|
At your service!
|
03/07/2008
|
| |
Everyone knows standard operating systems can’t function in real time, just as everyone knows the Moon is made of cheese. It’s a popular misconception which may have once been partially true.
Real time applications come in a variety of shapes and sizes. Providing a server which handles requests from many thousands of cash dispensers requires that average response is fast enough to prevent the customers losing patience.
Handling broadcast tv in real time requires the worst case response time to be less than the vertical scan rate (either 16.67 or 20ms). Missing a single response will cause the outgoing service to go off the air. In this case, average (as opposed to worst case) response time does not matter.
When Linux began life as a kernel for a general purpose operating system, it already had the vital feature that, once developed, would make it ideal for real time use. That feature is preemption – the ability of the kernel to stop a running task and to start another in its place. Unlike OSs initially intended for the desktop, it is not at the mercy of a task that will not let go of the processor. The kernel can pause a long running task when a real time response is needed.
Early versions of the Linux kernel were certainly not suitable for real time use, unless hacked to separate real time code from the kernel proper. It was prone to spending many milliseconds tidying up memory allocation or traversing internal data structures, only wandering back to run useful code when it had done what it thought important.
The kernel has changed radically over the last few years. It is now leaner, meaner and capable of precise real time performance.
|
| |
Author
Robert Billing
|
| |
| |
|
This material is protected by Findlay Publications copyright 2008.
See Terms and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the sales team.
|
| |
|
| Email this article |
| |
|
|
|
|