to lock onto a file. I've
suggested a sync command as in Linux or at least a system call that would
let one force the write that normally happens after 2 seconds or so.. The
GREED people just said that a diskchange would do the job ... But this
takes the whole partion/device offline for some time and might result in a
requester asking for that partion (especially if you run something like
rc5des on that device)..
While I find PFS3 to be both secure and fast I also find it missing things
like this.
Another example of things going wrong with PFS3 is the Apache server.. Which
(at least in the past - I havn't tested recently) would fail when the log
files had their place on a PFS partion.
Afaik SFS don't have this problem
Regards
Jens //
\\/Kirk
millennium
Everything should be made as simple as possible, but not simpler.
-- Albert Einstein