Sending a frame of any size on 100 Mbps requires a node to use the medium for 1/10 the time it would use it on 10 Mbps. This reduces contention for the medium.
Whether this reduces collisions depends on the sending patterns of the nodes. They could all send at the same time anyway, but that doesn't tend to happen. The frames arrive randomly on most Ethernets. If there's more time between frames, due to the shorter duration of transmissions, it's more likely that two senders will be able to send when they want and without encountering a collision. Although this isn't necessarily true, it is what tends to happen. Remember that this entire discussion started because his provider suggested he move to 100 Mbps to reduce the excessive collisions. That's good, classic advice. Using full-duplex is also good advice since it's a point-to-point link. He is doing both now. The problem is solved. I just didn't want us to forget what a collision really means and act like it's just a configuration mistake. Priscilla Symon Thurlow wrote: > > Actually, I disagree with a comment made way down in this > email!!!: > > One fix to the problem is to increase the capacity. > >>> > >>By jumping > >> > >>>10-fold from 10 Mbps to 100 Mbps, the risk of collisions, > >>> > >>especially > >> > >>>excessive collisions, goes way down. Since each frame takes > >>> > >>1/10th the > >> > >>>amount of time to send, the odds that some other station is > >>> > >>sending when > >> > >>>another station transmits (or retransmits) go way down. > >>> > > This may be true if the frame is 100Mb in size, but if it is a > 1500 byte > frame, then surely there is no difference in the capability of > a 10Mb or > 100Mb Ethernet to pass the frame? > > Symon > > > -----Original Message----- > From: Larry Letterman [mailto:[EMAIL PROTECTED]] > Sent: 03 December 2002 23:02 > To: [EMAIL PROTECTED] > Subject: Re: Collision [7:58389] > > > Not the last word, but you imply that the collisions are only > due to > capacity... > and I can have the wrong match on both ends and get plenty of > collisions > > with no > capacity issue.....and I reserve the last word as always for > you...where > > have you been lately, > I have missed you !!!! :) > > Priscilla Oppenheimer wrote: > > >Larry Letterman wrote: > > > >>set one end to 100 half and the other to 100 full and see what > >>happens....:) > >> > > > >I mentioned the duplex mismatch problem too, but it has > nothing to do > >with his question or problem. > > > >The key to troubleshooting is to address the actual problem, > not some > >assumption you make about the problem. > > > >Thought you would have the last word, eh? :-) > > > >Priscilla > > > >>Priscilla Oppenheimer wrote: > >> > >>>Larry Letterman wrote: > >>> > >>>>Most likely the previous 10/half interface on the switch > and the > >>>>router were not > >>>>linked at the same speed/duplex or the other router had an > >>>>issue with > >>>>the setting. > >>>> > >>>No, the switch and router were set to the same thing, which > >>> > >>was 10 Mbps half > >> > >>>duplex, if you read his messages. He was using a 2500 router. > >>> > >>Those routers > >> > >>>predate the full duplex standard. In fact they may predate > 100 > >>> > >>Mbps also. He > >> > >>>had no choice but to upgrade the router, which he did. > >>> > >>>He was seeing lots of collisions, including excessive > >>> > >>collisions where the > >> > >>>frame got dropped because even after 15 retries it > encountered > >>> > >>a collision. > >> > >>>Collisions are normal in shared (half-duplex) Ehternet, but > >>> > >>excessive > >> > >>>collisions are not. Collisions are caused by the stations on > >>> > >>the shared link > >> > >>>simultaneously sending. Excessive collisions are due to a > >>> > >>shortage of > >> > >>>capacity. One fix to the problem is to increase the capacity. > >>> > >>By jumping > >> > >>>10-fold from 10 Mbps to 100 Mbps, the risk of collisions, > >>> > >>especially > >> > >>>excessive collisions, goes way down. Since each frame takes > >>> > >>1/10th the > >> > >>>amount of time to send, the odds that some other station is > >>> > >>sending when > >> > >>>another station transmits (or retransmits) go way down. > >>> > >>>Increasing capacity used to be the only way we could upgrade > >>> > >>an Ethernet > >> > >>>network. Then the full-duplex standard came about. It can > only > >>> > >>be used on a > >> > >>>point-to-point link where each side has its own dedicated > >>> > >>transmit path. In > >> > >>>other words, it's no longer shared Ethernet. There's no need > >>> > >>to sense > >> > >>>carrier to see if anyone else is sending, because there isn't > >>> > >>anyone else. > >> > >>>It's not multiple access. Receiving while you're sening is > >>> > >>legitimate, so > >> > >>>there's no need to check for collisions. It's no longer > >>> > >>CSMA/CD. Of course > >> > >>>the collision rate goes down. Collisions really have no > >>> > >>meaning in this > >> > >>>environment. If there are collisions, then there's probably a > >>> > >>duplex mismatch. > >> > >>>So, anyway, he improved matters in two ways: upgrading the > >>> > >>capacity and > >> > >>>moving to full duplex. > >>> > >>>I just wanted to add this theory discussion. It's not right > to > >>> > >>say (as > >> > >>>someone else did) that collisions are "caused by" a > >>> > >>half-duplex setting. > >> > >>>Collisions are caused by two stations sending at once, which > >>> > >>tends to happen > >> > >>>more and more frequently when there's not enough capacity to > >>> > >>support the > >> > >>>sending behavior of the nodes on the shared network. To fix > >>> > >>the problem, you > >> > >>>can increase capacity or you can make the network not shared > >>> > >>by connecting > >> > >>>just two devices and using full duplex. > >>> > >>>_______________________________ > >>> > >>>Priscilla Oppenheimer > >>>www.troubleshootingnetworks.com > >>>www.priscilla.com > >>> > >>>>To be safe I would set the switch port and the router > >>>> > >>interface > >> > >>>>to > >>>>100/full or 10/full > >>>>and there should be no issues then. > >>>> > >>>>and yes, the fast ethernet in the 26XX/36XX routers are a > better > >>>>solution..... > >>>> > >>>>Larry Letterman > >>>>Cisco IT-LAN , San Jose > >>>> > >>>>Cliff Cliff wrote: > >>>> > >>>>>Today, We are put 3660 router to their end, having > >>>>> > >>>>Fastethernet card, and > >>>> > >>>>>connected to their switch. > >>>>> > >>>>>They change their switch port as following: > >>>>> > >>>>>interface FastEthernet0/14 > >>>>>load-interval 30 > >>>>>duplex full > >>>>> > >>>>>so far, after observe serveral hours, there is no collision > >>>>> > >>as > >> > >>>>well as not > >>>> > >>>>>error message in our router. > >>>>> > >>>>>So, what's wrong? Is the fastethernet is better? or > previous > >>>>> > >>>>setting that I > >>>> > >>>>>have is wrong? > ============================================= > > This email has been content filtered and > subject to spam filtering. If you consider > this email is unsolicited please forward > the email to [EMAIL PROTECTED] and > request that the sender's domain be > blocked from sending any further emails. > > ============================================= > > Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=58519&t=58389 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
