Subject: Re: [linux-audio-dev] mucos client/server implementation issues , I need opinions
From: Benno Senoner (sbenno_AT_gardena.net)
Date: la joulu 18 1999 - 07:56:01 EST
On Fri, 17 Dec 1999, Ben & Leslie Allfree wrote:
> > The API should be fullduplex, and allow a tree-like client/server
> > structure. That means every client can be the "server" of other clients.
>
> What about requiring some kind of IP-multicast hub for optimal functionality? Then
> nobody has to be a server except for their own data. If IP-multicast is not enabled,
> then yes, I think a tree-like broadcast algorithm is the fastest way to do an
> all-to-all broadcast. However, consider the topology of a tree-like network. It
> means buying special hardware, so why not just go with IP-multicast at that point?
> IMHO, most people will probably have a bus or star topology if they have not
> purchased special hubs.
yes IP multicast is not a bad idea,
but I was mainly talking about multiple clients (processes) inside of one box.
but of course shared memory is "multicast capable", for example I have the
server sampling the data from the soundcard,
ad n clients (on the same box) can read the data from shmem in parallel.
>
> Also, I wonder about shared memory. Shared memory has high overhead and is highly
> dependent on network bandwidth; is it necessary for this application?. Why can't
> clients just send or request data as message packets from other clients?
Of course: I was talking about shmem only for local clients, remote clients
will use some suitable network protocol.
There will be an intermediate "network server" which will communicate with other
"network clients" taking care of buffering etc.
Multicast works for clients on the same level, but often you have
interdependencies therefore for optimal performance you must use a mix of
multicast and single cast.
Benno.
>
> Take care,
>
> Ben
This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:23:26 EST