Rotational speed

The average rotational speed on current technology disk drives is 3,600rpm, although newer high-performance drives offer 4,400rpm and even 5,400rpm. Rotational speed has a direct bearing on latency and data transfer rates.

Latency is the time taken for a certain desired sector of the disk to be positioned underneath the read/write head. Average latency is equivalent to the time taken for half a revolution of the disk: for drives with a rotational speed of 3,600rpm, therefore, latency is around 8 milliseconds.

The data transfer rate of a disk drive is determined not only by the rotational speed, but also by other factors such as head and media design. As a general rule, though, current drive technology employing a rotational speed of 3,600rpm will support a data transfer rate of 3MBytes/s and a 5,400rpm drive will enable 4.5MBytes/s.

Developments in read/write circuitry will certainly facilitate faster speeds, however this will also necessitate parallel developments in bearings and spindle motors. Additionally, as the consequences of head touch increase with faster speeds, it will also require much lower mass sliders with improved aerodynamics and geometry to maintain a stable flying height.

Finally, with disks rotating faster, reads and writes also have to be performed faster if there is to be no loss of track capacity: therefore, the read/write frequency has also to be increased.

Seek time

One of the primary elements affecting seek, or positioning, times is the actuator whose task it is to move the read/write head to the desired position of the disk. There are two basic types of actuator: linear which moves radially to different positions, and rotary which pivots like the arm of a phonograph.

With radial movement the head is always kept at a tangent (90 degrees) to the platter, but it has a high inertia requiring proportionately higher power consumption to maintain response levels.

The most commonly used type is the rotary actuator which, conversely, has a lower effective inertia (and consequently lower power consumption), enabling tracks to be indexed more quickly. This arc-type movement also gives the advantage of greater balance and precision, which in turn means improved resistance to shock and vibration.

Manufacturers use different methods to determine average seek times: Fujitsu's figure, for example, is the average time taken to perform all possible seeks; some other manufacturers quote figures based on a proportion of the maximum seek time.

Servo system

In most of today's hard disk drives, the actuator is moved by a voice coil motor upon receiving signals from the controller on input from the host system and from the servo system which constantly supplies information on the head's current position. This information is written on the disks surface during the production of the drive. Servo performance has a direct bearing on track densities and positioning times.

There are three common servo referencing methods.

Dedicated Servo System Dedicated servos use a "closed loop" system, in which a separate, dedicated head and surface is used to constantly monitor positioning information: this provides the ability to immediately sense and correct off-track situations (e.g. as in the case of shock). As the read/write heads do not spend time processing this information, it also leads to faster seek times.

Embedded Servo System Embedded servos, on the other hand, use an "open loop" system, which samples rather than constantly monitors positioning information: this is achieved by placing the servo information in between the data blocks. This provides fine-tuning information, enabling higher track densities and ultimately higher disk drive capacities. The embedded servo is more cost efficient, but slower than the dedicated servo.

Hybrid Servo System As the name suggests, the hybrid servo system combines the above two systems by having a separate servo head and surface plus embedded servo information for each data head. This provides the best of both worlds: fast seek times and the high track densities.

Cache

The term "cache" has been used for many years to describe most disk related buffer memory applications. More recently it has been used to define the intelligent use of memory to improve the efficiency of disk to host I/O rate.

Read-ahead cache

In the majority of business applications, it is normal to transfer multiple blocks from disk to host. Therefore, after the first sector has been located and is being transferred from disk to host a pre-determined number of successive blocks are also read (hence the term read-ahead), but instead of being transferred directly to the host are held in cache memory. If the next host request is for a successive block it will be read directly from cache, not disk, thereby minimising the disk service time. If the subsequent host command is for a block not held in cache then the cache register is normally flagged as such (so that it can be overwritten) or flushed.

The effectiveness of the cache for any specific applications is referred to as the "hit ratio" and is the ratio of successful reads from cache compared to the number of requests. The caching parameters need to be variable to allow the optimisation of cache to specific applications. In general terms, large contiguous or sequential files benefit the most from caching, whereas small, random files benefit the least.

Segmented cache

In applications where the disk is receiving multiple commands/requests (e.g. multi-user and multi-tasking environments), there is an obvious advantage in being able to segment the cache so that each command can be given its own working area of memory. Some segmented caches come predefined, while others can be tailored by the user, in terms of both the number and size of memories within the cache. Most segmented caches are also read-ahead caches, so offer the same advantages described above.