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

Re: still troubles



> > This is right, it should also be noted that the jnodeos can run just fine
> > on a regular JVM, it just doesn't have resource controls.  So you can
> > develop ANTS applications without having to go for the full bore
> > oskit/moab/janosvm/jnodeos/ANTS thing, unless you're really interested in
> > resource controls.  Also note, I don't believe we've actually run ANTS on
> > a bare machine yet, we've only done some custom apps on the jnodeos.
> >
> > tim stack
> 
> Thanks for the info..
> 
> I have a small working ANTS application, and was wondering 
> where the jnodeos calls would be placed...

ANTS wraps the jnodeos so an application should rarely have to deal with
jnodeos classes.

> How do I associate a given ANTS protocol with its resources ??
> In theory when developing a new service, I should be able to limit its 
> resource consumption, but I haven't quite figured out how...

This happens automatically in ANTS, the classes that make up a protocol
are run through a hash to create a unique identifier that is placed
in the header of the packets for that protocol.  ANTS then uses the
jnodeos to create a `flow' (a unix like process) for each protocol and
then all packets with the unique identifier are dispatched to this flow
where they are handled.  Currently, the resources for these flows are
predetermined and there is no real path to configure them...

> Over the weekend I'll have a closer look at the NodeOS papers, 
> and hopefully will come out with a closer understanding of the NodeOS 
> objects....
> I am evaluating the use of ANs for the transport of video streams, 
> so I am looking into some way to realize ressource reservation...
>
> My ANTS app is only a classifier-dropper and I wanted to 
> run it on a operational active router to get a performance comparison
> with a similar thing in a linux router, that's why I am trying to build 
> the whole thing and run ANTS on it :)
> 
> The idea behind my project is to be able to selecively drop packets, but 
> guarantee that some of them will be forwarded (with reserved resources).

I believe something like this has been done by the PANDA people if you
want to check them out.

> cheers
> 
> Andreas

tim stack




[ Janos ] [ OSKit ] [ Network Testbed ] [ Flick ] [ Fluke ]
Flux Research Group / Department of Computer Science / University of Utah