Page 2 of 3
Sectors, tracks and cylinders
The geometry of data storage on a drive is very simple but goes a long way to explaining how it works.
A stack of magnetic disks are read and written by a stack of heads to allow access to a cylinder of tracks.
Each track is divided up into a number of sectors each of which stores a relatively small amount of data – usually 512Bytes.
The data is recorded on the disk as set of concentric rings called tracks. The head is moved by an actuator to a given position and then the data is just read or written as the disk spins underneath the head.
A drive can be built to read and write an entire track in one operation but it’s more flexible to divide a track up into small equal-sized chunks called sectors.
Each sector has a standard header that contains information about which sector it is and where it is on the entire disk – this is the sector header - and following this is an area where data can be stored. So to read or write a sector the steps are:
- The head is moved to the correct distance to position it over a track.
- It reads the sector headers from the disk until it finds the one it is looking for.
- Having found the start of the sector it waits until the data portion moves under the head and either reads the data or writes new data.
In practice it turns out to be more cost effective to build disk drives with multiple magnetic surfaces using a stack of disks all mounted on the same spindle.
Each disk has a top and bottom magnetic surface and a matching stack of heads can be used to read the same track on the top and bottom of each disk. This clever idea increases the amount of storage a disk provides but still only uses one motor to spin the stack and one head actuator to move all of the heads at once. The set of tracks that the stack of heads can read or write at the same time is called a cylinder, or a cylinder of tracks.
With these basic ideas of what goes inside a disk drive you can begin to understand some of the problems in making a drive fast and able to store a lot of data. What a drive can do is limited by mechanics rather than electronic considerations.
For example, the rate at which data can be read or written depends on how fast the disk moves under the head and how tightly packed the data is. You can work out the consequences using nothing more than simple arithmetic.
Consider a 3.5” disk rotating at 5400 rpm. At the outer edge the disk will be passing under the head at roughly 90kph or 50mph.
Typically a single track will store around 500KBytes of data and this means that it takes around (1/90) 1/100th of a second to read a track giving a data rate of (45) 50MBytes/s. Notice that this is the fastest possible data rate and in practice it would go down to 15MByte/s for a track at the centre of the disk which is moving more slowly.
There is also the sad fact that many of the bits transferred don’t correspond to useable data but to the headers and error detection codes used. Real data rates are generally 50% of the raw data rate calculated.
If you increase the spin speed then the speed that data can be read or written also increases. For example, 5400rpm is typical for a low cost or small 2.5 inch disk drive but for a little extra cost your desktop machine could be using a 7200 rpm drive, which gives a raw date rate of (60) 80MBytes/s.
So why not increase the spin speed even more? The answer is that there comes a point where the disk would shatter and fly apart due to centripetal force. The fastest disks currently available spin at an amazing 15,000 rpm and have raw transfer speeds close to (125) 200MBytes/s.
But remember that raw transfer speeds aren’t the same as the amount of useful data transferred. Even a top of the range 15,000 rpm drive can only read and write useful data at around 50MBytes/s.
Spin speed definitely affects how fast a disk can be read or written but there are other factors.
For example, how fast can the head move from one track to another? This affects how long you have to wait before the head starts to read or write data at a specified track. There are many different ways to measure how fast a head moves but it’s usually quoted in the time taken to move from one place to another, i.e. how long to “seek” and find a new track.
Drive specifications typically quote three seek times:
- Track-to-track seek time – the time it takes to move from the current track to the next track
- Average seek time - the average time it takes to move the head from a random location to another random location
- Full track – the time it takes to move from the first to the last track i.e. the maximum time you will ever have to wait.
For a modern drive track-to-track is down to a millisecond (1/1000s) or less. Average and full track times depend on the number of tracks that the disk has and this varies but it should be 10millisecond or less.
You can sometimes hear the heads being moved from one place to another as a rapid series of clicks. The whine due to the spinning is usually much more audible.
Once the heads have been moved to a new position there is a final delay in getting the data – waiting for the sector that you want to pass underneath. This is called the latency and it is on average equal to the time it takes the disk to rotate half a turn. Again, the faster the spin the smaller the latency.