chris@yarra.oz.au (Chris Jankowski) (08/27/90)
Most Ethernet and IEEE.802.3 LANs design guidelines recommend use
of multiport repeaters to increase LAN robustness and manageability
of fault finding.
Assume an asymmetrical configuration in the sense that there is
a large server on one segment and a large number of PCs talking to it.
The PCs are distributed among the other ports hanging off the multiport
repeater.
Just consider the following example:
Assume that N-1 PCs each on different Ethernet segment generates a packet
directed to the server at the same time. Even if the server's Ethernet
segment is quiet only one of those packets can be delivered at a time.
What happens with the others?
Does the N port repeater buffer the remaining frames?
If not (and I believe that store and forward is a function of a bridge)
what other strategy can a multiport repeater use?
Or am I missing something important altogether?
Thanks for your help.
-m------- Chris Jankowski - Senior Systems Engineer chris@yarra.oz{.au}
---mmm----- Pyramid Technology Corporation Pty. Ltd. fax +61 3 820 0536
-----mmmmm--- 11th Floor, 14 Queens Road tel. +61 3 820 0711
-------mmmmmmm- Melbourne, Victoria, 3004 AUSTRALIA (03) 820 0711
micron n. - a unit of length of one milionth of a meter,
worth $2,000,000,000 since the fault in the Hubble space telescope
mirror has been identified.abstine@image.soe.clarkson.edu (Arthur Stine) (08/27/90)
From article <65056@yarra.oz.au>, by chris@yarra.oz.au (Chris Jankowski): > > Most Ethernet and IEEE.802.3 LANs design guidelines recommend use > of multiport repeaters to increase LAN robustness and manageability > of fault finding. > Assume an asymmetrical configuration in the sense that there is > a large server on one segment and a large number of PCs talking to it. > The PCs are distributed among the other ports hanging off the multiport > repeater. > Just consider the following example: > Assume that N-1 PCs each on different Ethernet segment generates a packet > directed to the server at the same time. Even if the server's Ethernet > segment is quiet only one of those packets can be delivered at a time. > What happens with the others? If they all generate a packet at the same time, they will all see a collision. A MPR is a repeater, not a store-forward device. The server will see exactly the same line activity as any of the PC's at the same instant in time (relative to the propagation delay of the wire...) > Does the N port repeater buffer the remaining frames? No, it doesn't buffer any frames > If not (and I believe that store and forward is a function of a bridge) > what other strategy can a multiport repeater use? It just repeats the bits, just like any other ethernet repeater. It is just a multiport version ... > Or am I missing something important altogether? > > Thanks for your help. > > -m------- Chris Jankowski - Senior Systems Engineer chris@yarra.oz{.au} > ---mmm----- Pyramid Technology Corporation Pty. Ltd. fax +61 3 820 0536 > -----mmmmm--- 11th Floor, 14 Queens Road tel. +61 3 820 0711 > -------mmmmmmm- Melbourne, Victoria, 3004 AUSTRALIA (03) 820 0711 > > micron n. - a unit of length of one milionth of a meter, > worth $2,000,000,000 since the fault in the Hubble space telescope > mirror has been identified. -- Art Stine Sr Network Engineer Clarkson U ABStine@CLVMS.Clarkson.Edu
lars@spectrum.CMC.COM (Lars Poulsen) (08/27/90)
In article <65056@yarra.oz.au> chris@yarra.oz.au (Chris Jankowski) writes: >Assume an asymmetrical configuration in the sense that there is >a large server on one segment and a large number of PCs talking to it. >The PCs are distributed among the other ports hanging off the multiport >repeater. >Just consider the following example: >Assume that N-1 PCs each on different Ethernet segment generates a packet >directed to the server at the same time. Even if the server's Ethernet >segment is quiet only one of those packets can be delivered at a time. >What happens with the others? >Does the N port repeater buffer the remaining frames? >If not (and I believe that store and forward is a function of a bridge) >what other strategy can a multiport repeater use? A repeater is a very different animal from a bridge (though you may want to have them combined in some fashion). A repeater is a two-way device, so all of the segments connected by a multiport repeater have THE SAME TRAFFIC on them. So while PC#A is sending its message, PC#B and PC#C see CARRIER, and wait before transmitting. -- / Lars Poulsen, SMTS Software Engineer CMC Rockwell lars@CMC.COM
kwe@bu-it.bu.edu (Kent England) (08/27/90)
In article <65056@yarra.oz.au>, chris@yarra.oz.au (Chris Jankowski) writes: > > Most Ethernet and IEEE.802.3 LANs design guidelines recommend use > of multiport repeaters ... > Just consider the following example: > Assume that N-1 PCs each on different Ethernet segment generates a packet > directed to the server at the same time. Even if the server's Ethernet > segment is quiet only one of those packets can be delivered at a time. > What happens with the others? > Does the N port repeater buffer the remaining frames? > If not (and I believe that store and forward is a function of a bridge) > what other strategy can a multiport repeater use? > Or am I missing something important altogether? > There have been several messages lately about multiport bridges (a new market niche) and what with 10BaseT multi-ports et al, perhaps it is time to re-cover some fundamentals: "Repeaters" are bit level devices. That means that all they do is receive digital signals (pulses), regenerate them, and pass them on to all other ports. They add a little delay into the picture, but presumably less than a bit time (negligible). Repeaters propagate collisions. A "bridge" is a frame-level device. It can read Ethernet addresses, can mate Ethernet to other media (like broadband or TR), and can filter frames for reducing traffic loading. Until the advent of some of the new multiport bridges, all bridges would swallow a frame from one port before scanning it and either forwarding the frame or dropping it. This allowed for the decoupling of the collision process between the two bridged segments. At least one multi-port bridge maker (Kalpana) claims to forward frames as soon as the dest-addr can be parsed and compared to the table. This begs the question (first heard from Walt Haas in Utah) about what that does to the collision process. I believe that Walt is asking Kalpana what they do about that. Kalpana wants to minimize latency as an aid to improving performance for simple LAN protocols on Ethernet. As chips that do bridging become cheaper, we are bound to see some multi-port repeater makers come out with multi-port bridged concentrators, as well as more players in the high-performance multi- port bridge game. --Kent
henry@zoo.toronto.edu (Henry Spencer) (08/28/90)
In article <65056@yarra.oz.au> chris@yarra.oz.au (Chris Jankowski) writes: >Assume that N-1 PCs each on different Ethernet segment generates a packet >directed to the server at the same time. Even if the server's Ethernet >segment is quiet only one of those packets can be delivered at a time. >What happens with the others? Exactly what happens if they're all on one Ethernet: they collide. Repeaters just pass signals from one Ethernet to another, they do nothing to avoid collisions. (Well, barring attempts to detect and disconnect Ethernets that are defective.) -- Committees do harm merely by existing. | Henry Spencer at U of Toronto Zoology -Freeman Dyson | henry@zoo.toronto.edu utzoo!henry
pat@hprnd.HP.COM (Pat Thaler) (08/28/90)
> Assume an asymmetrical configuration in the sense that there is > a large server on one segment and a large number of PCs talking to it. > The PCs are distributed among the other ports hanging off the multiport > repeater. What happens in this case is roughly the same thing as would happen if all the stations were connected to the same coax segment. An 802.3 repeater's job is to restore the signal quality in terms of amplitude and jitter without interfering with operation of the CSMA/CD access method. > Just consider the following example: > Assume that N-1 PCs each on different Ethernet segment generates a packet > directed to the server at the same time. Even if the server's Ethernet > segment is quiet only one of those packets can be delivered at a time. > What happens with the others? > None of the packets gets delivered. A collision occurs. What an 802.3 repeater does when it receives multiple transmissions concurrently from several ports is transmit to all ports. This causes a collision on any segment where another transmitter is active. All transmitting stations detect collision and the CSMA/CD algorithm is used to resolve the collision. This is the same thing that would happen if they were all on the same segment except that the worst case delay is longer. The parameters of the Ethernet/802.3 CSMA/CD access method such as minimum frame size are set large enough to allow for the delay of 4 repeaters and 5 segments with at most 3 of those being coax. See section 10.7.1 of IEEE 802.3 for details on allowed topologies. Pat Thaler
k2@charly.bl.physik.tu-muenchen.de (Klaus Steinberger) (08/28/90)
chris@yarra.oz.au (Chris Jankowski) writes: >Most Ethernet and IEEE.802.3 LANs design guidelines recommend use >of multiport repeaters to increase LAN robustness and manageability >of fault finding. >Assume an asymmetrical configuration in the sense that there is >a large server on one segment and a large number of PCs talking to it. >The PCs are distributed among the other ports hanging off the multiport >repeater. >Just consider the following example: >Assume that N-1 PCs each on different Ethernet segment generates a packet >directed to the server at the same time. Even if the server's Ethernet >segment is quiet only one of those packets can be delivered at a time. >What happens with the others? None of those packets will go through, you will get a collision, same as if your PC's are on the same segment, and are sending at the same time. >Does the N port repeater buffer the remaining frames? No. >If not (and I believe that store and forward is a function of a bridge) >what other strategy can a multiport repeater use? >Or am I missing something important altogether? A repeater is only a electrical amplifier (some simplification, because it must handle bidirectional traffic), so it has no sense of the packets. Sincerely, Klaus Steinberger Klaus Steinberger Beschleunigerlabor der TU und LMU Muenchen Phone: (+49 89)3209 4287 Hochschulgelaende, D-8046 Garching, West Germany BITNET: K2@DGABLG5P Internet: k2@charly.bl.physik.tu-muenchen.de
pat@hprnd.HP.COM (Pat Thaler) (08/29/90)
> "Repeaters" are bit level devices.... > They add a little delay into the picture, but > presumably less than a bit time (negligible). Repeaters are allowed to add a start-up delay of 8 bit times (not including the delays of the MAUs/transceivers). The steady-state delay can be somewhat longer than that since lost preamble bits are restored. Negligible compared to store and forward delays. Pat Thaler