Intro to Computer Systems

Chapter 8: Mass Storage

Attributes of Mass Storage

Mass storage is a critical part of a modern computer system; virtually all tasks require the long-term storage of large volumes of information. There have been a number of different technologies used to perform this task in computing, such as:

This topic will concentrate on magnetic, optical, and solid state mass storage.

Performance Metrics

Although mass storage technologies may differ significantly in how they go about recording and retrieving data, the issues and metrics used to judge performance are common to all: at the end of the mass storage device is a computer requesting the data, and the way the data is stored is usually of little interest to what it needs the data for.

Random vs Sequential Access

Computer systems can request data for all sorts of reasons, however they generally fall into a proportion of two data access patterns: sequential and random access.

Like computer memory, there is latency involved in getting to the data that needs to be retrieved - however with mass storage it is more commonly referred to as access time.

Read and Write Rates

Another aspect of data access that varies from one type of mass storage to another is how long it takes (and consequently, how fast) it can perform read and write operations.

Many forms of media take longer to write a bit of data, then it does to read it. This is usually due to the fact that many recordable storage systems make use of chemical or magnetic properties to identify 0s from 1s - and it takes longer to modify the state of a chemical compound or magnetic field, than it does to merely observe its current state.

Some media types, especially those based on rotating discs, may have different performance depending on the part of the disc surface; since each part of the disc is travelling at the same rotational speed (rpm), for any given unit of time, a read/write head that's over the outer edges of the platter will see more data than when it is nearer to the inner edge.

Burst and Sustained Rates

There is also a difference between what the maximum data rate the storage device can sustain for short bursts of data, versus the data rate it can sustain for long periods of time.Techniques such as data caching and optimised algorithms for operating the apparatus inside the storage device, can make very significant differences between the two.

Generally speaking, quoted peak transfer rates assume best-world conditions; e.g. a minimal need for seeking, the cache algorithms are working well, etc. This may be true for small bursts of data, but will not be the case for very long.

Sustained data rates are usually significantly lower than burst rates, and are a better reflection of the realistically achievable speed, taking into account the design (and inherent limitations) of the device.