Newell Ryan D SrA 18 CS/SCBT wrote: > > > A collision could happen at the other end of the network > segment. > > I thought on 10BaseT net a NIC was notified of a collision by > its RX pin > getting data. So if Station A was transmitting and it was on > bit 27 and > station B > started TX and by the time it got Station As first bit and was > on bit 2. Is > the collision said to happen at the location the data crossed > on the 'bus' > or at the NIC?
Technically, the collision happens where the data crossed, which on old coax cables caused a signal reflection and resulted in extra voltage that the sender could sense. The theory and numbers come from these old networks and thinking of it that way might help you understand it. In a hubbed network, I don't think the signal reflection is really relevant, though. Plus the collision could happen on the backplane of a hub. You are right that on 10Base-T the NIC decides that there is a collision when it receives data while sending. > Back to example....Now that Station B knows of > the collision > it will finish its preamble and will send a jam signal. So will > Station A. I > can see how round trip would make sense. > > > News of the collision has to travel back to the senders. > > Would it be one of the senders sending jam signals? No. The jam signal just extends the length of the time that they are both sending. It extends the length of the time that the collision event occurs so that nobody else jumps in. > > >The signal travels outwards; the collision news travels back. > > Not really sure what you mean. My wording was old-fashioned. For 10Base-T, it should say "news of the other transmitter travels back to the sender." > > I have been reading your book and the Ethernet book. I have > been trying to > figure this out all weekend. If a bit is 17.7 meters long and > the max of a > distance of a 10BaseT net is 500 meters 500 Meters?? It's 2500 meters. In one example of such a network, there can be 5 segments, 4 repeaters (hubs), but only 3 segments can have end systems. That's the infamous 5-4-3 "rule." It makes a lot of assumptions. Really, the size of the network depends on round-trip propagation delay for the particular equipment, cables, and cable lengths. > with 4 hubs (20 bit > times) that > gives a grand total of 105 bit times. Is this the propagation > delay of the > cable? I've been trying to compare this to the definitive > guides method and > it is just not making sense in my mind. Seems like I'm over > complicating a > simple process. No, it's really not simple. I admire you for working it out. Most people just ingore the details! :-) Good luck with it. Let me know if you have any further questions. It's hard to explain it without spending huge amounts of time on the answers, but I'll try! _______________________________ Priscilla Oppenheimer www.troubleshootingnetworks.com www.priscilla.com > > > -----Original Message----- > From: Priscilla Oppenheimer [mailto:[EMAIL PROTECTED] > Sent: Monday, February 24, 2003 4:51 AM > To: [EMAIL PROTECTED] > Subject: RE: Ethernet Slot Time and Delay [7:63581] > > > Some descriptions of Ethernet refer to a segment as one side of > a hub, i.e. > just one link. The propagation delay information for a hubbed > networks takes > into account the small amount of time for a repeater to repeat. > The repeater > doesn't do much, but it does regenerate the preamble and > signal. A set of > link "segments" connected via hubs is all one collision domain. > > Anyway, read my book! Please! :-) It covers all of this in gory > detail. > > An earlier version of the Ethernet chapter is also available at > http://www.certificationzone.com/. > > _______________________________ > > Priscilla Oppenheimer > www.troubleshootingnetworks.com > www.priscilla.com > > > Priscilla Oppenheimer wrote: > > > > Newell Ryan D SrA 18 CS/SCBT wrote: > > > > > > If two 10 Base T Ethernet stations transmit at the same they > > > receive data on > > > their receive pins. Will both stations send out a 32 bit jam > > > sequence? > > > > Yes. > > > > > If both stations do send a jam signal, why is the slot time > > > closely related > > > to round trip propagation delay? I would think it would be > one > > > way. > > > > A collision could happen at the other end of the network > > segment. News of the collision has to travel back to the > > senders. The signal travels outwards; the collision news > > travels back. > > > > The goal is to make sure that the sender is still sending when > > the news travels back, even if the news had to come from the > > far end of the network segment. If the sender weren't still > > sending, it wouldn't know that its transmission got damaged > and > > wouldn't back off and retransmit. You would lose the feature > of > > the NIC ensuring succussful transmission, which happens in a > > microsecond time span, and have to depend on an upper layer > > figuring out that there's a missing ACK, which happens in a > > millisecond or worse time span. So, slot time is dependent on > > round trip time because it considers the time for news of the > > collision to travel back. > > > > Both senders transmit a jam signal to busy out the network for > > another 32 bit times. At least one of them has to do it, but > > they can't know that the other one did, so they both do it. > > > > Your question doesn't make sense, but hopefully there's some > > info in that which will help you. > > > > > > > > Ethernet, The Definitive Guide page 182 they have some > values > > > to use to > > > figure out propagation delay on 10 MB networks. There is a > > base > > > value to > > > start with and from there you add delay per meter. Why is > the > > > base value not > > > zero? > > > > Even light in a vacuum takes some time to travel any distance. > > It travels 299,792,458 meters per second to be exact, but > > still, it's not zero. A signal on a network cable travels > about > > 2/3 the speed of light. > > > > I don't know what base value you are referring to, but zero > > times anything is zero, so I doubt they could use a base value > > of zero regardless. > > > > > Also between segments the numbers do not make any sense. > > > Going from > > > Base to Max I understand but between segments. > > > > A collision domain stops at the boundary between network > > segments. A network segment is devices connected via hubs or > > coax cable. In fact, it might help you to remember that > > Ethernet was originally a long bus, like a link of Christmas > > tree lights. The signal propagated outwards from the sender in > > both directions and travelled to the end of the segment, and > > hopefully not back if the segment was terminated correctly. > But > > if there was a collision, the signal did bounce back. > > > > All hubs are is a way to gather this Christmas tree string of > > lights into a manageable structure. But when first learning > > CSMA/CD details, it helps to think of the segment like that > > string of lights. > > > > There's a ton of information about CSMA/CD in my book > > "Troubleshooting Camus Networks." > > > > Also, "Ethernet: The Definitive Guide" really is definitive. > > It's the bible. Don't doubt it. Try to understand it. > > > > _______________________________ > > > > Priscilla Oppenheimer > > www.troubleshootingnetworks.com > > www.priscilla.com > > Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=63639&t=63581 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
