3.3 General Notes on Consumer Devices and Embedded Systems
Over the past thirty years, the performance of computer hardware has doubled approximately every 12 to 24 months. This has led to unprecedented miniaturization and widespread usage of computer technology. The price of the microprocessor in today's personal computers is only a few cents per megahertz, and the price is still decreasing rapidly. Today's small wireless devices such as cell phones and two-way pagers have microprocessors that are similar to those that were commonly used in high-end desktop computers less than fifteen years ago.
Ultimately, computing power and storage capacity will become extremely cheap. Inexpensive devices, home appliances, and consumer products such as microwave ovens, shopping carts, wallets, and toys will include powerful processors and megabytes of storage memory.
Thanks to Moore's Law, software developers for desktop computers and servers haven't had to worry very much about processing power or efficiency of their software. Many performance problems have been resolved simply by waiting for a more powerful computer to become available.
However, the consumer device space and embedded software development are different. Even though the processing power of consumer devices and embedded systems is increasing, the majority of these systems are still very resource-constrained and limited in processing power. And, unlike in the desktop and enterprise computing area, the situation is not likely to change dramatically in the next few years. There are several reasons for this:
Moore's Law does not apply to the battery.
Most consumer devices are battery-operated. Even though battery technology has advanced substantially in the past years, the advances in battery technology are not as dramatic and predictable as in microprocessor development. Low power consumption is very important in wireless devices, because 75 to 85 percent of the battery power typically needs to be reserved for the radio transmitter and RF signal processing. This places severe power constraints on the rest of the system.
High-volume production.
Typically, consumer devices such as cellular phones are manufactured in extremely large quantities (millions or even tens of millions of units). They are sold to price-conscious consumers at prices that are very low, often subsidized. To improve their profit margins, device manufacturers want to keep the per-unit costs of the devices as low as possible. Additional processing power or precious dynamic memory will not be added unless the consumers are willing to pay for the extra capabilities.
Specialized nature of devices.
Consumer devices are typically highly specialized. Each device is usually highly customized for its intended usage: cell phones for voice communication, pagers for alphanumeric message retrieval, digital cameras for photography, and so on. Again, to keep the price of the devices reasonable, device manufacturers are not willing to add general-purpose features and capabilities that would raise the cost of the device, unless the features are well-justified from the viewpoint of the target market and the consumer.
In general, a fundamental difference between desktop/enterprise computing and consumer devices is that in the consumer device space one solution does not fit all. Unlike personal computers, which can run tens of thousands of different kinds of applications, and are not customized particularly well for any single application, consumer devices are typically highly specialized for their intended use and are designed to do only one or a few things well. This means that unlike desktop and enterprise computers, whose performance is typically easy to measure in terms of their processing power, the performance and value of consumer devices and embedded systems is measured in terms of how well they serve their intended usage. The specialized nature of the devices has to be kept in mind when developing software for these devices.
There are further factors differentiating the consumer device space from desktop computers. These include:
Small screen size.
The majority of consumer devices have miniscule screens compared to desktop computers or servers. Some devices have no screens at all.
Different usage models.
The usage models of small devices vary considerably. Some are stylus-operated, some have tiny phone keypads operated with one hand only, some have small QWERTY keyboards operated using two thumbs, some are voice-operated/hands-free, and so on.
Mobility.
Compared to personal computers, consumer devices are operated in varying and sometimes unusual circumstances, for example, while in traffic, while roller skating or skiing, at night, on a beach, in a snowstorm, and so on.
Limited network bandwidth with intermittent connections.
Personal computer users have become accustomed to relatively reliable, rather high-speed networks based on TCP/IP and other well-established network protocols. In contrast, many wireless devices have intermittent, low-bandwidth network connections.
In summary, the rules that apply to desktop and enterprise software development do not generally apply to consumer devices and embedded software development. Even though from the historical viewpoint it would be quite easy to predict the future of small devices to be similar to the past history of personal computers, it is by no means obvious that the history will actually repeat itself in this context. Rather, at least for the time being, a fundamentally different mindset is required to address the peculiarities of consumer devices, embedded systems, and the consumer market.
|
