here's a little question. We all know that to be considered a wire speed a device in full duplex should have a thoughput almost equal to the speed of its interface.
To be more clear, the theoretical value of the thoughput on a 10Mbps FDX interface should be of 7.6 Mbps when packets of 64 bytes are sent.
My calculation is the following:
10Mbps(speed of the interface)*( 64byes(lengh of packet)/64bytes(lengh of packets) 8bytes(preamble)+12bytes(interpacket gap) )
or 10Mbps * 64/84 = 7.6
What do we consider wire speed for HDX interfaces. Should it be the FDX speed/2 (i.e 3.8Mbps) for 64 bytes packets in 10HDX or is there another calculation?
Is there a certain standard defining what is wire speed in FDX and HDX mode? (i.e is a device that has a thoughput of of 7.5 or 7.4 for 64 bytes packets considered wire speed). If not what is a resonable assumption to say that a device is at wire speed? Is a 1% difference (of thoughput) between the theretical value and the tested value too little?
Full duplex actually means that it's 10/100/1000 on both transmit receive, so in theory, a 10 Mbps is actually 20 Mbps, and half duplex would be 10 Mbps. You also have taken into account the packet header and length, but there is also the ethernet frame that will introduce even more overhead. I think I've heard estimates that ethernet only runs at about half it's advertised speed.
The full duplex question is interesting. Jwj is right that you basically get 100Mb/s transmit plus 100Mb/s receive but remeber that most software can't actually receive and transmit at the same time - for example TCP will send you a basket full of data but eventually it wants an acknowledgment back and won't send you any more until it gets one. So although it's tempting to think of full duplex as giving you twice the bandwidth in practice you won't achieve that. What full duplex does do, however, is eliminate the collisions and overheads associated with reversing a single hallf duplex link each time the conversation changes direction; the replies shoot straight back unhindered along the second link and the software can get on with it again