Tim wrote: > Wayne, > > >> Not surprised. I couldn't see a way to broadcast using SOCK_STREAM. >> Googled all over trying to find an example. Does it follow that since I >> would like to use SOCK_STREAM for the rest of the communications during >> the optimization job, the multicast side would need to be on a second >> network interface? >> > > Honestly, I've never implemented anything with multicast. It is just > a neat way to efficiently distribute traffic. > > I've done a little digging, since this is something I want to learn as > well, and I believe you just need to set some socket options which > instruct the kernel which multicast groups you want to subscribe to. > Then in the background the kernel will notify the appropriate routers > what group traffic you should receive, etc. I'm a little fuzzy on > what group you would want to pick for simple local experiments... > > In any case, it shouldn't require using a different interface. Since I would need to select a multicast class D address for the multicast interface, I was wondering if I could use (TCP/IP) for the second socket on that address / interface.
Thanks you very much for your links. The C code example is great. Time available is such that I could send all 24 to 36 distributed processing helpers the files individually at the start of each job, but if there was a easy way to send all the same file at the same time, I would have . Jobs occur a the rate of a few / minute and I would like to get the answer in a few seconds. Wayne _______________________________________________ PLUG mailing list [email protected] http://lists.pdxlinux.org/mailman/listinfo/plug
