Size of an embedded Linux system

The size of an embedded Linux system is determined by a number of different factors. First, there is physical size. Some systems can be fairly large, like the ones built out of clusters, whereas others are fairly small, like the Linux wristwatches that have been built in partnership with IBM. The physical size of an embedded system is often an important determination of the hardware capabilities of that system (the size of the physical components inside the finished device) and so secondly comes the size of the components with the machine. These are very significant to embedded Linux developers and include the speed of the CPU, the size of the RAM, and the size of the permanent storage (which might be a hard disk, but is often a flash device—currently either NOR or NAND, according to use).

In terms of size, we will use three broad categories of systems: small, medium, and large. Small systems are characterized by a low-powered CPU with a minimum of 4 MB of ROM (normally NOR or even NAND Flash rather than a real ROM) and between 8 and 16 MB of RAM. This isn’t to say Linux won’t run in smaller memory spaces, but it will take you some effort to do so for very little gain, given the current memory market. If you come from an embedded systems background, you may find that you could do much more using something other than Linux in such a small system, especially if you’re looking at “deeply embedded” options. Remember to factor in the speed at which you could deploy Linux, though. You don’t need to reinvent the wheel, like you might well end up doing for a “deeply embedded” design running without any kind of real operating system underneath.

Medium-size systems are characterized by a medium-powered CPU with 32 MB or more of ROM (almost always NOR flash, or even NAND Flash on some systems able to execute code from block-addressable NAND FLASH memory devices) and 64–128 MB of RAM. Most consumer-oriented devices built with Linux belong to this category, including various PDAs (for example, the Nokia Internet Tablets), MP3 players, entertainment systems, and network appliances. Some of these devices may include secondary storage in the form of NAND Flash (as much as 4 GB NAND Flash parts are available at the time of this writing; much larger size arrays are possible by combining more than one part, and we have seen systems using over 32 GB of NAND, even at the time that we are writing this), removable memory cards, or even conventional hard drives. These types of devices have sufficient horsepower and storage to handle a variety of small tasks, or they can serve a single purpose that requires a lot of resources.

Large systems are characterized by a powerful CPU or collection of CPUs combined with large amounts of RAM and permanent storage. Usually these systems are used in environments that require large amounts of calculations to carry out certain tasks. Large telecom switches and flight simulators are prime examples of such systems, as are government research systems, defense projects, and many other applications that you would be unlikely to read about. Typically, such systems are not bound by costs or resources. Their design requirements are primarily based on functionality, while cost, size, and complexity remain secondary issues.

In case you were wondering, Linux doesn’t run on any processor with a memory architecture below 32 bits (certainly there’s no 8-bit microcontroller support!). This rules out quite a number of processors traditionally used in embedded systems. Fortunately though, with the passage of time, increasing numbers of embedded designs are able to take advantage of Linux as processors become much more powerful (and integrate increasing functionality), RAM and Flash prices fall, and other costs diminish. These days, it often makes less economic sense to deploy a new 8051 microcontroller design where for a small (but not insignificant) additional cost one can have all the power of a full Linux system—especially true when using ucLinux-supported devices. The decreasing cost of System-On-Chip (SoC) parts combining CPU/peripheral functionality into a single device is rapidly changing the cost metrics for designers of new systems. Sure, you don’t need a 32-bit microprocessor in that microwave oven, but if it’s no more expensive to use one, and have a built-in web server that can remotely update itself with new features.

Source of Information : OReilly Building Embedded Linux Systems