What are the requirements of realtime operating systems?

Answer 1

Technically, all operating systems are "realtime." The term realtime just means it happens immediately, versus being delayed. Most operating systems today are actually a blend of realtime and "batch." Batch processing is doing something later, usually waiting until you have more than one thing (a "batch") to do. An example of batch processing is when a company processes your payment. They don't post each payment to their computer system - they wait until all the payments are in for the day (or at least a good number of them) and then post them as a group - so thousands of payments get posted at one time (posted just means processed through some financial system). Some businesses, such as banks, will use both realtime and batch processing for the same items. For example, if you make a deposit with a teller, you will probably be able to go to Starbucks next door, sign on the internet, and see the the deposit posted to your account already. This is "realtime" processing. However...this is probably only a "memo" post. This means it was only posted to the online system so that you can get to it through the internet or telephone. During the night, your deposit will be sent to a central site and batch processed with thousands of other deposits. The night post will be to the "real" posting system for checks, the "Demand Deposit Accounting" system, or DDA. The biggest use "realtime" operating systems is actually in what is called "embedded" systems. These are operating systems and application code that is running on a single board and embedded in something physical - a car, a toaster, a TV set, a DVD player, an iPod. Today, if it is electronic, it probably has an embedded system (or even several). Embedded systems are primarily realtime operating systems - they collect some statistics in realtime and allow access to them later in batch mode (here are all the errors that occurred on your car since it was last serviced...) but their real job is to control the device they are embedded in - and do it in "realtime" or immediately. This means they must be: - dependable - they must operate in all sorts of conditions - hot, cold, dirty, greasy, noisy, vibrations etc. - timely - they must be fast enough to think about an action, and get it done right away. For example, if you are an embedded system in a car that controls skidding, you must be able to react to the skid in "realtime" - as it is happening, and control brakes, steering and gas as appropriate to what is happening (skid starting, car is curving to right, skid is ending). - flexible - in the previous example, the embedded system to control skids must be able to handle a car going any speed, curving right, left or going straight, know when it is "over compensating" and correct it etc. It must be able to make the same decisions a knowledable human would make in the same circumstances. - sized properly - again, the previous example. The embedded system must be able to recognize and react to thousands of varying facts (speed, direction, condition and status of mechanical systems etc.). You must have a processor fast enough to process the data in realtime, and enough storage to accept, and display, conditions and status of "analog" devices. Note - most realtime systems are in what we know as "analog" devices. This just means they have something mechanical about them. Most parts of a car are analog, as are most TV set these days. This is versus "digital", where parts don't move, signals flow to devices that are able to convert the final result to analog (for instance, a digital TV will receive a broadcast signal, convert it to a certain channel and pass it on to the screen to display. Even screens such as plasma and LCD are ultimately analog - they must decide which dots (pixels) to display, in what color and what shade. The decision can be made digitally, controlling the nuances of the display in realtime.