Renner, Michael wrote: > Hi everybody! > > I'm examining right now the possibilities of RTNet for my new project. > > I'm looking for procedure, handling a scenario like this: > > > ********** ************ *********** > * Master *---------------* router 1 *----------------* Slave w * > ********** | ************ *********** > | > | > | ************ *********** > |-------* router 2 *----------------* Slave x * > | ************ *********** > | > | > : > : > | ************ *********** > |-------* router n *----------------* Slave y * > ************ | *********** > | > | *********** > |---------* Slave z * > *********** > > I want to use one master, handling several slaves with variable routes. > The router should work with RTnet, too. The router's will have two > Ethernet-Devices, > and only should manage the traffic between the master and one or more slaves > (belonging to the routes of the slaves) with rtroute. > > The system should work with a classic TDMA-procedure, where the master works > synced > with the slaves, without managing the routers. > > How could such a system be configured ? > > Must I startup the system, handling the master, all routers and all slaves > and working when > with tdmacfg with only handling the TDMA with master and slaves ?? > > Does the routing on the routers work, without beeing involved in the > TDMA-slotting ? > > Any help is welcome, thanks a lot.... >
TDMA as it is now does only work within local network segments (i.e. broadcasting domains). The Sync frame will not make it automatically across the routers. And as that frame is not based on IP, there is also no "trick" (routing entry) to forward it between both networks. If you depend on the two-level hierarchy of the figure above, the routers will have to become masters as well. I could imagine, e.g., to extend TDMA in a way that the cycle of a master instance on one real-time NIC (right hand side of the routers above) gets sychronised on the master's clock coming in on some other real-time NIC (left hand side of that routers). Should be straight forward and should not require significant changes to the current implementation. Is this a feature which would help you? Note that the routers will introduce additional errors to a global clock distributed from the root master to the second-level slaves. Regarding the relation of TDMA and IP routing: yes, they are independent. TDMA manages any kind of traffic, including IP. And the real-time IPv4 routing does not depend on TDMA, it would work with other RTmac disciplines or even without any at all. Jan
signature.asc
Description: OpenPGP digital signature
