Partitions: What are they and what are they good for?
Partitions are synonymous with Hard Disk Drives (HDD) for
PCs. All hard drives have them, but there seems to be a fair amount of confusion
about what they really are and what they can be used for.
Background
Before we can understand partitions
and how they work, we need to establish some concepts and terms related to the physical
structure and operation of the hardware.
HDDs contain two elements related to data storage and retrieval; platters
and heads.
Each platter has two sides, a head is required for each side.
The read/write process is very similar to that used for magnetic tape, such as cassette
tapes. Cassette tape is coated with magnetic material, and the heads are mounted in the
cassette deck. There are a total of four 'heads' usually contained in a single metal
structure. This construction provides for four 'tracks' on the tape, two for each
direction. This read/write mechanism is designed so that, when you flip the tape over, the
heads are positioned over the two tracks for the other direction.
On a HDD, however, the magnetic material is coated on the platters. In place of the
linear motion of the tape moving across the heads, the platters of a disk drive are
rotated so that they move under the heads. After low-level
formatting of the magnetic media, each head (for each platter) has its own track rotating
underneath it. The heads are mounted on an access
arm that moves, in and out, to access multiple tracks on each recording surface (the
platter sides). Each track holds multiple fixed-length records called Sectors. In
what I believe to be a carry-over from the early floppy days, a sector is normally 512 bytes.
 In order to properly
visualize the addressing structure used to locate specific records on a HDD, we need to
introduce some more concepts. For a single position of the access arm, all of the heads,
and therefore the tracks, are lined up vertically. Because the platters are round, this
corresponds nicely to the three-dimensional mathematical concept of a cylinder. This
accounts for all of the area under the heads at that time. Each different position of the
heads forms a cylinder with a slightly different radius. This leads to the concept of a
cylinder address, which specifies which of the possible positions of the access arm is to
be used.
Within each cylinder we have multiple heads, each with its own surface to record on. So
the second part of our addressing scheme is to specify which of the possible heads is to
be used.
The final piece to the puzzle is to determine which record or sector to use within the
track by using the sector (record) number.
Putting all these together, we have the C-H-S (Cylinder-Head-Sector) form of addressing
a HDD used by employing interrupt 13
(int 13h), the BIOS
disk access routine. With those three values you can locate any specific sector on the
disk.
The CHS
addressing scheme worked fine when the size of the media was small (at one time a 10MB HDD
was the largest available). Advances in hardware fairly quickly exceeded the capacity of
the CHS schema within the constraints of the sizes initially chosen as limits for the
values. The initial data structures defined to hold CHS addresses were 10 bits for the
cylinder address (0-1023), 8 bits for head address (0-255) and 6 bits for sector address
(0-63). This gives the maximum size of a disk under C-H-S as:
(64 sectors x 256 heads x
1024 cylinders x 0.5K bytes/sector) / 1024 KB/MB = 8192MB (8GB)
All is not lost however; there is an alternate method of addressing disks. Since we
know that the disk is made up of fixed length records and the controller for the drive can
perform calculations, we can also simply number the sectors on the disk sequentially, and
leave it up to the drive controller to convert it into the appropriate C-H-S address for
the attached drive. This addressing scheme is known as Logical Block Address or LBA and
was introduced with the Int 13h Extensions. To take advantage of LBA, support for these
extensions is required in the system BIOS and the operating system. The maximum number of
'blocks' that can be maintained is still a constraint. However, 512 byte records are quite
small and disks can have a very large number of them. This leads us to the next concept:
clusters.
A cluster
is the smallest unit of allocation, that is, the smallest amount of disk space that can be
allocated to a file. If you extend the concept backwards, the original cluster size was
512 bytes--the same as a single sector size. By grouping sectors into clusters, we can
reduce the size of the numbers that the file system has to keep track of for large disks.
Before we go any farther, we need to make a distinction between the physical data
storage on the disk and its organization and use. The physical data only has meaning
within the context of its structure. Within a partition, this structure is known as a file
system, and there are a lot of them (see Bob's treatise on partition types in the Arcanum). The hardware
is still going to be dealing with cylinders, heads and sectors. The concept of clusters
and allocating space to files belongs to the file system.
MBR and Partition Tables
Following the use of a partition utility to partition a hard drive, the very first
sector of a PC HDD is assigned a special use, and a special name, the Master
Boot Record or MBR. For our purposes the part of the MBR we're concerned with is the
Partition Table.
The partition table is a fixed-length data structure that describes the partitions on
the disk. It can contain a maximum of four 16-byte partition entries. Each entry contains:
- the partition type
- a flag to indicate if this is the 'active' partition (there can be only one active
partition)
- the starting and ending addresses of the partition for C-H-S drives or the starting
sector and number of sectors in the partition for an LBA drive.
Most all of today's drives will be LBA. All of the partitions defined in the partition
table of the MBR are, by definition, primary partitions.
By now some of you are probably thinking, "Only four partitions? That seems like
an awfully small number" or "I know I've built more than four partitions on a
drive". Well, that's sort of true. What solves those two quandaries is one of the
primary partition types, the Extended partition. The Extended partition type is a special
case that is designed solely to contain logical drives. One Extended partition can contain
many logical drives. Although the partition table can contain up to four partition
entries, Win9x software will only 'see' two primary partitions on a given HDD, one active
primary partition (the boot drive) and one DOS extended partition which can contain
multiple logical drives. Under Windows XP, disk management functions will see the hidden
partition, but it will not be an available as a disk and will not be assigned a drive
letter.
Boot Partition Size Limit
It is unlikely that anyone will run into this boot partition size limit with current
hardware and software. If your system BIOS does not support int 13h extensions, any
partition that will be used to load (boot) an Operating System must be contained within
the first 8GB of the drive. This limitation is imposed due to the facilities available at
the time the OS loads. At that time disk access is limited to what is provided by
int 13h. Without the int 13h extensions, and support for them in the OS loader, this
limits the size of the boot partition to 8GB.
What are Partitions Good For?
Now that we've got a fairly good understanding of what partitions are, how can we
use them effectively? That's going to be difficult to answer. Each person is going to have
different needs that will affect their choice of a partitioning scheme. I can only discuss
the reasons I have chosen for my partitioning strategy.
Do you need partitions? The easy answer to that is yes, you can't use a HDD without at
least one primary partition. And if you buy a PC 'off the shelf' it will likely only have
one primary partition. You'll have a C: drive that encompasses the entire HDD. For most
users this is probably fine.
Probably the most popular reason for having multiple partitions is to allow the
installation of multiple Operating Systems, or multiple versions of the same operating
system, on a machine without having them interfere with each other. This is known as a
multi-boot configuration and it can be configured in many ways. The most basic is to do it
manually. You use a partitioning tool to select which partition is marked 'active' and
thus, which OS boots up. There are also several 'boot managers' that can be used to
display a menu of your choices at each startup so you can pick the one you want each time.
Segregating data is another good use of partitions. By isolating your data from the OS
and other programs you provide a more conducive environment for a 'Nuke & Pave' (well, for the Pave part
anyway, if you use the debug script to remove the partition table, no partitions on that
drive will survive.)
With established partitions, the OS and programs can be wiped out and replaced with a
fresh installation without affecting the data on the other partition(s). Instead of having
to do a number of restores to replace your data you simply need to inform the re-installed
products where your configuration and data files still reside.
Partitions, and some partition management software, can also be handy for recovering
from system modifications gone bad. A Win9x-based system's inability to see multiple
primary partitions can be exploited to provide a quickly available backup of a running
system. Prior to making the changes, the currently active partition is copied to another
'hidden' primary partition. In the event the system is unable to boot following your
changes, you use your partition management software (on floppies) to mark the previously
hidden partition as active and you are back in business. I haven't experimented yet to see
how XP handles swapping active partitions while it isn't looking. I'm not sure you can use
this trick under XP. Although the drive letters straighten themselves out in the BIOS, the
statements in boot.ini reference the partition(s) by physical number which does not
change. This means that the statement in boot.ini would have to be changed to correctly
reference the partition you now want to boot from. For example, the statement to boot from
the first physical partition on the disk would be something like:
multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Microsoft Windows XP
Professional" /fastdetect
If you then 'hide'
that partition and 'unhide' the second physical partition on the disk, you would need to
change that to:
multi(0)disk(0)rdisk(0)partition(2)\WINDOWS="Microsoft Windows XP
Professional" /fastdetect
Drive Letters
The assignment of drive letters for a DOS or Win9x system is fixed. The boot drive
will be the active primary partition on the primary master drive (in an IDE system) and it
will be referenced as drive C:. If there are any other HDDs that contain primary, non DOS
Extended partitions they will be assigned consecutive drive letters. For example if you
have all four IDE devices populated with HDDs containing a primary partition, the primary
master will be drive C:, the primary slave will be drive D:, the secondary master will be
drive E: and the secondary slave will be drive F:. Once all primary partitions have been
accounted for, we go back to the beginning and look for DOS Extended partitions containing
logical drives. Continuing our previous example let's assume each drive also has a DOS
Extended partition on it. The primary master has two logical drives defined in its
extended partition so they become drives G: and H:. The primary slave only has one logical
drive in its extended partition so it becomes drive I:. The secondary slave has three
logical drives, which become drives J:, K: and L:. The secondary slave only has one
logical drive, M:.
This assignment method can be very confusing if you add to, or change, your HDD
configuration. Many drive letters will change which tends to confuse installed products
and users. One method of reducing the drive letter changes is to only define a primary
partition on the physical boot drive. The physical boot drive is the only one that is
required to have an active primary partition; it is perfectly legal to have Extended
partitions only on all other disks. The only drawback that I'm aware of is the loss of a
small amount of disk space. The defined Extended partition will start with logical
cylinder 1 so one cylinder's worth of space will not be usable on the drive (for an IBM
40GB drive, you lose 7.9MB).
A summary of Microsoft OS Partition Information can be found here. |
Forums 
