Difference between revisions of "Real Time Tech Zone 2013"

From eLinux.org
Jump to: navigation, search
(Presentation notes)
(Presentations)
 
(17 intermediate revisions by 4 users not shown)
Line 4: Line 4:
  
 
<nowiki>[[My New Page]]</nowiki>
 
<nowiki>[[My New Page]]</nowiki>
 +
 +
[[Real_Time_Tech_Zone_2013#Problems_and_Issues]]
  
 
Save this page/section and your new page link will appear in Red.  Click it and begin editing your new page.
 
Save this page/section and your new page link will appear in Red.  Click it and begin editing your new page.
Line 23: Line 25:
 
=== Problems and Issues ===
 
=== Problems and Issues ===
 
What issues do you have with "Real Time"?
 
What issues do you have with "Real Time"?
 +
 +
*  The primary issue that I have with the real-time patch is why its still a patch and not in mainline.  We seem to be asymptotically approaching inclusion.  But as with any asymptote, you never seem to reach zero.  TG and SR and others are working to keep the patch caught up with reasonably recent kernels, but that's a lot of work that detracts from making further progress on the core issues of determinism in Linux.
  
 
What are some possible solutions?
 
What are some possible solutions?
 +
 +
*  The best thing would be to finally round the corner on the inclusion of the real-time patch into mainline.  I realize that there differing views on the future of Linux.  But, since the RT patch has seemingly minimal impact on mainline if it's not enabled, maybe 2013 will be the year that it finally makes mainline?  One can only hope.
  
 
What's going on in the wider community to address these issues?
 
What's going on in the wider community to address these issues?
 +
 +
*  Stephen Rostedt had a good presentation on the RT patch and where it stands in his mind.  Check out his presentation from the conference.  There are many of us who use the RT patch on our systems.  But, it's unclear to many of us who don't follow the LKML daily as to what progress is being made in the kernel community.  Can anyone enlighten the rest of us?
  
 
== People ==
 
== People ==
 
Who at ELC is interested in "Real Time"?
 
Who at ELC is interested in "Real Time"?
 +
 +
* Frank Rowand
 +
* Tim Bird
 +
* Mike Anderson
  
 
== Presentation notes ==
 
== Presentation notes ==
Line 36: Line 48:
 
Put your notes about the "Real Time" presentations from ELC here...
 
Put your notes about the "Real Time" presentations from ELC here...
  
=== Tim's notes ===
+
=== Presentations ===
This is a test of simultaneous editing....
+
 
 +
* Wed 2/20 Frank Rowand '''Using and Understanding the Real-Time Cyclictest Benchmark'''
 +
** A conceptual description of the cyclictest algorithm
 +
** How to use cyclictest to understand the real time latency of your system
 +
** What does cyclictest data typically look like for different architectures?
 +
** cyclictest options are very powerful and must be selected with care
 +
** There are several different formats of data, which provide different views of system behavior
 +
** There are debug options to help investigate causes of large latencies
 +
* Wed 2/20 Steven Rostedt '''Understanding PREEMPT-RT (The Real-Time Patch)'''
 +
** Steven is the Stable RT Branches maintainer
 +
** The math of trebuchets
 +
** Definition of real time
 +
** Varying levels of preemption available in Linux
 +
** The technology in the PREEMPT_RT Linux kernel that enable real time
 +
*** Locks
 +
*** Preemption
 +
*** Priority inheritance
 +
*** Threaded interrupts
 +
*** Disabling interrupts
 +
*** percpu
 +
*** nohz
 +
*** user space considerations
 +
** Current issues that are being worked on
 +
* Wed 2/20 Brent Roman '''Making Linux Do Hard Real-Time'''
 +
** An experience of how PREEMPT_RT Linux is used in a scientific context
 +
* Thu 2/21 Yoshitake Kobayashi '''Deadline Miss Detection with SCHED_DEADLINE'''
 +
** SCHED_DEADLINE is being driven by real time use cases
 +
** Definition and description of deadlline scheduler
 +
** Comparison of deadline algorithms
 +
** How to use the deadline scheduler
 +
** Example of how it works
 +
** Granularity improvement
 +
** Deadline miss detectioin
 +
** SCHED_DEADLINE is being driven by real time use cases
 +
 
 +
=== Presentations that incidentally touched on real time ===
 +
 
 +
* Thu 2/21 Robert Rose '''Keynote: SpaceX | Moore's Law to Mars'''
 +
** The space vehicles use PREEMPT_RT Linux
 +
* Thu 2/21 Alison Chaiken '''Embedded Linux Takes on the Hard Problems of Automotive'''
 +
** Are real-time IPC guarantees possible with IP?
 +
** CANBUS vs IP networking
 +
* Fri 2/22 Mathieu Poirier '''In Kernel Switcher: A Solution to Suppport ARM's New big, LITTLE Implementation'''
 +
** Power saving and real time are like oil and water. Real time is not yet a consideration in the big, LITTLE world.
 +
* Fri 2/22 Pantelis Antoniou '''Adventures in (Simulated) Asymmetric Scheduling'''
 +
** Q: Is the deadline schedular on the radar as a technology that can assist big.LITTLE scheduling? A: No
 +
* Fri 2/22 Gregoire Gentil '''Lessons Learned In Designing a Self-Video Self-Hovering Nano Copter'''
 +
** PREEMPT_RT is one scheduling solution that was explored
  
 
[[Category:ELC]]
 
[[Category:ELC]]
 
[[Category:2013]]
 
[[Category:2013]]
 
[[Category:Tech Zone]]
 
[[Category:Tech Zone]]

Latest revision as of 17:10, 22 February 2013

Robot64x64.png ELC2013 Tech Zone Contest

To add a link to your new page use the following wiki link format:

[[My New Page]]

Real_Time_Tech_Zone_2013#Problems_and_Issues

Save this page/section and your new page link will appear in Red. Click it and begin editing your new page.


Welcome to the Real-Time Tech Zone for ELC 2013!!

This page is dedicated to preserving information and discussions about Linux Real-time from the Embedded Linux Conference. You should think of this page like a "scratchpad" to record your thoughts and discussions from the conference. Most of us take notes at an event, at a talk that interests us, or after we learn something new and useful from another attendee. Here's a place to record those notes in public, so that not only you can access them after the event, but everyone else can as well.

Discussions

What discussions did people have in the tech zone?

Problems and Issues

What issues do you have with "Real Time"?

  • The primary issue that I have with the real-time patch is why its still a patch and not in mainline. We seem to be asymptotically approaching inclusion. But as with any asymptote, you never seem to reach zero. TG and SR and others are working to keep the patch caught up with reasonably recent kernels, but that's a lot of work that detracts from making further progress on the core issues of determinism in Linux.

What are some possible solutions?

  • The best thing would be to finally round the corner on the inclusion of the real-time patch into mainline. I realize that there differing views on the future of Linux. But, since the RT patch has seemingly minimal impact on mainline if it's not enabled, maybe 2013 will be the year that it finally makes mainline? One can only hope.

What's going on in the wider community to address these issues?

  • Stephen Rostedt had a good presentation on the RT patch and where it stands in his mind. Check out his presentation from the conference. There are many of us who use the RT patch on our systems. But, it's unclear to many of us who don't follow the LKML daily as to what progress is being made in the kernel community. Can anyone enlighten the rest of us?

People

Who at ELC is interested in "Real Time"?

  • Frank Rowand
  • Tim Bird
  • Mike Anderson

Presentation notes

What presentations at ELC, on the subject of "Real Time" did people like? Why?

Put your notes about the "Real Time" presentations from ELC here...

Presentations

  • Wed 2/20 Frank Rowand Using and Understanding the Real-Time Cyclictest Benchmark
    • A conceptual description of the cyclictest algorithm
    • How to use cyclictest to understand the real time latency of your system
    • What does cyclictest data typically look like for different architectures?
    • cyclictest options are very powerful and must be selected with care
    • There are several different formats of data, which provide different views of system behavior
    • There are debug options to help investigate causes of large latencies
  • Wed 2/20 Steven Rostedt Understanding PREEMPT-RT (The Real-Time Patch)
    • Steven is the Stable RT Branches maintainer
    • The math of trebuchets
    • Definition of real time
    • Varying levels of preemption available in Linux
    • The technology in the PREEMPT_RT Linux kernel that enable real time
      • Locks
      • Preemption
      • Priority inheritance
      • Threaded interrupts
      • Disabling interrupts
      • percpu
      • nohz
      • user space considerations
    • Current issues that are being worked on
  • Wed 2/20 Brent Roman Making Linux Do Hard Real-Time
    • An experience of how PREEMPT_RT Linux is used in a scientific context
  • Thu 2/21 Yoshitake Kobayashi Deadline Miss Detection with SCHED_DEADLINE
    • SCHED_DEADLINE is being driven by real time use cases
    • Definition and description of deadlline scheduler
    • Comparison of deadline algorithms
    • How to use the deadline scheduler
    • Example of how it works
    • Granularity improvement
    • Deadline miss detectioin
    • SCHED_DEADLINE is being driven by real time use cases

Presentations that incidentally touched on real time

  • Thu 2/21 Robert Rose Keynote: SpaceX | Moore's Law to Mars
    • The space vehicles use PREEMPT_RT Linux
  • Thu 2/21 Alison Chaiken Embedded Linux Takes on the Hard Problems of Automotive
    • Are real-time IPC guarantees possible with IP?
    • CANBUS vs IP networking
  • Fri 2/22 Mathieu Poirier In Kernel Switcher: A Solution to Suppport ARM's New big, LITTLE Implementation
    • Power saving and real time are like oil and water. Real time is not yet a consideration in the big, LITTLE world.
  • Fri 2/22 Pantelis Antoniou Adventures in (Simulated) Asymmetric Scheduling
    • Q: Is the deadline schedular on the radar as a technology that can assist big.LITTLE scheduling? A: No
  • Fri 2/22 Gregoire Gentil Lessons Learned In Designing a Self-Video Self-Hovering Nano Copter
    • PREEMPT_RT is one scheduling solution that was explored