Thread: Hidden secrets of modern data storage

MOSt of us don't give our hard drives a second thought; after all, it's just something on which to store files and photos. They generally only come to our attention when they start irritating us by making noises or driving us to despair when they
fail and lose countless hours of important work. But have you ever wondered how they work or what's inside them? In this feature we will explain how these mass-produced yet precision engineered products came about.

A brief history
The hard disk has been around for more than 50 years now. The grandfather of all hard disks was IBM's 350 storage unit (or 'disk file' as IBM called it), used in the 305 Ramac computer in 1956. It used 50 24in magnetic disks (platters) to hold five million 7-bit data characters and had one head for reading and writing data. But it was in 1973 that IBM introduced the technology on which most of today's drives are based. The 3340 or, to use its more popular name, the Winchester drive, introduced low-mass heads, lubricated disks and a sealed assembly. It had a capacity of 35MB or 70MB via two or four 14in disks.

1980 saw the first 5.25in full-height drive, the Seagate ST-506, which was the first hard drive for personal computers and held 5MB of data. Three years later Rodime introduced the 3.5in half¬height form factor that we all know today, with the R0352 which held 10MB of data on two 3.5in platters. 1988 saw the first low¬profile 3.5in drive, the Conner Peripherals CP3022, which had a capacity of 21MB on a single 3.5in platter. This form factor became the standard for modern drives. The first 2.5in drive appeared the same year, when a company called Prairie Tek launched the 220, storing 20MB on two 2.5in platters.

The start of the 1990s saw IBM bringing three important advances in the way drives access data. The IBM Redwing, an 857MB drive launched in 1990, was the first to use magneto¬resistive (MR) heads and a type of data decoding called PRML (partial response maximum likelihood). A year later IBM's Pacifica mainframe drive was the first drive to replace the magnetic oxide medium on the platter surface with a thin film medium.



The first 10,000rpm drive arrived in 1997 with Seagate's launch of the STl9101 Cheetah 9, which used eight 3.5in disks to provide 9.1GB of capacity - and a year later they had the first 10,000rpm drive with 3in disks, the Cheetah 18 (ST1l8202), which used 12 disks for its 18.2GB capacity.

The same year saw Hitachi bringing out an even faster drive, the DK3El T-91, which had a spin speed of 12,000rpm, and in 1999 IBM launched the lin Microdrive.
The turn of the century saw Sea gate reclaim the fastest spin speed crown with the Cheetah X15, rotating at 15,000rpm with a capacity of 18.3GB.

As well as capacities, drive interfaces improved, with the parallel ATA standard becoming faster, and then in 2003 the Serial ATA Working Group published the serial ATA (Sata) l.Oa specification. It's become well established, and now the drives in most new systems use the Sata interface.

Recording technology has changed too, with the first 'perpendicular recording' commercial drive appearing in 2005 in the form of a l.8in drive from Toshiba, but Seagate was first to the market with 2.5in and 3.5in drives (both in 2006) using the technology. It's this technology, which allows for closer packing of bits on the magnetic media, that enabled Hitachi to launch the first terabyte (1 TB) drive in 2007. Last year also saw the advent of commercially available Flash-based solid state drives, Hard drives work by storing and accessing data in a similar way to magnetic tape. On a disk, the data is stored by changing the magnetic polarity of small portions of its surface, which is coated with grains of cobalt-platinum. Think of a bar magnet - it can be pointing in one of two directions, with the north pole facing in the direction that disk is rotating, or against it.

In the most recent drives, the direction of magnetisation is perpendicular, so the north pole is either facing towards or away from the surface of the platter.
The direction of magnetisation indicates whether a binary '1' or a '0' is stored in that bit; the disk heads either apply a stronger magnetic field to change the polarity of a bit on the surface, when writing data, or detect the current state to read it.

Each platter has a head moving over its upper surface, and all the heads in a drive are moved by the same actuator - effectively accessing a cylinder at a time, across all the platters. The heads have to be extremely close to the surface in order to read or write the data, around 0.0003mm. To get that close to the disk surface without physically touching it, they ride on the air cushion created by the spinning discs. The only time the heads physically touch the discs is when the drive is switched off or if something drastic happens, say if the drive is dropped.

Most modern drives park the heads on a section of the disk reserved for that purpose, called the landing zone. To read or write data, the controller has to wait until the appropriate sector passes underneath the disk head. The head position is monitored to check it's correct, and any errors are fed back in a closed loop system, ensuring that the drive stays correctly aligned.

The heads themselves are built from magneto¬resistive materials - this means their electrical resistance changes depending on magnetic fields near them, so as the disk moves underneath the head the resistance reflects the pattern of Os and Is stored. When data is written to the disk a current is applied to the heads, which creates a magnetic field that aligns the polarity of the part of the disk's surface below the head.

Data is stored in sectors and tracks (put simply, tracks are concentric circles, which are divided along their length into sectors). Both of these are established by low-level formatting, when the starting and ending points of each sector
are written.

A sector contains a fixed number of bytes, usually 512 or 256, and these sectors are grouped together to form clusters. The actual file-storage operations and details such as the allocation of space are taken care of by the high-level formatting and the operating system.