interesting idea except that 10 1Mb lines with their own wave slice would have a "worse feel" than 1 10Mb line shared by 10 machines. Due do random statictics theory not all machines will xmit at the same time say 90% of the time. That means that the 1 machine that needs to borrow the other bandwidth can do so. Ethernet is all about statistical randomness and works because it uses randomness to its advantage. 10Mb ethernet can take an extreme beating of traffic and still function with very good responsiveness because even if the line it maxed at near 100% the back-off ethernet algorithm when a collision happens will ensure that even the infrequent talkers get just as much of a chance to get their packet on the network as the machines eating up all the traffic with heavy loads. Iv'e seen 500+ machines on a flat 10Mb segment (collision domain) with that many college kids copying warez and porn day and night and the performance is still extremely good. perhaps latency of 10-15ms to talk to other nodes on the segment under extreme loads and many hosts condending for a space on the line. On a 10Mb line an extra 15ms in delay may mean many transmit delays (deferred transmits) to send the packet but to us is still very good response. Also remember, collisons are NOT bad. They never cause lost packets because the nic will always re-send the packet. All it causes is a delay of sub-millisecond time. And unless your 10Mb segment is at about 85% or higher utilization a packet which causes a collision will probably only have to be tried once or twice more before being sucessful adding about 1-2ms if that. Most of the time when we claim that our network is performing bad because its over utilized its because we are dealing with apps that were written poorly to handle different traffic patterns. They do things like time out too soon or retransmit too quickly or too often and get themselves all confused. (eh hem...MS apps?) And broadcast storms?....a thing of the past. Most currently used protocols dont do much broadcasting anymore to the point where it could kill a segment...although i will point a small finger at MS and netbios again. Cheers At 09:27 PM 3/26/01 -0600, you wrote: > > > > > I was wondering if anyone has >done this already or if it would be impossible. Say you got a network of >10 hosts on a subnet going to a router. As it stands now we transmit to >all the hosts on the subnet simultaneously and this increases traffic on >an ethernet network. We know this slows things down a bit in some cases >because each host is competing for "time" on the network. Is it possible >to modulate each of the hosts's transmissions on the subnet so that each >host is transmitting at a different frequency on the same line? If you >could do that, all the hosts could talk at the same time to the router >without signals bumping into one another if they are all talking at a >different frequency. Wouldn't this significantly increase the amount of >speed that a host could operate on if we didn't have to worry about >contention? > > > - Jamie > > > > > >-- > > "It's pretty hard to stop a man who eats his toast every morning." > > > > > >_______________________________________________ >tclug-list mailing list >tclug-list at mn-linux.org >https://mailman.mn-linux.org/mailman/listinfo/tclug-list >