[vserver] RE : [vserver] VServer and Multicasting

From: <mourad.alia_at_orange-ftgroup.com>
Date: Fri 21 Jan 2011 - 18:32:12 GMT
Message-ID: <5837_1295634732_4D39D12C_5837_21637_1_AFD6B5459394674D90CD2227EC14ED8B42DBEE1DD3@PUEXCB2C.nanterre.francetelecom.fr>

I am working with Julien in the same project (see my previous email on OpenVZ vs VServer) :

Our software is a java application representing a VoIP client which able of managing multiple calls at the same time. This application uses many server distributed over many servers (out of the scope here) to create and to route calls. In our case, we are obliged to manage such calls from one instance which is solely hosted on a VServer and which is assigned to one client ID. Consequently, we should assign many adresses to one guest. Everything is working fine without the JMX component. When enabling JMX (part of the software) which uses the socket 0.0.0.0.(all interfaces) only the first guest is started.

I hope this will help in udertanding our problem,

-- Mourad
________________________________________
De : Michael S. Zick [mszick@morethan.org]
Date d'envoi : vendredi 21 janvier 2011 18:32
À : vserver@list.linux-vserver.org
Objet : Re: [vserver] VServer and Multicasting

On Fri January 21 2011, Furgerot Julien wrote:
> Sorry for these obscures questions out of any context.
> It's a proprietary software developped by my firm.
>

Hmm...
Guess that means it is unlikely that I designed and/or wrote it. ;-)

> This software can
> manage multiple VoIP calls simultaneously, one per multicast address.
>

Still don't see your problem, why not 256 vserver contexts, one per
each of 256 individual multicast address.

> It's why we need to bind to several, dynamicaly assigned multicast IP
> addresses, not only one per guest.
>

So go ahead and assign it, only one of the 256 vserver contexts will
handle the processing - nearly identical to having only one thread handle
the processing.

> With your help, we find a workaround to addresses assignment by
> creating one interface by IP.
>

Great, now do that another 255 times.

Mike
> But now, we would like to force
> "any_addr" (0.0.0.0) binding to a choosen IP/interface, because a
> component in our software we don't manage (java JMX) bound to 0.0.0.0.
> In my experiments with vservers, only the first started vserver can
> bound to any_addr, so others can't bind to the same port.
>
> Maybe have I miss something with vserver configuration ?
>
> Sincerely,
> Julien Furgerot
>
> On Fri, Jan 21, 2011 at 5:47 PM, Michael S. Zick <mszick@morethan.org> wrote:
> > On Fri January 21 2011, Furgerot Julien wrote:
> >> Thank you for this reply.
> >>
> >> However, recall, that my software is a VoIP application which could
> >> use different (a range of) multicast adresses during its lifecycle.
> >> These addresses are allocated on demand by another software. Thus,
> >> each instance is configured to be potentially linked to one of these
> >> adresses. Furthermore, one can have many simultaneous VoIP
> >> communications where each one uses one given multicast address. Except
> >> if there is a solution to resolve this multiple multicast adresses
> >> bindings, I can't see how this could be handled.
> >>
> >> What do you think ?
> >>
> >
> > Still can not see your problem in your description above.
> > Does a single, VoIP call use multiple addresses during its lifetime?
> >
> > I would think not. Once the call is put up, it will use whatever
> > address it was assigned until the call is torn down.
> >
> > Or, at least that was the way they used to work.
> >
> > Do you mean by "my software" something you invented yourself?
> > Or do you mean "the software I am using"? What software?
> >
> > Mike
> >
> >>
> >> Julien
> >>
> >> On Fri, Jan 21, 2011 at 3:02 PM, Michael S. Zick <mszick@morethan.org> wrote:
> >> > On Fri January 21 2011, Furgerot Julien wrote:
> >> >> On Fri, Jan 14, 2011 at 5:47 PM, Herbert Poetzl <herbert@13thfloor.at> wrote:
> >> >> > services binding to 0.0.0.0 inside a Linux-VServer guest
> >> >> > will be automagically limited to the assigned IP addresses,
> >> >> > which in turn means, if you assign different IP addresses
> >> >> > to different guests, they will live happily side by side
> >> >> > even if the services inside the guests bind to 0.0.0.0
> >> >>
> >> >> You are right, I have tested when the VM is bound to one IP address
> >> >> and it works fine !
> >> >>
> >> >> However, in my configuration each VServer is bound to many IP
> >> >> addresses in order to be able to receive/send from/to many multicast
> >> >> addresses that are allocated on demand. Thus, I was wondering whether
> >> >> it is any hint so that to restrict sockets on 0.0.0.0 to be bound to
> >> >> only one of these associated IP addresses ? Is there any patch that
> >> >> can overcome this problem ?
> >> >>
> >> >
> >> > Why not just run a vserver per multicast address?
> >> >
> >> > Your whatever-it-is application is probably running an instance
> >> > per multicast address anyway (perhaps as a thread).
> >> >
> >> > If you "hashify" the on-disk files, you'll only have a single
> >> > copy of those files (on-disk and in-memory) -
> >> > So even running a few hundred context-per-address vservers would
> >> > probably not be all that resource intensive.
> >> >
> >> > Mike
> >> >> Again, thank you for all,
> >> >>
> >> >> Julien
> >> >>
> >> >>
> >> >
> >> >
> >> >
> >>
> >>
> >
> >
> >
>
>

*********************************
This message and any attachments (the "message") are confidential and intended solely for the addressees.
Any unauthorised use or dissemination is prohibited.
Messages are susceptible to alteration.
France Telecom Group shall not be liable for the message if altered, changed or falsified.
If you are not the intended addressee of this message, please cancel it immediately and inform the sender.
********************************
Received on Fri Jan 21 18:32:26 2011

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Fri 21 Jan 2011 - 18:32:26 GMT by hypermail 2.1.8