is multicast *really* to all stations at once with a switch?

Rick Jones wrote: (snip)

Instead use a repeater and only connect the transmit outputs to the inputs of the switch under test. A repeater should reliably transmit on all ports at the same time. You can even vary the cable length to add small delays.

Then again, if you work at HP just ask the people in the ethernet switch section.

-- glen

This will be imeplementation dependent.

Some switches (e.g. Cisco Catalyst 5000/5500, and non-fabric 6500) have an internal bus that sends all incoming traffic to *every* port. It is buffered there until a signal is sent on the control bus telling the ports what to do with the traffic. Send or discard.

In that case the traffic will be send simultaneously from every port, as long as there is no contending traffic.

This architecture though is not used with newer cisco switches since I don't think that is scales well enough for current throughput requirements.

I woudl imagine that the internal bandwidth in a modern switch would be sufficent to make the delays small.

You can still get linecards for the 6500 that operate on the "old"

If you did decide to implement it, it would be worth considering giving the users a broadcast option (all Fs). Note also that on a multicast enabled network (i.e. where multicast traffic is not treated as an all Fs broadcast) the end nodes have to join the group (IGMP) before the switch will send the traffic towards that host.

I think that IGMP enabled stack is an extra cost add on for windows. A quick look suggests that it might be included now. Tibco is widely used by multicast developers but will not be cheap.

I just want to make certain my knowledge horizon doesn't begin and end with a single vendor :)

rick jones