[News] [Rival] Windows Vista = XP with Less Applications, Less Speed, More Bugs

In comp.os.linux.advocacy, Erik Funkenbusch
<erik@despam-funkenbusch.com>
wrote
on Tue, 20 Nov 2007 19:21:48 -0600
<zau1gsyhs1xi.dlg@funkenbusch.com>:[color=blue]
> On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:
>[color=green][color=darkred]
>>> Better yet, run a file system that doesn't need defragmentation.[/color]
>>
>> Absolutely.
>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>> filesystems don't fragment.
>>
>> I'm not joking, that's *just* what he said in another group. I replied to
>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?[/color]
>
> There are 2... well, maybe 3 possible reasons the statement "linux
> filesystems don't need defragmenting" could be true.
>
> 1) Linx filesystems don't fragment (we know that's not true, and you are
> admitting to as much here)[/color]

Any system can fragment, given code sequences such as the following:

byte page[8192];

int f1 = open("file0","w");
for(int i = 0; i < 10000; i++)
{
write(f1, page, sizeof(page));
int f2 = open(sprintf("tmp%d", i), "w");
write(f2, page, sizeof(page));
close(f2);
}
close(f1);
for(int i = 0; i < 10000; i++)
{
unlink(sprintf("tmp%d", i));
}

(for purposes of simplicity I'm using char * sprintf()
instead of the sequence char buf[]; sprint(buf, ...), and
ignoring error issues.)

If the system's smart enough file0's later pages might
be allocated from a different pool designed for use
with larger files. However, that's arguably the best
it can do without incurring a rather large performance penalty
while shuffling blocks around.
[color=blue]
>
> 2) Linux filesystems do not suffer any performance degradation from
> fragmentation (that's a silly argument)[/color]

Pedant point: no filesystem suffers performance
degradation; the entire system does, but it depends on
what the user and the system are doing. In particular,
a badly fragmented paging file -- assuming one creates
such on a mostly full volume for some reason -- is not an
issue if it's never used.
[color=blue]
>
> 3) Linux filesystems automatically defrag themselves (which is, in effect
> saying #1, thus also not true).[/color]

Pedant point: no filesystem defragments itself. One could
of course state that the filesystem management code --
in the case of Linux, a module loaded during startup --
makes a reasonable attempt to ensure that file blocks are
allocated in such a way as to resist fragment creation.
Also, the design of the ext2 filesystem tries to minimize
fragmentation in the first place, by using groups.
[color=blue]
>
> If none of those are true, then your statement is provably false.[/color]

None of these are true and the statement is not provably false.
The problem here is that "need" is a conditional term.
[color=blue]
> If Linux
> filesystems can fragment, and fragmentation results in degraded
> performance, then by definition Linux filesystems will need defragmenting
> at some point in time.
>
> So, either you were claiming that linux fileysstems don't fragment, or you
> were claiming that they don't suffer any performance penalty from
> fragmentaiton. Which falsehood were you claiming?
>
> Yes, i'm being a pedant here because you are being a pedant. You might
> have meant "Linux filesystems don't need defragmenting often" or maybe
> "Linux filesystems don't fragment often, so defragmenting isn't usually
> necessary" or maybe even "Linux filesystems fragment so infrequently that
> most people just don't bother to defragment them".[/color]

I for one am curious about that issue. There's a number of factors
here, not the least of which is how the system writes the files.
There's also the issue of how the system reads them afterwards; it
does no good to make the files contiguous if they're stored in an
order not consistent with the system's usage model while doing so.

I am also curious as to how often one might expect to have
to worry about it. Windows systems are defragmented as
often as once a month, as I understand it.

--
#191, [email]ewill3@earthlink.net[/email]
Been there, done that, didn't get the T-shirt.

--
Posted via a free Usenet account from [url]http://www.teranews.com[/url]

On Tue, 20 Nov 2007 18:08:52 -0800, The Ghost In The Machine wrote:
[color=blue]
> In comp.os.linux.advocacy, Erik Funkenbusch
> <erik@despam-funkenbusch.com>
> wrote[/color]
[color=blue][color=green]
>> The hard disk is designed for that. It was not designed for millions
>> of APM power cycles, like Ubuntu puts it through.[/color]
>
> Interesting claim. Cite?
>[/color]

I think he's referring to the spindown problem:

<http://ubuntudemon.wordpress.com/2007/10/26/laptop-hardrive-killer-bug/#comment-31234>

In comp.os.linux.advocacy, AZ Nomad
<aznomad.2@PremoveOBthisOX.COM>
wrote
on Wed, 21 Nov 2007 02:05:05 GMT
<slrnfk74il.btj.aznomad.2@ip70-176-155-130.ph.ph.cox.net>:[color=blue]
> On Tue, 20 Nov 2007 15:35:20 -0800, The Ghost In The Machine <ewill@sirius.tg00suus7038.net> wrote:
>
>[color=green]
>>In comp.os.linux.advocacy, AZ Nomad
>><aznomad.2@PremoveOBthisOX.COM>
>> wrote
>>on Tue, 20 Nov 2007 21:35:08 GMT
>><slrnfk6kof.vlb.aznomad.2@ip70-176-155-130.ph.ph.cox.net>:[color=darkred]
>>> On Tue, 20 Nov 2007 13:16:58 -0800, The Ghost In The Machine <ewill@sirius.tg00suus7038.net> wrote:
>>>
>>>
>>>>In comp.os.linux.advocacy, AZ Nomad
>>>><aznomad.2@PremoveOBthisOX.COM>
>>>> wrote
>>>>on Tue, 20 Nov 2007 20:45:42 GMT
>>>><slrnfk6hrp.vlb.aznomad.2@ip70-176-155-130.ph.ph.cox.net>:
>>>>> On Tue, 20 Nov 2007 12:46:17 -0600, Erik Funkenbusch <erik@despam-funkenbusch.com> wrote:
>>>>>
>>>>>
>>>>>>On Tue, 20 Nov 2007 10:26:44 -0800, alt wrote:
>>>>>>
>>>>>>>> [url]http://www.cnet.com.au/laptops/laptops/0,239035649,339276024,00.htm[/url]
>>>>>>>>
>>>>>>>> For a laptop, those are darned good specifications.
>>>>>>>> And yet Vista Aero -- or is it just regular old Vista? --
>>>>>>>> runs "really quite awful", apparently.
>>>>>>>
>>>>>>> No Aero (as far as I could tell). It was just constant though. Open up
>>>>>>> anything and you got the "hourglass". Lots of UAC interruptions it would
>>>>>>> seem, but I was also doing admin operations.
>>>>>>>
>>>>>>> But it just felt slow. Like something else was more important than the
>>>>>>> user.
>>>>>
>>>>>>Vista takes a few days to "settle down". It has to index search the drive,
>>>>>>it does a full disk defrag, and a number o f other disk intensive
>>>>>
>>>>> disk defrag on a new instalation? What the **** are you running, windows
>>>>> 3.0?
>>>>>
>>>
>>>>Well, can't be too careful; there might be something else
>>>>on that machine, after all, like a virus or a worm or ...
>>>>wait for it ... Linux.
>>>
>>>>Oh, the horror.
>>>
>>>></sarcasm>
>>>
>>>>Does it just defragment? Or does it attempt to seek out
>>>>and destroy? How can we tell? At least with Linux I can
>>>>conceptually pull up the source code for fsck.ext2 and
>>>>peruse it -- and if necessary rebuild it.
>>>
>>> Better yet, run a file system that doesn't need defragmentation.[/color][/color]
>[color=green]
>>Well, yes. :-) Of course fsck.ext2 isn't a defragger
>>anyway; it's a consistency checker. (There is a defragger
>>but I don't know its name and haven't used it. I use
>>reiserfs anyway.)[/color]
>
> Modern file systems don't need defragmentation. That late 70's technology
> hasn't quite made its way into microsoft.[/color]

Depends on the objectives. I for one would think that
the defragmentation of a file system is a given if one is
setting up a CD readonly affair that will be installed on
thousands of machines, for example -- this is of course a
one-time-only sort of thing unless one needs to tweak it
repeatedly for some reason.

Of course that's not the normal usage model any more than
NASCAR racing is the normal usage model for the family
sedan, or for that matter even most sports cars.

I'm quite curious as to how badly performance degrades,
admittedly, and have no real good notion on to how to test
it -- especially since different users have different
partitioning allocations. Linux has a good caching
system, which means that the more RAM one has, the less
likely fragmentation will make all that much difference,
as files will be kept in RAM until flushed out, and no
read is needed. No read, no head seek.

I'm not sure XP's caching is quite as reliable but I don't
have a good notion as to exactly how it decides when to
throw out old pages. As for Vista -- don't have it, not
sure I want it.

--
#191, [email]ewill3@earthlink.net[/email]
/dev/signature: No such file or directory

--
Posted via a free Usenet account from [url]http://www.teranews.com[/url]

On Tue, 20 Nov 2007 19:21:48 -0600, Erik Funkenbusch <erik@despam-funkenbusch.com> wrote:

[color=blue]
>On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:[/color]
[color=blue][color=green][color=darkred]
>>> Better yet, run a file system that doesn't need defragmentation.[/color]
>>
>> Absolutely.
>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>> filesystems don't fragment.
>>
>> I'm not joking, that's *just* what he said in another group. I replied to
>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?[/color][/color]
[color=blue]
>There are 2... well, maybe 3 possible reasons the statement "linux
>filesystems don't need defragmenting" could be true.[/color]
[color=blue]
>1) Linx filesystems don't fragment (we know that's not true, and you are
>admitting to as much here)[/color]
nobody is saying that.


[color=blue]
>2) Linux filesystems do not suffer any performance degradation from
>fragmentation (that's a silly argument)[/color]
nobody is saying that either


[color=blue]
>3) Linux filesystems automatically defrag themselves (which is, in effect
>saying #1, thus also not true).[/color]
three strikes. You're out.


Which word in "doesn't need defragmentation" don't you understand?

On Tue, 20 Nov 2007 18:26:48 -0800, The Ghost In The Machine <ewill@sirius.tg00suus7038.net> wrote:

[color=blue]
>In comp.os.linux.advocacy, Erik Funkenbusch
><erik@despam-funkenbusch.com>
> wrote
>on Tue, 20 Nov 2007 19:21:48 -0600
><zau1gsyhs1xi.dlg@funkenbusch.com>:[color=green]
>> On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:
>>[color=darkred]
>>>> Better yet, run a file system that doesn't need defragmentation.
>>>
>>> Absolutely.
>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>> filesystems don't fragment.
>>>
>>> I'm not joking, that's *just* what he said in another group. I replied to
>>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?[/color]
>>
>> There are 2... well, maybe 3 possible reasons the statement "linux
>> filesystems don't need defragmenting" could be true.
>>
>> 1) Linx filesystems don't fragment (we know that's not true, and you are
>> admitting to as much here)[/color][/color]
[color=blue]
>Any system can fragment, given code sequences such as the following:[/color]
[color=blue]
> byte page[8192];[/color]
[color=blue]
> int f1 = open("file0","w");
> for(int i = 0; i < 10000; i++)
> {
> write(f1, page, sizeof(page));
> int f2 = open(sprintf("tmp%d", i), "w");
> write(f2, page, sizeof(page));
> close(f2);
> }
> close(f1);
> for(int i = 0; i < 10000; i++)
> {
> unlink(sprintf("tmp%d", i));
> }[/color]
[color=blue]
>(for purposes of simplicity I'm using char * sprintf()
>instead of the sequence char buf[]; sprint(buf, ...), and
>ignoring error issues.)[/color]
[color=blue]
>If the system's smart enough file0's later pages might
>be allocated from a different pool designed for use
>with larger files. However, that's arguably the best
>it can do without incurring a rather large performance penalty
>while shuffling blocks around.[/color]
[color=blue][color=green]
>>
>> 2) Linux filesystems do not suffer any performance degradation from
>> fragmentation (that's a silly argument)[/color][/color]
[color=blue]
>Pedant point: no filesystem suffers performance
>degradation; the entire system does, but it depends on
>what the user and the system are doing. In particular,
>a badly fragmented paging file -- assuming one creates
>such on a mostly full volume for some reason -- is not an
>issue if it's never used.[/color]
[color=blue][color=green]
>>
>> 3) Linux filesystems automatically defrag themselves (which is, in effect
>> saying #1, thus also not true).[/color][/color]
[color=blue]
>Pedant point: no filesystem defragments itself. One could
>of course state that the filesystem management code --
>in the case of Linux, a module loaded during startup --
>makes a reasonable attempt to ensure that file blocks are
>allocated in such a way as to resist fragment creation.
>Also, the design of the ext2 filesystem tries to minimize
>fragmentation in the first place, by using groups.[/color]
[color=blue][color=green]
>>
>> If none of those are true, then your statement is provably false.[/color][/color]
[color=blue]
>None of these are true and the statement is not provably false.
>The problem here is that "need" is a conditional term.[/color]
[color=blue][color=green]
>> If Linux
>> filesystems can fragment, and fragmentation results in degraded
>> performance, then by definition Linux filesystems will need defragmenting
>> at some point in time.
>>
>> So, either you were claiming that linux fileysstems don't fragment, or you
>> were claiming that they don't suffer any performance penalty from
>> fragmentaiton. Which falsehood were you claiming?
>>
>> Yes, i'm being a pedant here because you are being a pedant. You might
>> have meant "Linux filesystems don't need defragmenting often" or maybe
>> "Linux filesystems don't fragment often, so defragmenting isn't usually
>> necessary" or maybe even "Linux filesystems fragment so infrequently that
>> most people just don't bother to defragment them".[/color][/color]
[color=blue]
>I for one am curious about that issue. There's a number of factors
>here, not the least of which is how the system writes the files.
>There's also the issue of how the system reads them afterwards; it
>does no good to make the files contiguous if they're stored in an
>order not consistent with the system's usage model while doing so.[/color]
[color=blue]
>I am also curious as to how often one might expect to have
>to worry about it. Windows systems are defragmented as
>often as once a month, as I understand it.[/color]

I just checked my FS's. 0.5% fragmentation. I didn't say it doesn't fragment,
or that there's no performance penalty, or that it unfragments as it goes. Just
that it doesn't require defragmentation.

On Wed, 21 Nov 2007 03:14:45 GMT, AZ Nomad wrote:
[color=blue]
> On Tue, 20 Nov 2007 18:26:48 -0800, The Ghost In The Machine <ewill@sirius.tg00suus7038.net> wrote:
>
>[color=green]
>>In comp.os.linux.advocacy, Erik Funkenbusch
>><erik@despam-funkenbusch.com>
>> wrote
>>on Tue, 20 Nov 2007 19:21:48 -0600
>><zau1gsyhs1xi.dlg@funkenbusch.com>:[color=darkred]
>>> On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:
>>>
>>>>> Better yet, run a file system that doesn't need defragmentation.
>>>>
>>>> Absolutely.
>>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>>> filesystems don't fragment.
>>>>
>>>> I'm not joking, that's *just* what he said in another group. I replied to
>>>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>>>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?
>>>
>>> There are 2... well, maybe 3 possible reasons the statement "linux
>>> filesystems don't need defragmenting" could be true.
>>>
>>> 1) Linx filesystems don't fragment (we know that's not true, and you are
>>> admitting to as much here)[/color][/color]
>[color=green]
>>Any system can fragment, given code sequences such as the following:[/color]
>[color=green]
>> byte page[8192];[/color]
>[color=green]
>> int f1 = open("file0","w");
>> for(int i = 0; i < 10000; i++)
>> {
>> write(f1, page, sizeof(page));
>> int f2 = open(sprintf("tmp%d", i), "w");
>> write(f2, page, sizeof(page));
>> close(f2);
>> }
>> close(f1);
>> for(int i = 0; i < 10000; i++)
>> {
>> unlink(sprintf("tmp%d", i));
>> }[/color]
>[color=green]
>>(for purposes of simplicity I'm using char * sprintf()
>>instead of the sequence char buf[]; sprint(buf, ...), and
>>ignoring error issues.)[/color]
>[color=green]
>>If the system's smart enough file0's later pages might
>>be allocated from a different pool designed for use
>>with larger files. However, that's arguably the best
>>it can do without incurring a rather large performance penalty
>>while shuffling blocks around.[/color]
>[color=green][color=darkred]
>>>
>>> 2) Linux filesystems do not suffer any performance degradation from
>>> fragmentation (that's a silly argument)[/color][/color]
>[color=green]
>>Pedant point: no filesystem suffers performance
>>degradation; the entire system does, but it depends on
>>what the user and the system are doing. In particular,
>>a badly fragmented paging file -- assuming one creates
>>such on a mostly full volume for some reason -- is not an
>>issue if it's never used.[/color]
>[color=green][color=darkred]
>>>
>>> 3) Linux filesystems automatically defrag themselves (which is, in effect
>>> saying #1, thus also not true).[/color][/color]
>[color=green]
>>Pedant point: no filesystem defragments itself. One could
>>of course state that the filesystem management code --
>>in the case of Linux, a module loaded during startup --
>>makes a reasonable attempt to ensure that file blocks are
>>allocated in such a way as to resist fragment creation.
>>Also, the design of the ext2 filesystem tries to minimize
>>fragmentation in the first place, by using groups.[/color]
>[color=green][color=darkred]
>>>
>>> If none of those are true, then your statement is provably false.[/color][/color]
>[color=green]
>>None of these are true and the statement is not provably false.
>>The problem here is that "need" is a conditional term.[/color]
>[color=green][color=darkred]
>>> If Linux
>>> filesystems can fragment, and fragmentation results in degraded
>>> performance, then by definition Linux filesystems will need defragmenting
>>> at some point in time.
>>>
>>> So, either you were claiming that linux fileysstems don't fragment, or you
>>> were claiming that they don't suffer any performance penalty from
>>> fragmentaiton. Which falsehood were you claiming?
>>>
>>> Yes, i'm being a pedant here because you are being a pedant. You might
>>> have meant "Linux filesystems don't need defragmenting often" or maybe
>>> "Linux filesystems don't fragment often, so defragmenting isn't usually
>>> necessary" or maybe even "Linux filesystems fragment so infrequently that
>>> most people just don't bother to defragment them".[/color][/color]
>[color=green]
>>I for one am curious about that issue. There's a number of factors
>>here, not the least of which is how the system writes the files.
>>There's also the issue of how the system reads them afterwards; it
>>does no good to make the files contiguous if they're stored in an
>>order not consistent with the system's usage model while doing so.[/color]
>[color=green]
>>I am also curious as to how often one might expect to have
>>to worry about it. Windows systems are defragmented as
>>often as once a month, as I understand it.[/color]
>
> I just checked my FS's. 0.5% fragmentation. I didn't say it doesn't fragment,
> or that there's no performance penalty, or that it unfragments as it goes. Just
> that it doesn't require defragmentation.[/color]

So what you meant then is that you find the level of fragmentation and
typical performance of a fragmented linux filesystem to be acceptable. Or
perhaps you're saying that you don't feel the need to defragment.

That's a subjective argument. I'm sure many people don't "need" to ever
defragment a Windows OS either.

On Wed, 21 Nov 2007 03:13:27 GMT, AZ Nomad wrote:
[color=blue]
> On Tue, 20 Nov 2007 19:21:48 -0600, Erik Funkenbusch <erik@despam-funkenbusch.com> wrote:
>
>[color=green]
>>On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:[/color]
>[color=green][color=darkred]
>>>> Better yet, run a file system that doesn't need defragmentation.
>>>
>>> Absolutely.
>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>> filesystems don't fragment.
>>>
>>> I'm not joking, that's *just* what he said in another group. I replied to
>>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?[/color][/color]
>[color=green]
>>There are 2... well, maybe 3 possible reasons the statement "linux
>>filesystems don't need defragmenting" could be true.[/color]
>[color=green]
>>1) Linx filesystems don't fragment (we know that's not true, and you are
>>admitting to as much here)[/color]
> nobody is saying that.
>
>
>[color=green]
>>2) Linux filesystems do not suffer any performance degradation from
>>fragmentation (that's a silly argument)[/color]
> nobody is saying that either
>
>
>[color=green]
>>3) Linux filesystems automatically defrag themselves (which is, in effect
>>saying #1, thus also not true).[/color]
> three strikes. You're out.
>
>
> Which word in "doesn't need defragmentation" don't you understand?[/color]

Which part of "subjective argument" don't you understand?

In comp.os.linux.advocacy, AZ Nomad
<aznomad.2@PremoveOBthisOX.COM>
wrote
on Wed, 21 Nov 2007 03:14:45 GMT
<slrnfk78l8.btj.aznomad.2@ip70-176-155-130.ph.ph.cox.net>:[color=blue]
> On Tue, 20 Nov 2007 18:26:48 -0800, The Ghost In The Machine <ewill@sirius.tg00suus7038.net> wrote:
>
>[color=green]
>>In comp.os.linux.advocacy, Erik Funkenbusch
>><erik@despam-funkenbusch.com>
>> wrote
>>on Tue, 20 Nov 2007 19:21:48 -0600
>><zau1gsyhs1xi.dlg@funkenbusch.com>:[color=darkred]
>>> On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:
>>>
>>>>> Better yet, run a file system that doesn't need defragmentation.
>>>>
>>>> Absolutely.
>>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>>> filesystems don't fragment.
>>>>
>>>> I'm not joking, that's *just* what he said in another group. I replied to
>>>> someone that modern linux filesystems don't need defragmenting. The idiot Quack
>>>> troll claimed that I said linux filesystems don't fragment! See how dumb he is?
>>>
>>> There are 2... well, maybe 3 possible reasons the statement "linux
>>> filesystems don't need defragmenting" could be true.
>>>
>>> 1) Linx filesystems don't fragment (we know that's not true, and you are
>>> admitting to as much here)[/color][/color]
>[color=green]
>>Any system can fragment, given code sequences such as the following:[/color]
>[color=green]
>> byte page[8192];[/color]
>[color=green]
>> int f1 = open("file0","w");
>> for(int i = 0; i < 10000; i++)
>> {
>> write(f1, page, sizeof(page));
>> int f2 = open(sprintf("tmp%d", i), "w");
>> write(f2, page, sizeof(page));
>> close(f2);
>> }
>> close(f1);
>> for(int i = 0; i < 10000; i++)
>> {
>> unlink(sprintf("tmp%d", i));
>> }[/color]
>[color=green]
>>(for purposes of simplicity I'm using char * sprintf()
>>instead of the sequence char buf[]; sprint(buf, ...), and
>>ignoring error issues.)[/color]
>[color=green]
>>If the system's smart enough file0's later pages might
>>be allocated from a different pool designed for use
>>with larger files. However, that's arguably the best
>>it can do without incurring a rather large performance penalty
>>while shuffling blocks around.[/color]
>[color=green][color=darkred]
>>>
>>> 2) Linux filesystems do not suffer any performance degradation from
>>> fragmentation (that's a silly argument)[/color][/color]
>[color=green]
>>Pedant point: no filesystem suffers performance
>>degradation; the entire system does, but it depends on
>>what the user and the system are doing. In particular,
>>a badly fragmented paging file -- assuming one creates
>>such on a mostly full volume for some reason -- is not an
>>issue if it's never used.[/color]
>[color=green][color=darkred]
>>>
>>> 3) Linux filesystems automatically defrag themselves (which is, in effect
>>> saying #1, thus also not true).[/color][/color]
>[color=green]
>>Pedant point: no filesystem defragments itself. One could
>>of course state that the filesystem management code --
>>in the case of Linux, a module loaded during startup --
>>makes a reasonable attempt to ensure that file blocks are
>>allocated in such a way as to resist fragment creation.
>>Also, the design of the ext2 filesystem tries to minimize
>>fragmentation in the first place, by using groups.[/color]
>[color=green][color=darkred]
>>>
>>> If none of those are true, then your statement is provably false.[/color][/color]
>[color=green]
>>None of these are true and the statement is not provably false.
>>The problem here is that "need" is a conditional term.[/color]
>[color=green][color=darkred]
>>> If Linux
>>> filesystems can fragment, and fragmentation results in degraded
>>> performance, then by definition Linux filesystems will need defragmenting
>>> at some point in time.
>>>
>>> So, either you were claiming that linux fileysstems don't fragment, or you
>>> were claiming that they don't suffer any performance penalty from
>>> fragmentaiton. Which falsehood were you claiming?
>>>
>>> Yes, i'm being a pedant here because you are being a pedant. You might
>>> have meant "Linux filesystems don't need defragmenting often" or maybe
>>> "Linux filesystems don't fragment often, so defragmenting isn't usually
>>> necessary" or maybe even "Linux filesystems fragment so infrequently that
>>> most people just don't bother to defragment them".[/color][/color]
>[color=green]
>>I for one am curious about that issue. There's a number of factors
>>here, not the least of which is how the system writes the files.
>>There's also the issue of how the system reads them afterwards; it
>>does no good to make the files contiguous if they're stored in an
>>order not consistent with the system's usage model while doing so.[/color]
>[color=green]
>>I am also curious as to how often one might expect to have
>>to worry about it. Windows systems are defragmented as
>>often as once a month, as I understand it.[/color]
>
> I just checked my FS's. 0.5% fragmentation. I didn't say
> it doesn't fragment, or that there's no performance penalty,
> or that it unfragments as it goes. Just that it doesn't
> require defragmentation.[/color]

I for one would hope that the problem manifests as a slow, subtle
degradation, with defragmentation occuring manually. However,
0.5% doesn't mean much to me; is that:

* the number of files that have more than 1 fragment?
* the number of blocks belonging to such files?
* the estimated slowdown if 'find / -xdev -type f | xargs cat'
were to be run on the volume, versus a volume that has
0% fragmentation?

One of the more interesting WindowsSins(tm) is that it
maintains a log file during a C++ build -- in Visual
Studio, anyway. The log file is going to have the worst
fragmentation during a heavy build, as it is being written
to while the other files are being deleted and recreated
(or maybe just overwritten).

If the system is smart enough the .OBJ files will have
few fragments, and the DLLs and EXEs will have high I/O
but not too much fragmentation either, as they're being
written sequentially. However, that log file is going to
get messy.

The equivalent for log files in Unix would be

make all > logfile.txt 2>&1

which would lead to a fragmented logfile, but most
developers building interactively would at most do

make all 2>&1 | less

which creates an in-memory copy pushed through the pipe.
No fragmentation here. Of course, batch builds might
have their own issues but /var/tmp should be on its own
volume anyway, and /bin/install virtually guarantees
contiguity -- it's a copy-move sequence -- if the filesystem
isn't too stupid.

Another WindowsSin(tm) is incremental links. These of course
tack on/patch the .EXE file and can speed things up nicely
during the build proper; however, the file is now going to
have at least two fragments -- and the rest might not be all
that optimally distributed if the source code was edited.

/bin/ld does, however, support an incremental link -- but
the notion differs from Windows'; basically, it's an alias
for -r, which generated an output file that can be fed to
*another* link later on (with relocation records and symbolic
name references and such). Ergo, one gets two contiguous
files instead.

(Nowadays, one would probably use dynamically loadable
libraries -- .so -- instead, and executables have
undefined references that the runtime loader resolves.)

It was also instructive to defragment a disk, then run a
complete rebuild using Visual Studio. The results were
rather ugly, but I don't remember the exact amount of
clutter. To be sure, Windows did state that "one doesn't
have to defragment the drive", but it wasn't zero.

--
#191, [email]ewill3@earthlink.net[/email]
Useless C++ Programming Idea #8830129:
std::set<...> v; for(..:iterator i = v.begin(); i != v.end(); i++)
if(*i == thing) {...}

--
Posted via a free Usenet account from [url]http://www.teranews.com[/url]

On 2007-11-21, Erik Funkenbusch <erik@despam-funkenbusch.com> claimed:
[color=blue]
> So what you meant then is that you find the level of fragmentation and
> typical performance of a fragmented linux filesystem to be acceptable. Or
> perhaps you're saying that you don't feel the need to defragment.
>
> That's a subjective argument. I'm sure many people don't "need" to ever
> defragment a Windows OS either.[/color]

Unless they want it to run worth a crap. Or perhaps they're having
someone else do it. Or maybe it doesn't stand up to malware and bitrot
long enough for them to see deficinecies caused by fragmentation.

I used a program to defrag (called 'defrag') on linux a long, long time
ago. I thought I needed it. Since I stopped using it I've seen zero
difference in performance between using it and not using it.

I've seen the previous system (500MHz, 384M) report fragmentation of
more than 25% since then. Probably higher, but that's the ballpark
number I remember. I've had no slowdown that I could detect. If I
allowed the Windows machine (1.7GHz, 512M) at work to get to that level
it would take forever just to open the defrag program to fix it.

--
Hey Santa! How much for your list of naughty girls?

Sinister Midget wrote:
[color=blue]
> On 2007-11-21, Erik Funkenbusch <erik@despam-funkenbusch.com> claimed:
>[color=green]
>> So what you meant then is that you find the level of fragmentation and
>> typical performance of a fragmented linux filesystem to be acceptable. Or
>> perhaps you're saying that you don't feel the need to defragment.
>>
>> That's a subjective argument. I'm sure many people don't "need" to ever
>> defragment a Windows OS either.[/color]
>
> Unless they want it to run worth a crap. Or perhaps they're having
> someone else do it. Or maybe it doesn't stand up to malware and bitrot
> long enough for them to see deficinecies caused by fragmentation.
>
> I used a program to defrag (called 'defrag') on linux a long, long time
> ago. I thought I needed it. Since I stopped using it I've seen zero
> difference in performance between using it and not using it.[/color]

There is a linux defragging application which is about 10 years old, & only used
on ext2 fs. It's never been updated, as it is simply not needed.
[color=blue]
> I've seen the previous system (500MHz, 384M) report fragmentation of
> more than 25% since then. Probably higher, but that's the ballpark
> number I remember. I've had no slowdown that I could detect. If I
> allowed the Windows machine (1.7GHz, 512M) at work to get to that level
> it would take forever just to open the defrag program to fix it.[/color]

Unlike windoze, there is NO defragment application for ext3 or Reiser journaled
filesystems.

Here's why linux doesn't need it:
[url]http://geekblog.oneandoneis2.org/index.php/2006/08/17/why_doesn_t_linux_need_defragmenting[/url].



--
Operating systems: FreeBSD 6.2 (64bit), PC-BSD 1.4,
Testing: FreeBSD 7.0-BETA 2
Linux systems: Kubuntu 7.10 "Gutsy" amd64,
Debian 4.0, PCLinuxOS 2007.

AZ Nomad wrote:

<snip>
[color=blue]
> I just checked my FS's. Â*0.5% fragmentation. Â*[/color]

This is at 0.2% ATM. One of the other systems I checked which is *older* &
running Ubuntu 6.06LTS (June 06) is at 0.4%. I intend updating it when the next
LTS version is released....might upgrade the machine too.
[color=blue]
> I didn't say it doesn't fragment, or that there's no performance penalty, or
> that it unfragments as it goes. Â*Just that it doesn't require defragmentation.[/color]

I didn't say that it doesn't fragment either, just I just stated that it does
not require *defragmenting*. But as usual Quack took 2+2 & made 6, saying that
I said linux doesn't fragment. What a buffoon.

--
Operating systems: FreeBSD 6.2 (64bit), PC-BSD 1.4,
Testing: FreeBSD 7.0-BETA 2
Linux systems: Kubuntu 7.10 "Gutsy" amd64,
Debian 4.0, PCLinuxOS 2007.

AZ Nomad wrote:
[color=blue]
> On Tue, 20 Nov 2007 19:21:48 -0600, Erik Funkenbusch
> <erik@despam-funkenbusch.com> wrote:
>
>[color=green]
>>On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:[/color]
>[color=green][color=darkred]
>>>> Better yet, run a file system that doesn't need defragmentation.
>>>
>>> Absolutely.
>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>> filesystems don't fragment.
>>>
>>> I'm not joking, that's *just* what he said in another group. I replied to
>>> someone that modern linux filesystems don't need defragmenting. The idiot
>>> Quack troll claimed that I said linux filesystems don't fragment! See how
>>> dumb he is?[/color][/color]
>[color=green]
>>There are 2... well, maybe 3 possible reasons the statement "linux
>>filesystems don't need defragmenting" could be true.[/color]
>[color=green]
>>1) Linx filesystems don't fragment (we know that's not true, and you are
>>admitting to as much here)[/color]
> nobody is saying that.
>
>
>[color=green]
>>2) Linux filesystems do not suffer any performance degradation from
>>fragmentation (that's a silly argument)[/color]
> nobody is saying that either[/color]

Nope, but OTOH they do not suffer any *noticeable* effects.
[color=blue][color=green]
>>3) Linux filesystems automatically defrag themselves (which is, in effect
>>saying #1, thus also not true).[/color]
> three strikes. You're out.
>
>
> Which word in "doesn't need defragmentation" don't you understand?[/color]

The idiot's misinterpreted what you said, just the same way as Quack did with
me. WTF is it with these windoze tards, don't they understand what they read?

--
Operating systems: FreeBSD 6.2 (64bit), PC-BSD 1.4,
Testing: FreeBSD 7.0-BETA 2
Linux systems: Kubuntu 7.10 "Gutsy" amd64,
Debian 4.0, PCLinuxOS 2007.

HangEveryRepubliKKKan <Justice@executethebushtraitor.com> wrote:[color=blue]
>
> "AZ Nomad" <aznomad.2@PremoveOBthisOX.COM> wrote[color=green]
>> Better yet, run a file system that doesn't need defragmentation.[/color]
>
> There is no such thing.[/color]

Remind me not to hire you into my IT department.

I've had Unix and Linux systems running for years with the file
system fragmentation never rising over 2 percent. These are systems
running ftp servers, news feeds, routing mail... etc. Lots of file
reads and writes of various sizes. It is a stone cold fact; modern
inode based files systems do not suffer from the fragmentation that
FAT based or derived systems do (that includes NTFS).

Cheers,

Thad

On Wed, 21 Nov 2007 13:13:22 +0000, William Poaster <wp@linux.amd64.eu> wrote:

[color=blue]
>AZ Nomad wrote:[/color]
[color=blue][color=green]
>> On Tue, 20 Nov 2007 19:21:48 -0600, Erik Funkenbusch
>> <erik@despam-funkenbusch.com> wrote:
>>
>>[color=darkred]
>>>On Tue, 20 Nov 2007 23:59:41 +0000, William Poaster wrote:[/color]
>>[color=darkred]
>>>>> Better yet, run a file system that doesn't need defragmentation.
>>>>
>>>> Absolutely.
>>>> Oops! Careful now! We'll have the Quack troll saying, that we said linux
>>>> filesystems don't fragment.
>>>>
>>>> I'm not joking, that's *just* what he said in another group. I replied to
>>>> someone that modern linux filesystems don't need defragmenting. The idiot
>>>> Quack troll claimed that I said linux filesystems don't fragment! See how
>>>> dumb he is?[/color]
>>[color=darkred]
>>>There are 2... well, maybe 3 possible reasons the statement "linux
>>>filesystems don't need defragmenting" could be true.[/color]
>>[color=darkred]
>>>1) Linx filesystems don't fragment (we know that's not true, and you are
>>>admitting to as much here)[/color]
>> nobody is saying that.
>>
>>
>>[color=darkred]
>>>2) Linux filesystems do not suffer any performance degradation from
>>>fragmentation (that's a silly argument)[/color]
>> nobody is saying that either[/color][/color]
[color=blue]
>Nope, but OTOH they do not suffer any *noticeable* effects.[/color]
[color=blue][color=green][color=darkred]
>>>3) Linux filesystems automatically defrag themselves (which is, in effect
>>>saying #1, thus also not true).[/color]
>> three strikes. You're out.
>>
>>
>> Which word in "doesn't need defragmentation" don't you understand?[/color][/color]
[color=blue]
>The idiot's misinterpreted what you said, just the same way as Quack did with
>me. WTF is it with these windoze tards, don't they understand what they read?[/color]

It's not that. It's a diversionary tactic away from the previous discussion
of vista being dog slow. This whole fragmentation discussion spawned from
a windows appologist theorizing that vista piss poor performance was due
to it doing a defrag in the background.

William Poaster wrote:
[color=blue]
> I'm not joking, that's *just* what he said in another group. I
> replied to someone that modern linux filesystems don't need
> defragmenting.[/color]

If they don't need defragmenting, why do defragmenters even exist for
Linux/OSS filesystems in the first place?

uh duh... my names Dumb Willie.... uh duh.

[email]thad05@tux.glaci.delete-this.com[/email] wrote:[color=blue]
> HangEveryRepubliKKKan <Justice@executethebushtraitor.com> wrote:[color=green]
>>
>> "AZ Nomad" <aznomad.2@PremoveOBthisOX.COM> wrote[color=darkred]
>>> Better yet, run a file system that doesn't need defragmentation.[/color]
>>
>> There is no such thing.[/color]
>
> Remind me not to hire you into my IT department.[/color]

I'll remind you never to hire Mr. ext3 or Mrs. XFS:

[url]http://www.webservertalk.com/archive202-2005-5-1037740.html[/url]
[url]http://oclug.on.ca/archives/oclug/2002-August/022746.html[/url]
[url]http://lopsa.org/pipermail/tech/2006-October/000994.html[/url]

[color=blue]
> I've had Unix and Linux systems running for years with the file
> system fragmentation never rising over 2 percent. These are systems
> running ftp servers, news feeds, routing mail... etc. Lots of file
> reads and writes of various sizes. It is a stone cold fact; modern
> inode based files systems do not suffer from the fragmentation that
> FAT based or derived systems do (that includes NTFS).
>
> Cheers,
>
> Thad[/color]

DFS wrote:
[color=blue]
> William Poaster wrote:
>[color=green]
>> I'm not joking, that's *just* what he said in another group. I
>> replied to someone that modern linux filesystems don't need
>> defragmenting.[/color]
>
> If they don't need defragmenting, why do defragmenters even exist for
> Linux/OSS filesystems in the first place?[/color]

Oh, you can certainly point us to those, right, DumbFull****?
[color=blue]
> uh duh... my names Dumb Willie.... uh duh.[/color]

No. It is DumbFull**** for a reason
--
There are no stupid questions, but there are a lot of inquisitive idiots.

On 2007-11-21, William Poaster <wp@linux.amd64.eu> claimed:[color=blue]
> Sinister Midget wrote:
>[color=green]
>> On 2007-11-21, Erik Funkenbusch <erik@despam-funkenbusch.com> claimed:
>>[color=darkred]
>>> So what you meant then is that you find the level of fragmentation and
>>> typical performance of a fragmented linux filesystem to be acceptable. Or
>>> perhaps you're saying that you don't feel the need to defragment.
>>>
>>> That's a subjective argument. I'm sure many people don't "need" to ever
>>> defragment a Windows OS either.[/color]
>>
>> Unless they want it to run worth a crap. Or perhaps they're having
>> someone else do it. Or maybe it doesn't stand up to malware and bitrot
>> long enough for them to see deficinecies caused by fragmentation.
>>
>> I used a program to defrag (called 'defrag') on linux a long, long time
>> ago. I thought I needed it. Since I stopped using it I've seen zero
>> difference in performance between using it and not using it.[/color]
>
> There is a linux defragging application which is about 10 years old, & only used
> on ext2 fs. It's never been updated, as it is simply not needed.[/color]

That's the one. I used it 3 or 4 times before I found out it wasn't
necessary. I stopped using it and waited to see the result. The result
was, I'd wasted my time using it.
[color=blue][color=green]
>> I've seen the previous system (500MHz, 384M) report fragmentation of
>> more than 25% since then. Probably higher, but that's the ballpark
>> number I remember. I've had no slowdown that I could detect. If I
>> allowed the Windows machine (1.7GHz, 512M) at work to get to that level
>> it would take forever just to open the defrag program to fix it.[/color]
>
> Unlike windoze, there is NO defragment application for ext3 or Reiser journaled
> filesystems.
>
> Here's why linux doesn't need it:
> [url]http://geekblog.oneandoneis2.org/index.php/2006/08/17/why_doesn_t_linux_need_defragmenting[/url].[/color]

That won't stop Ewik trying to "teach" people that linux is just as
awful as Windwoes. Since when has correction ever stopped him from
repeating the same "errors"?

--
You are a rare treasure. Maybe someone will bury you.

On 2007-11-21, [email]thad05@tux.glaci.delete-this.com[/email] <thad05@tux.glaci.delete-this.com> claimed:[color=blue]
> HangEveryRepubliKKKan <Justice@executethebushtraitor.com> wrote:[color=green]
>>
>> "AZ Nomad" <aznomad.2@PremoveOBthisOX.COM> wrote[color=darkred]
>>> Better yet, run a file system that doesn't need defragmentation.[/color]
>>
>> There is no such thing.[/color]
>
> Remind me not to hire you into my IT department.
>
> I've had Unix and Linux systems running for years with the file
> system fragmentation never rising over 2 percent. These are systems
> running ftp servers, news feeds, routing mail... etc. Lots of file
> reads and writes of various sizes. It is a stone cold fact; modern
> inode based files systems do not suffer from the fragmentation that
> FAT based or derived systems do (that includes NTFS).[/color]

Unless they're getting full. _That's_ more of a problem than the
resulting fragmentation, though. In fact, I've had fragmentation on a
nearly full disk that would have made XP almost stop dead. The reason
it got so full without my notice was precisely because there was no
noticeable performance hit.

Reaching 100% full, now *that's* a problem. One I've had a couple of
different times.

--
Generally speaking, it's dangerous to generalize.

On 2007-11-21, Peter Köhlmann <peter.koehlmann@t-online.de> claimed:[color=blue]
> DFS wrote:
>[color=green]
>> William Poaster wrote:
>>[color=darkred]
>>> I'm not joking, that's *just* what he said in another group. I
>>> replied to someone that modern linux filesystems don't need
>>> defragmenting.[/color]
>>
>> If they don't need defragmenting, why do defragmenters even exist for
>> Linux/OSS filesystems in the first place?[/color]
>
> Oh, you can certainly point us to those, right, DumbFull****?[/color]

This is the only one I know about:

[url]http://freshmeat.net/projects/defrag/[/url]

Notice the /added/ and /updated/ dates. Notice even further all of the
discussions below. Looks to me like it wouldn't even work with some
twiddling.

Maybe DuFS knows of some more.
[color=blue][color=green]
>> uh duh... my names Dumb Willie.... uh duh.[/color]
>
> No. It is DumbFull**** for a reason[/color]

He's so dumb he thinks he has "Willie" someplace in his online moniker?

--
Behind every great man stands a woman; correcting him.