[aclug-L] Re: Linux as an alternative to a Windows desktop

[Nate Bargmann <n0nb@xxxxxxxxxxxxxxxx>]

* Jonathan Hall <flimzy@xxxxxxxxxx> [2002 Dec 16 19:48 -0600]:
> That is noticable not because of the location of the swap file in relation
> to the disk, but in relation to other data accessed on the disk.
> 
> A hard disk spins at a constant speed, so unless you store different amounts
> of data on different cylinders, the access time should be identical on the
> inner and outter tracks.

Now this thread has taken an interesting tangent.

Just sitting here thinking about it, I wonder if there have been any
studies or if empirical data exists regarding the reliability of the
inner vs outer tracks.  In theory, at least, the inner tracks *should*
have less magnetic material per unit of data than the outer tracks,
which would make the inner tracks prefered for things like swap and
other temporal data in my mind.

> What takes time is moving the heads, which happens when you place your swap
> partition away from the rest of your data.

Precisely.

Again, I wonder what data exists regarding the optimal positioning of
the swap partition.  Adjaceant to /usr?  Doubtful since /usr is usually
quite static.  Adjaceant to /var?  Probably best on a web server where
system web files and database files likely exist (at least on Debian).
Adjaceant to /home?  This might make sense on a workstation with a large
amount of private data in /home.  Perhaps this would make sense on a
remote login server as well.  Interesting.

> In practice, it's usually easiest to put your swap partition "in front of"
> your most heavily used data partition.  For this reason, I usually put /,
> swap, then /usr.  Because filesystems generally fill up from front to back,
> most of my most heavily used data will be at the front of /var, and thus
> close to the swap partition, so moving the heads between these two areas of
> the disk should not take long.

Looks like we agree.

> In theory, if you had a back-to-front filling filesystem, it would make
> sense to put your swap partition at the end of it--perhaps on the outtermost
> cylinders of the drive.

Of course the optimal placement of the swap partition seems to depend on
the type of data the kernel swaps out.  On this workstation my
assumption is that it is binary code for applications that haven't been
active for some time.  So does it really matter where on the disk the
swap physically is unless it contains a lot of dynamic data as on a web 
server?

Beats me.

- Nate >>

-- 
 Wireless | Amateur Radio Station N0NB          | "We have awakened a
 Internet | n0nb@xxxxxxxxxxxxxxxx               | sleeping giant and
 Location | Bremen, Kansas USA EM19ov           | have instilled in him
  Amateur radio exams; ham radio; Linux info @  | a terrible resolve".
             http://www.qsl.net/n0nb/           | - Admiral Yamamoto
-- This is the discussion@xxxxxxxxx list.  To unsubscribe,
visit http://www.complete.org/cgi-bin/listargate-aclug.cgi