This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.
Subject: [OT] Cascaded eth hubs From: Gustav Schaffter <gustav@schaffter.com> Date: Tue, 29 Aug 2000 10:56:52 +0200 Sorry for the off topic question, but I really don't know where to turn otherwise. (If somebody has any serious suggestions...) I have two ordinary ethernet hubs. One (hub B) is uplinked to the other (hub A). Questions: - Is it possible to cascade a third hub (hub C) by uplinking it to hub B? - Is it possible to cascade a third hub (hub C) by uplinking it directly to hub A? - Will all NICs be able to reach each other in both the above connection models? Specifically in the second case, where both hub C and hub B are uplinked to hub A, will hub C and hub B 'see' each other through hub A? I've learnt that there is a maximum length of about five meters for an uplink cable. - Is there any way to (by adding some equipment?) lengthen this distance? === Subject: Re: [OT] Cascaded eth hubs From: Tom Minchin <tom@interact.net.au> Date: Tue, 29 Aug 2000 20:04:29 +1100 On Tue, Aug 29, 2000 at 10:56:52AM +0200, Gustav Schaffter wrote: > Hi, > > Sorry for the off topic question, but I really don't know where to turn > otherwise. (If somebody has any serious suggestions...) > > I have two ordinary ethernet hubs. One (hub B) is uplinked to the other > (hub A). > > Questions: > > - Is it possible to cascade a third hub (hub C) by uplinking it to hub > B? > Yes - with a cross over cable (or if the hub supports it just a straight through and set crossover on the port) > - Is it possible to cascade a third hub (hub C) by uplinking it directly > to hub A? yes > > - Will all NICs be able to reach each other in both the above connection > models? Specifically in the second case, where both hub C and hub B are > uplinked to hub A, will hub C and hub B 'see' each other through hub A? > mostly, some NICs and cheap hubs have problems. Very rare though. > > I've learnt that there is a maximum length of about five meters for an > uplink cable. > > - Is there any way to (by adding some equipment?) lengthen this > distance? > Should be right for 50-100 metres. === Subject: Re: [OT] Cascaded eth hubs From: Tom Minchin <tom@interact.net.au> Date: Tue, 29 Aug 2000 23:39:05 +1100 On Tue, Aug 29, 2000 at 02:16:16PM +0200, Gustav Schaffter wrote: > Tom, > Tom Minchin wrote: > <snip> > > > I've learnt that there is a maximum length of about five meters for an > > > uplink cable. > > > > > > - Is there any way to (by adding some equipment?) lengthen this > > > distance? > > > > > > > Should be right for 50-100 metres. > 50 - 100 meters !?!?! Do you, or anyone else, have first hand experience > using that long cables for uplinks? > > The two manuals I've read from my two hubs (Micronet & Linksys) both > talk about 5 meter. Yeah, our record was 90 meters (across the length of a building and up 4 floors). === Subject: Re: [OT] Cascaded eth hubs From: "Jeff Hogg" <tastin@catchnet.net> Date: Tue, 29 Aug 2000 08:50:03 -0500 Yep, got a 16 port 100mbit hub and 8port 10mbit hub uplinked via a 35meter straight through cable. No probs so far even though I've mixed up brands and speeds a bit. === Subject: Re: [OT] Cascaded eth hubs From: Chris Watt <cnww@chebucto.ns.ca> Date: Tue, 29 Aug 2000 10:34:32 -0300 At 10:56 29/08/00 +0200, Gustav Schaffter wrote: >Hi, > >Sorry for the off topic question, but I really don't know where to turn >otherwise. (If somebody has any serious suggestions...) > >I have two ordinary ethernet hubs. One (hub B) is uplinked to the other >(hub A). >- Is it possible to cascade a third hub (hub C) by uplinking it to hub >B? Yes, usually, but it's not recommended. According to spec you're not supposed to uplink hubs more than two deep (i.e. if you want a Hub C you should uplink both it and Hub B to Hub A if at all possible), and if you want to have a 3-deep stack you should "ideally" be using a switch on top. Personally I would say if you're under physical restrictions on hub placement the cheapest thing to do within spec would be to uplink hubs A and C to hub B in the middle (so it's only two deep). My networking guru tells me about 3-deep hubs: "Some work, some don't, some sometimes work". >- Is it possible to cascade a third hub (hub C) by uplinking it directly >to hub A? Sure. You can potentially have every port other than the uplink port (if it has one) connecting to another hub either with a crossover cable or to another hub's uplink port. The problem with this is that you very rapidly start to run into traffic crowding problems (especially if you have more than 1 server) because the hubs just blindly bounce packets around your entire network all the time. Generally if you're going to connect a bunch of populated hubs to a center it's best to use a switch in the center (thankfully it's now possible to get small 10/100 ethernet switches for only slightly more than the cost of a regular hub) which will a) allow machines connected directly to it to use full duplex mode (good for servers) and b) only forward packets to whichever port can communicate with the packet's destination MAC address (so you only get them on the segment of the LAN on which they're wanted, and you don't interfere with traffic on other parts of the network. This also makes packet sniffers useless because no packets are round to sniff). >- Will all NICs be able to reach each other in both the above connection >models? Specifically in the second case, where both hub C and hub B are >uplinked to hub A, will hub C and hub B 'see' each other through hub A? In the config where B and C uplink to A, it will work perfectly. If you have B uplinked to A and C uplinked to B it might work, or it might not. In any case systems plugged into any of the hubs should be able to communicate normally with any other system. >I've learnt that there is a maximum length of about five meters for an >uplink cable. This depends on the network technology in use. My home network consists of a 5-port 10/100 hub uplinked to an 8-port 10/100 switch. According to the hub's documentation it can use a 100 meter uplink cable if it's serving only fast ethernet (100mbit/s) devices, or 10 meters if one or more of the devices is 10mbit/s. The documentation for my switch simply says to "use 100-ohm Cat 5 UTP or STP cables as supported by the IEEE 802.3u Type 100BASE-TX standard", I don't want to go through the project 802.3 docs at my local CS library to see if it has a cable-length restriction in it at the moment, but I can tell you that the ISP in my "from" field has hubs which are cascaded using cables installed in the wall, well in excess of 10 meters, and this does not appear to be uncommon. They do not have a packet loss problem (we did for a while, but it turned out that one of my associates had misplaced a wire in the patch panel when setting up one of the wall cables to connect to a firewall in our machine room). >- Is there any way to (by adding some equipment?) lengthen this >distance? The most obvious answer is: Buy a hub/switch with a length restriction that is large enough to handle what you want to do. For really long runs consider using a router/bridge (a cheap Linux box works well for this) and shifting to a different media (e.g. DSL or fibre) for the long cable run. As long as you're working in a single small to medium-sized (say less than 1/2 square km (5.4*10^6 sq.ft.) per floor and less than 20 floors) building you should be able to just use ordinary (or fast) ethernet provided you pick your hardware properly. n.b. Most of these cable length restrictions can just be ignored most of the time. The trick is being able to predict when ;) >50 - 100 meters !?!?! Do you, or anyone else, have first hand experience >using that long cables for uplinks? Not sure exactly, but I have a hub on the 3'rd floor of a building with it's uplink port connected to a roughly 4M cable, connected to a wall-jack which has a cable at least 20M long and possibly somewhat longer (it does not go straight down), hence to a patch-panel in the basement of the same building and to a 10M cable running in a cable guide along the wall to another hub (and hence to a firewall). This works perfectly and I'd guess the total cable length should be around 40M. . . === Subject: Re: [OT] Cascaded eth hubs From: Andrew So hing-pong <andrewso@netvigator.com> Date: Tue, 29 Aug 2000 22:22:30 +0800 (HKT) in fact, the limitation of the length of twisted-pair cable is not so clear. All depend on the environment along the cable. If most noise or electrical signal around the cable, then the connection length will be reduced becaue of the attenuation. === Subject: Re: [OT] Cascaded eth hubs From: Graham Hemmings <rhlists@ntlworld.com> Date: Tue, 29 Aug 2000 17:18:26 +0000 At 09:04 29/08/2000, you wrote: >On Tue, Aug 29, 2000 at 10:56:52AM +0200, Gustav Schaffter wrote: > > I have two ordinary ethernet hubs. One (hub B) is > > uplinked to the other (hub A). > > Questions: > > > > - Is it possible to cascade a third hub (hub C) by uplinking it to hub > > B? > > > >Yes - with a cross over cable (or if the hub supports it just a straight >through and set crossover on the port) > > > - Is it possible to cascade a third hub (hub C) by uplinking it directly > > to hub A? > > yes There is a limit of 1 or 2 repeaters in an ethernet segment, so 3 hubs in a chain is your limit - whether it is 1 or 2 repeaters depends on the type of hub... If you designate one hub as a "root" and then patch other hubs into this root, you should be fine for as many as you like. ===