[Ubuntu-be] Remember - IRC meeting tomorrow Thursday 28/10/2010 at 21.00 hr.

[Egon Geerardyn]

Dear Jan,

To answer your questions:

a) the memory requirements of Linux vary a lot with what programs you use
(and what for), but a basic Linux desktop (e.g. writing letters, some small
spreadsheets, some e-mail, some photo's and some web browsing) is perfectly
feasible with the amount of RAM you mention. Regarding regular Ubuntu, I
haven't tried a really low-end setup (I did check with 512 MiB RAM and
fairly decent Mobile Pentium 4 and quite low-end Mobile Celeron (P4
generation)). Works really well for most people; but don't expect it to beat
any new equipment. In any case: some variant of Ubuntu (e.g. Xubuntu or
Lubuntu) have lower requirements than Ubuntu or will perform better on older
hardware. This is because they have "less" features (they are just more
optimized towards memory and processor usage). On the other hand, Linux can
be tailored to fit your needs (add this, remove that, ...) while in Windows
only very few people bother to slim down their installation.

Overall Linux has better memory management algorithms which the user can
tweak a little. For most Windows folk, the effeciency of an algorithm is
measured by how much free memory you have.  That actually is a very bad
measure, as the amount of free (or unused) memory does not really contribute
to the performance. Not having any free memory does slow your computer
terribly, but as long as you have a bit of free memory it should work fine.
In layman's terms, the free memory is head room.

During the Windows XP era, the comparison between Linux and Windows was
fairly simple: the Windows memory management was prehistoric. Very little
caching (that's a mechanism to make a computer faster) and swapping even if
the computer doesn't need it. With Windows Vista, the algorithms used were
really improved a lot: Vista caches a lot more data, which should make it
faster. Windows 7 has improved little, only the overall performance of the
system was improved, since Vista was really a resource hog.

In less technical terms (for your pupils): suppose you are making your
homework (data) on your desk (RAM space). Then you have two general
approaches. Some people (i.e. Linux) will get the assignment and directly
after that they will think about what they might also need so they go to the
closet in the other room and take  (cache) their pens, some text books, a
dictionary, a calculator, writing paper, ... out of the closet if it's not
already on the desk. So they have most of the things they need at hand and
they start to work. Other people (i.e. Windows XP) will take the assignment
and sit down to think about the assignment and they don't understand a word
in the text. They fetch the dictionary in the other room, return with the
dictionary and work on the assignment until they want to write down
something so they fetch paper and pens, and so on and so on. The first kind
of people might have less space free but will have more at hand to quickly
solve their homework; while the other people will have lots of free space on
their desk but they will have to run a lot to get other things they need
(and will lose a lot of time while their desk is large enough to hold a few
extra items). In this case, as Linux and Windows 7/Vista notice that your
free space is becoming too small to work comfortably, they will put away the
parts of the cache they expect not to use (i.e. remnants from other
assignments).

A bit more background about memory; the actual need for memory is based on 2
large factors: we want a lot of storage capacity (loads of data (e.g.
videos) to process and we want to pay as little as possible. Some memories
are quite cheap per storage unit (e.g. hard disks, dvd-r, ...), nowadays it
costs next to nothing to store a few gigabytes on a HD; but on the other
hand these devices are really slow. On the other hand, electronic memories
are really fast, you just store a very small bunch of electrons in a
transistor (or next to one); no need for any mechanical parts. But these
memories are really expensive to store a lot of data in (the level 1 (L1)
cache of your processor is a very nice example). Of course some memories are
in between both extremes (e.g. your RAM memory is a lot faster than a hard
disk but not as expensive as those really fast memories). To get the
positive sides (fast and a large storage for little money) from both
extremes, we apply a little trick. When our processor needs a bit of data,
it only looks in the L1 cache (the fast one), if we notice that the needed
piece of data is not available we try to get it from a slower memory (L2
cache) or even a slower memory (L3 cache; RAM memory or finally the hard
disk). The piece of data is then copied to all the fast memories such that
if we need that same piece of information a little bit later, we can access
it really fast (because it will be in the L1 cache). On the other hand, when
data goes stale in this fast memory (the fast memory is full, so the
computer tries to free up some space for new data), the data is written back
to the slower memory. That is the principle of caching.

Swapping is more or less the opposite. You know you have a fixed amount of
RAM available in your computer, so when you try to use more than this
amount, the computer will grind to a halt. It has no more possibilities to
store any data; so to circumvent this, most operating systems use a certain
file in windows called the page file, in Linux a swap file) or even a whole
partition (which is used the most in Linux) to store parts of the memory. So
the computer tries to free up memory by writing it to the hard disk; but as
I already told you: we have a lot of space on the hard disk, but it is
extremely slow. That's why your computer becomes really slow when your RAM
is full; your operating system will frantically try to write the memory to
the hard disk (which you can actually hear in most cases).

Regarding swapping, I've noticed that most Windows system almost always have
something in their swap file. With Linux you can tune this behavior (you can
tell Linux how hard he has to swap), but the default setting is quite
decent. In most cases Linux only mostly swap space if the memory is getting
too full, so you always retain full RAM speed. (I have tweaked my settings a
bit, I have about 4.5GiB (out of 8) used RAM at the moment, but only 28 MiB
of swap used (less than what the default settings should give)). Windows has
another characteristic in my experience: even when your memory is not full,
Windows will use a lot of swap space (I expect about 1GiB of swap in my
case). So Windows is more agressive to "swap out" (write to disk) your RAM
contents while that is not necessary.

http://www.codinghorror.com/blog/2006/09/why-does-vista-use-all-my-memory.html




b) Defragmenting was something that during the mid-nineties was a really hot
topic. This was because of the FAT file system; which was a rather "dumb"
file system. I won't go into details (because it has been a few years since
I revised the technical side of file systems). Fragmenting is a phenomenon
that occurs when you save and delete a lot of files on a partition. Let's
say you want to save a certain file on you disk, so you look at your disk
and notice that at the end you don't have enough space left to store that
file; but in between some older files, you have deleted other files so in
total free space you have enough room to fit in the new file. What you do
is, you fragment the new file (cut it in pieces) and put it in the free
holes you find on your disk. When reading the file from the disk (or
overwriting the file with a newer version); between each part of the file,
your hard disk will have to look for the next part. With FAT this was rather
tedious, so it took a long time, so your computer was slow when your disk
got fragmented. The same thing happens when you add content to a file that
has other files after it: you will have to write the new content in another
chunk of file. With newer file systems (NTFS for Windows, EXTFS for Linux),
the designers thougth of ways to lessen these problems. I don't really know
about the details there or any comparison between Windows and Linux
regarding this issue. A possible strategy is to keep some free space after
each file, such that when you append to the file, the data written directly
after this file. That is something that was implemented (or should be
implemented?) in ext4fs which has quite recently become the standard file
system in Ubuntu.

http://geekblog.oneandoneis2.org/index.php/2006/08/17/why_doesn_t_linux_need_defragmenting
http://www.msversus.org/microsoft-windows-performance.html


c) Windows has no real package management so no real packages. An
application (OS independent) generally makes use of libraries to provide
certain functions (e.g. to draw figures on the screen, to print, ... instead
of reinventing the wheel every time). Windows has some standard locations
for libraries (e.g. C:\Windows\system32 if I'm not mistaken) to store these
common parts for your programs. When a program is distributed, it has to
know whether these libraries are already on your computer or not. So if you
can't assume that it is there, you will have to either provide the libraries
you need with your program/installer (or package if you want to call it that
way). If you provide the library you can store it either in the standard
location which is shared for all programs or put it in the directory of your
program. E.g. if you rely on Microsoft Office 2007 to be installed in your
program, the program should check that the office libraries are installed.
When you distribute your program, you can either provide MS Office together
with your software (which might be a tad expensive) or your installer can
say to the user "I don't want to install the software unless you install MS
Office" (and then it just closes down). Above this problem, you might get
some problems with different versions of the same libraries.

Windows can only handle 1 version of a library in the shared location. So
suppose you have 2 programs that need a different version of a library, you
will have to install that library in the directory of one of the programs.
So if you're unsure whether to include a certain DLL, just include it and
your program will work. This of course causes that if most programmers work
that way, you will end up with a lot of different copies of the same
library. When a bug is found in that library it is very tedious to fix all
occurances of the library (in most cases, people don't even bother!). This
has happened a few years back when something was wrong in some graphical
library in Windows. A lot of programs had to be fixed one by one because of
this.

In most Linux distributions you have a package management system (dpkg for
Ubuntu with DEB-packages and user interfaces aptitude, apt-get, synaptic,
...). That defines a set of rules to represent both the files of a program,
its dependencies towards libraries or other programs and install/uninstall
procedures. So when your program needs some library, it is easy to specify
what library you want and even what version is good to be used. There is
little extra effort for the programmer to specify all of this (because it is
standard for the distribution (and in most cases also for some closely
related distributions (e.g. Ubuntu uses the Debian system, hence
.DEB-files)). Linux can also handle multiple versions of libraries quite
well.

The need for different versions of libraries is that newer libraries may add
or remove features that older software depends on. It is generally
discouraged to remove features from a library, but in some cases it has to
be done.

To summarize: Ubuntu offers one streamlined way to manage your installed
programs so program's don't have to include the same file every time and
there is a clean way to check everything. More so: package management will
even notice if two programs try to overwrite the same library (which might
cause both programs to malfunction). It allows for different library
versions and because most libraries are only present once (some applications
do use the strategy I described for Windows), it's easier to fix bugs. That
is IMHO the true power of (a certain) Linux (distribution), the package
management system: a unified installation/deinstallation/management system.
If you want to know a bit more about this subject: DLL Hell (or dependency
hell) is used to describe most problems related to libraries.


d) I wouldn't say that that is the most important safetly feature, but it
does help. Windows folk usually are used to click every next or OK button
that pops up without reading the messag because you get flooded with those
messages ("Are you sure you want to do this or that?", click "Next" 20 times
to install this or that piece of software, ...). The same with UAC in Vista
and Windows 7: it only requires you to click OK to become administrator on
most home computers (in companies you have to enter the administrator
credentials, which explains why in the office Windows can work). In Linux
you have to type your password, so that's when most people think about what
they are doing. In general, people are more reluctant to type in a password
than to click OK. Some people may find it tedious, that's a little down side
to this.

On the other hand: Linux has always been a multi user OS (like unix).
Windows was a single user OS in the beginning (Windows 1 till Windows Me)
and has been adapted (in Windows NT) to handle multiple users. But some
features from the old windows era are still present to be able to run old
software. In Linux the security features needed in a multi user setup were
there by design; in Windows they were built onto the existing code base
after a while, so some security code will be sub-optimal.

There actually is a lot of standardization, altough some distributions have
little differences in the details. Though not every software from any
distribution may work on another distribution (I'm currently trying to get a
quite old circuit simulator for Red Hat Linux and regular Unix to work on
Ubuntu 10.10; easy is a different ball game), but most recent programs and
most commonly used programs are distribution independent. A lot of software
is common to most distributions (X server, ssh server, Apache, Gnome, ...).
This common software base has caused some problems in the past (even with
Linux): due to a misconfiguration of OpenSSL (I thought, it was something
SSH <http://it.slashdot.org/article.pl?sid=08/05/13/1533212&from=rss>related
in any case) at Debian, cryptographic keys generated by computers
with Debian (and therefore also Ubuntu) were very weak and easy to crack. At
the moment that bug is fixed; but it's just an example to show that a
failure at one point can cause a lot of trouble when that part is used in a
lot of computers. Software is inherently prone to errors: a lot of functions
are used by other functions, so a security issue in one low-level function
can cause security issues in hundreds of programs.

But indeed diversity is a great way to ensure security. It's indeed harder
to find a bug that bother every single Linux computer because Linux
computers sometimes have different architectures (all Windows computers use
x86 or x86-64 (AMD64); Linux also has support for a lot of other platforms
(PowerPC, ARM, Cell, ...)). Actually the same mechanism is used by nature to
ensure the survival of species. Let's say you plant one kind of potato
plant, all more or less the same, then when a bug comes along that is
harmful to one plant, all plants could (and will) get infected. (That has
already happened a few times, e.g. Great
Famine<http://en.wikipedia.org/wiki/Great_Famine_%28Ireland%29>).
When you plant different kinds of potatoes, some plants will get infected
but others will survive. As with Linux: because of the differences, this
will get you some security. Some servers make use of this principle by not
using standard values for the installation.

e) There is no real absolute answer to this, because the security holes in
Windows do not exist in Linux and vice versa. But at the moment there has
been little succes in manufacturing large scale malware for Linux, sure,
some do exist (rootkits etc.). Hackers will mostly target Windows because it
is easier to succeed in overpowering at least a few machines. For Linux, a
lot of computers are servers (more security focussed, both on Linux and
Windows) and the defaults are quite sane (or enforced to be sane). In
Windows some defaults were/are not sane (passwords etc.), I can't say if the
defaults in Windows 7 are all sane, because I mostly choose to set them to
sane values anyway

http://www.msversus.org/microsoft-windows-performance.html

I hope this will help you a bit,

Kind Regards


Egon.



e) That depends on the software you use. Some will run faster in Linux, some
in Windows; I guess. But there is some proof that some applications run
faster in Wine (Windows API layer for Linux and Mac) than on real Windows
(but then on the other hand, Wine does not implement the Windows API
entirely and is slower in a lot of cases). In XP some programs might run
slower due to the caching differences; but in Vista and 7 performance should
be almost the same, I think. Haven't really performed any tests.

But what does make a difference is the fact that there is little to no
malware for Linux, so you don't really slow down your computer by surfing
the internet as is the case with Windows and you don't need a virus scanner,
malware scanner to keep your system running.

On Thu, Oct 28, 2010 at 11:14, Jan Bongaerts <jbongaerts at gmail.com> wrote:

> Unfortunately I won't be there again.
>
> I'm still compiling some introsession for a school, and could use some
> ideas on what to show
> 1) using the live CD
> 2) using a normally installed system.
>
> I would like to know the following background info, to answer some probable
> questions..
>
> a) Why does Linux need so little memory to run (min 384MB if I'm not
> mistaken), compared to Windows (min 1GB I think)?
>
> b) Why isn't it necessary to 'defragment' a Linux hard disk, like one needs
> to do in Windows?
>
> c)Confirm that software packages in Linux are much lighter because of the
> multi-package structure. If not, please give reason.
>
> d)Confirm that the most important safety feature in Linux is due to the
> fact that you always need a password to become root, and that the second
> most important reason is that there is little standardisation, so difficult
> to write malware that works on all flavours of the target software.
>
> e) Confirm that apps usually run faster in Linux than in Windows, because
> of the different memory management.
>
> Those are just some of the things I can think of now.
> I'd love to hear feedback from the experts here.
>
> Regards,
> Jan.
>
>
>
>
> On Wed, Oct 27, 2010 at 9:27 PM, jean7491-Events-Team <jean7491 at gmail.com>wrote:
>
>>  Hi to all,
>>
>> Remember - next IRC meeting  tomorrow Thursday 28/10/2010 at 21.00 hr.
>> on #ubuntu-be -- IRC (http://webchat.freenode.net/).
>>
>> See agenda in wiki https://wiki.ubuntu.com/BelgianTeam/IrcMeetings
>>
>> --
>> jean7491
>> Ubuntu Belgium Events Team
>>
>>
>> --
>> ubuntu-be mailing list / mailto:ubuntu-be at lists.ubuntu.com
>>
>> Modify settings or unsubscribe at:
>> https://lists.ubuntu.com/mailman/listinfo/ubuntu-be
>>
>
>
>
> --
> Microsoft programs are like Englishmen. They only speak Microsoft.
>
> --
> ubuntu-be mailing list / mailto:ubuntu-be at lists.ubuntu.com
>
> Modify settings or unsubscribe at:
> https://lists.ubuntu.com/mailman/listinfo/ubuntu-be
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ubuntu.com/archives/ubuntu-be/attachments/20101028/87e26a2d/attachment-0001.htm