Messages in this thread | | | Date | Fri, 07 Jul 2000 11:35:19 -0700 | From | kevin morgan <> | Subject | Re: a joint letter on low latency and Linux |
| |
Out of all the jitter on this subject there is some excellent news:
- A 2.4 preemption point patch will get produced by Ingo (note: I'm no fan of preemption points! But sometimes an ugly solution is better than no solution).
- 2.5 will move to the "right" (imo) solution, namely, a preemptible kernel leveraging the SMP capabilities of Linux. This is fantastic news, and should address the core need stated in the open letter (albeit not in 2.4).
I believe that under many important environments, such a system can actually have better throughput than a non-preemptible kernel based system (in contrast to those who believe responsiveness and throughput are fundamentally at odds in OS design).
MontaVista is working to measure the preemption delays in a prototype uniprocessor preemptible Linux kernel. We'll share results when we have them. If anyone else is working on this we'd be interested in collaboration.
As for all the debate around when and where and how RTLinux/RTAI solutions should be used...from a vendor's perspective, the customer is usually right, and our customers are requiring both RTLinux product from us, and "native Linux responsiveness" ala Ingo patches as product from us (usually not both from the same customer). We just explain the hows/whats/pros/cons of each and let them decide how to architect their solutions.
Long term, imo, "standard" Linux must have better user level process responsiveness under load to become a truly ubiquitus operating environment for modern desktops, servers and devices. Sounds like 2.5 and beyond should meet the need for the most part, and that's very encouraging.
-Kevin
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |