[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Publicity] Draft blog post on the ARINC653 scheduler

On 12/4/2013 7:29 PM, Dario Faggioli wrote:
> On mer, 2013-12-04 at 12:43 -0500, Nate Studer wrote:
>> Dario,
> Hi Nate!
>> It has been awhile, but I finally got a draft blog post giving some basic
>> details about the background and algorithm of the ARINC653 scheduler together
>> for your review.
>> http://blog.xen.org/?p=8453&preview=true
> So, first of all, thanks for doing this. Really!
> I went through the post, and it already looks pretty good.
> They're very minor issues but can you, perhaps, make the pictures a bit
> bigger. It would be perfect if you could make them fit the width of the
> text area of the blog. I see they're 300px right now... I think 450 or
> 500 would do.

I am not a WordPress expert by any means and I am having trouble getting them
bigger without making them fuzzy.  If you have some tips, feel free to share.

> Also, could you link to either an official specs of ARINC653 or, at
> minimum, its Wikipedia page (or something similar)? 
Could this
> "[EASA.2011/6 MULCORS - Use of Multicore Processors in airborne
> systems]" be a hyperlink to (perhaps to this page
> http://www.easa.europa.eu/safety-and-research/research-projects/large-aeroplanes.php
>  or to whatever else you think it's best)?

The standard itself is unfortunately not free, so the wikipedia page will have
to do.

I made the last reference a hyperlink as well to.

>> Any ideas on how to make the pictures look better are appreciated.
> Ok, here they are some more substantial thoughts and comments:
>  - in the background, you mention how important isolation is in avionics
>    systems and I think (also considering my background) I actually see
>    it. Still, could you provide an example on what the various
>    components needing to be isolated between each others are. It doesn't
>    have to be 100% accurate, just give people an idea what kind of
>    subsystems/software we're talking about. Also, what's the role of
>    virtualization in that? I mean, I got that an ARINC scheduler
>    guarantees strict temporal isolation, but, is using virt --and then
>    having an ARINC scheduler in the hypervisor-- common or is an actual
>    added value? Can the same be achieved by "just" an OS providing an
>    ARINC scheduler and the different components being different
>    processes? If no (i.e., if you think virt is actually better), why is
>    that?

I updated some of the background information to give some example applications,
and to better connect the background and development of ARINC653 to 

>  - I see you never mention the fact that the scheduler has been
>    implemented and is being maintained by DornerWorks. Well, we are an
>    Open Source community, so it definitely does not matter much what
>    company does the work but, in this case, I think it's a relevant
>    piece of information. At least only as an historical note, I think
>    I'd say that somewhere (perhaps linking here
> http://dornerworks.com/news/2010/12/dornerworks-releases-prototype-arinc-653-hypervisor-to-open-source-community/
>    or wherever you think it's appropriate).
>  - Related to the above, I think it would be nice to have a few words
>    about how and for what you have used it (up to what can be said in a
>    post like this, of course). Even better (related to the first point),
>    why Xen was chosen in the first place and it is being maintained. As
>    you know, we're starting to see interest in embedded virtualization,
>    and the view of an early adopter, like you are, on why Xen could be
>    the right choice there would be *gold*, or at least so I think.

I added a link to the above address, which should answer the question of what it
has been used for.  Most of our current work is being done with the US military,
and we have to jump through a lot of hoops before we say anything about it in a
public forum.  (It is easier to just say nothing.)

As to why we choose Xen, at the time, we choose Xen primarily because it was
Open Source and it was running on multiple platforms (x86, PPC, ARM).  Since
then KVM has added platforms and Xen has dropped some, so I guess we need to
come up with some new reasons to differentiate between KVM and Xen.

>  - I see you mention "worst case running times" and I think people may
>    ask *the* question: <<how in h%$l do you compute that?>>. However,
>    this is probably a topic for another post, as giving details about
>    this would be both too technical and too long... But personally, I'm
>    really curious on whether you actually do that, and, if yes, how,
>    especially with the hypervisor in your way! :-)

People have written Phd theses on it, so I doubt I could do it justice in a
single paragraph.  It is somewhat easier in aviation though, since things are
much more static.

> Again, I'm fine with the content as is... All I did above was sort of
> thinking out loud. :-) So, if, for any reason, you don't like or don't
> think any of the above is appropriate, just ignore me. :-)

Your comments are appreciated.

My biggest concern is really to keep it to a readable size and to keep it from
becoming too complex or esoteric, which is difficult when most Xen
developers/users are not familiar with it.

> Thanks again and Regards,
> Dario

Publicity mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.