[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RAID Sizes (was Re: Why do people in the UK put a u in the word color?)



On Mon, Apr 17, 2006 at 07:32:50AM -0700, Willie Wonka wrote:
> Andrei Popescu wrote:
> > Willie Wonka <floydstestemail@yahoo.com> wrote:
> > 
> 
> > > In this example, I'll use [Sector=512Bytes] and [Track=4096Bytes =
> 8
> > > Sectors].
> > > Data (File) that occupies more space than 1 sector (512Bytes), will
> > > fill up those sectors until the Track/Block/Cluster (8 sectors) is
> > > full, ...and a larger File will then  overflow onto the next
> > > Sectors/Track, and so on -- this is merely a consequence of
> > > *contiguous* writing of data.
> > 
> > You can't mix tracks and sectors with blocks/clusters. The former are
> > physical 'units' while the later are logical.
> 
> I think I'll leave this part of the topic alone for now, since I need
> to brush up on my understanding of the 'physical' (CHS) vs 'logical'
> (LBA) differences, but indeed a *Track* in Linux seems to contain 63
> sectors, as noticed again using 'hdparm'

The cylinder/track/sector used to make sense in the ancient days of DOS 
floppy disks.  The addressing technique was build into the hardware, and 
into the software.  I can still remember programming with data 
structures containing cylinder/head/sector numbers.

But hardware improved, eventually its capacity exceeded the old 
geometrical mode, so they had to fake it.  Software still accepted the 
CHS model, so the hardware had to, too, even though it became 
increasingly uncoupled form the physical layout.  Numbers like '63' 
became a codeword that indicated, ignore this number as meaningless in 
terms of disk geometry.

Then came linear block addressing, which officially accepted the 
demise of the geometry as a programming model.

-- hendrik

> 
> ~$ sudo hdparm -I /dev/hda
> ....
> 
>  Configuration:
>          Logical         max     current
>          cylinders       16383   65535
>          heads           16      1
>          sectors/track   63      63
>          --
>          CHS current addressable sectors:    4128705
>          LBA    user addressable sectors:  160836480
>          LBA48  user addressable sectors:  160836480 
> ....
> 
> > > Cylinders are ring-shaped, vertically aligned areas of the HDD -
> think
> > > of stacking doughnuts or rings; one on top of each other, the only
> > > difference (besides the obvious), is that no 2 stacks of
> > > cylinders/doughnuts/rings are the same physical size...yet they are
> > > stacked vertically (according to the platters). This all starts to
> get
> > > real *funky* once you start using LBA, instead of *phsyical*
> address. 
>  
> > And a track is one dough-nut. And because in reality the radius of
> the
> > dough-nut and hence also its length, the number of sectors/track is
> > variable. But the OS doesn't see this. The numbers are converted
> > inside the HDD logic and passed to the BIOS/OS as if the number of
> > sectors/track is constant. Otherwise a C/H/S address would make no
> > sense to the BIOS/OS.
>   
> I'll accept that info for now...  thanks;
> I'll digest it over time, and research a bit more, before again
> addressing this sub-topic ;)
> 
> > > > The smallest physical unit is the sector which is always 512 B.
> > > > When you format a partition you divide it in allocation units. In
> > > *nix
> > > > they are called blocks, in MS clusters. 
> > > 
> > > Yes, I concur; 
> > > but I'd refine it to *a group of sectors, which has a set size*
> > > perhaps.
> > 
> > and that size is always 2^x * 512B where x is a positive integer
> value
> > (zero allowed). How big it can get depends on filesystem limitations.
> 
> Yep ....Ok
>  
> > Bye
> > Andrei
> 
> I appreciated this dialog/dialogue :-)
> All I can think of now, because I'm hungry is
> (donuts/doughnuts/dough-nuts).
> 
> Regards
> 
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around 
> http://mail.yahoo.com 
> 
> 
> -- 
> To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org 
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
> 



Reply to: