Talk:Network switch/Archive 1

Function?

I would like to know which sentence of the paragraph "Function" describes the function of a switch. Further reading of the article reveals that "switch" is a "marketing term". My provider put one in my home, so it must be more than a "marketing term" or he would not have bothered to send over a technician to install one.

It is a marketing term. First, there is no rigorous definition of a switch. Most often, it is a bridge or router, optimized for Ethernet connectivity. It may include NAT or firewall functionality.

My dentist used to tell me he was giving me "Novocaine", a trade name for procaine, when it had been replaced by "Xylocaine", a trade name for lidocaine, for decades. At one time, routers did pure software lookup, where "switches", which were Ethernet-optimized bridges that were not as scalable as routers, did forward faster. Modern router design is such that this speed difference is essentially gone.

The Comcast cable access device in my house is a router with firewall, NAT, and VoIP functions. I've designed routers. I never knew anyone who designed a switch, other than to turn electricity on or off.

Personally, I'd be delighted if the "switch" article simply pointed to "router", with a caveat that some "switches" are L2 only. Howard C. Berkowitz 18:55, 1 December 2007 (UTC)

Let me point out that the author claims that "router" is a "marketing term" for a layer 3 switch. So, what *are* the technical terms? Vandersluism (talk) 23:47, 28 July 2008 (UTC)

Routers and switches are not the same. A router is a communications device designed to distribute incoming data based on its destination OSI layer 3 address. A bridge or a switch provides the same service based on the destination's layer 2 address. There are devices that provide both of these services (and more) in the same physical container. In this case some people refer to the device as a layer 3 switch. Layer 3 switches still make forwarding decisions based on layer 2 addresses for destinations that appear in its layer 2 forwarding table (the layer 3 component must use proxy ARP to participate in forwarding. Even if the layer 3 switch is constructed with more than a single layer 2 device, each of the layer 2 devices is still required to use the layer 3 capability to forward traffic between them. Docdave (talk) 23:24, 30 September 2011 (UTC)

In Geek speak - layer 2 forwards on MAC to PHY mapping table contents and layer 3 forwards on routing table contents. Traffic destined for off-LAN destinations must always traverse the layer 3 device even if the destination LAN exists as a VLAN on the same physical device. A router is not a marketing term for a layer 3 switch; a layer 3 switch is a marketing term for the combination of a layer 2 switch (micro segmentation bridge) and a layer 3 router Docdave (talk) 23:24, 30 September 2011 (UTC)

Basic introduction missing

Sorry, but all I've really understood from the entire article is that the thing is a computer gadget. But in plain English:

• What is it?
• Does every network have one?
• If your network doesn't have one of these what else might it have?

In really plain English, "switch" is really a marketing term, although, with some vendors, "switches" and "routers" may describe differently optimized product lines. Let me offer some basic definitions of what can generically be called a network relay. Repeaters operate at the physical layer, essentially passing all bits. Bridges operate at the data link layer, making forwarding decisions based on MAC layer headers. Routers operate at the network layer, making decisions on IP headers.

While there is no standard definition, a layer 2-only switch is a LAN-optimized bridge that avoids collisions, by connecting each device to its own full-duplex port. With many vendors, the only difference between a layer 2 and layer 3 switch is the software image. Layer 3 LAN switches generally offer higher Ethernet port density and less quality of service/buffering than "router" products. Assorted other optimizations, not clearly layer 2 or layer 3, include multicast management with IGMP snooping (or equivalents). You may also see QoS management at the end device level, especially for switches intended to support VoIP. Hcberkowitz 02:44, 26 June 2007 (UTC)

How does one remove the vandalism here, since it doesn't appear in the edit text? 207.99.73.226 19:10, 12 October 2006 (UTC)

I have tried to address these issues. --Boscobiscotti 05:47, 25 May 2007 (UTC)

basic intro added, removed maint. tags. --akc9000 ^{(talk • contribs • count)} 02:33, 30 June 2007 (UTC)

I agree... the basic problem with this article is that it fails to spell out in simple terms what a switch is, and just leaps straight into rambling periphrastic and IMO over-technical stuff.

I've studied this stuff at uni, but I tend to think that at least half of a wikipedia article should be intelligible to my mum... if she can't follow it, then I know that a basic information accessibility test has been failed!

I'm not suggesting you import the simple english version of the article; but, I don't think this article should start off talking about "network segments" (never heard of that term, not so keen on the article it leads to either), ISOOSI and layers in the first para... just state what it is in a sentence; what it does, in another sentence.

Hows about: "A network switch is a computer networking device that connects multiple computers and other devices in a network..."

better still: "A network switch is a small hardware device that joins multiple computers together within one local area network (LAN). Technically, network switches operate at layer two (Data Link Layer) of the OSI model. "

[1]

helpful-er? [2]

Think about questions your mum might ask... e.g.: "can I plug it into my router and have more ethernet things plugged into it?" —Preceding unsigned comment added by 87.115.101.107 (talk) 18:29, 12 July 2009 (UTC)

Improvement drive

This page lacks any useful information or detail. I'm going to make it a bit arch-specific, because the ethernet switch article simply redirects here. If this is a problem, someone can turn it into a disambiguation page if nessicary later. -- Gamera2 02:27, 27 Aug 2004 (UTC)

Switch Limitations Question

In the realm of very high-speed network switches, what is the limitation to higher bandwidth? I've seen Infiniband running at 10Gbps, why not 100Gbps? What's the limitation? -WikiNewbie —The preceding unsigned comment was added by 67.160.157.239 (talk • contribs) .

thats actually a limitation on standards and medium

Ethernet has yet to use Fiber nor Copper Cabling to reach such high speeds.

a Switch cannot support something until it exists\

The limitations are capacitance, silicone device current ratings, inductance and crosstalk. I'm not sure if you're any the wiser though - if you don't understand electronics, a few sentences wont explain it. —Preceding unsigned comment added by 86.4.152.167 (talk) 09:45, 19 October 2008 (UTC)

Categories of Switches

Managed switches, Unmanaged switches, "Smart" switches. How do they differ? —The preceding unsigned comment was added by 61.31.32.206 (talk • contribs) .

These terms are covered in marketing so exact definitions are hard to find but the following are roughly correct.

• unmanaged switch --A switch with no administrative interface that just acts as a black box gathering information on locations and forwarding packets with port settings either fixed at the factory or set by autoconfiguration.
• managed switch --A switch which has various settings that can be changed and stats gathering capabilities with some kind of management interface to access those functions.
• smart switch --I haven't heared this term before but i suspect it reffers to mechanisms to limit the damage that miscreants can cause by automatically blocking suspicious behaviour but i don't know for sure. Plugwash 01:36, 6 March 2007 (UTC)

• smart switch is definitely a 'dumb' marketing term--Boscobiscotti 06:56, 12 May 2007 (UTC)

• ok I take that back. I have not seen that term. But I have seen "web-managed" switch see:

google search web-managed switch AFAIK, this term is used for a switch which allows (like 'smart switch') a subset of management functions, and only provides a web-interface for management (no shell/CLI access, no SNMP) I think we should add this term to "Categories"--Boscobiscotti 04:28, 25 May 2007 (UTC)

see also discussion of web-managed switches as usual its a fairly ambiguous term which means whatever the vendor wants it to. It fits a niche "cheaper than a good managed switch, but I want to manage the basics" I have seen these from Dell, and HP--Boscobiscotti 04:38, 25 May 2007 (UTC)

Forwarding Method

Is the description of "Fragment Free" correct? I'm now aware of it checking the outgoing media. Fragment Free checks the first 64 bytes of a frame to ensure the frame is not a runt - a frame that collided with another frame and so transmission was stopped and the entire frame will be resent. —The preceding unsigned comment was added by 138.162.128.42 (talk • contribs) .

Fragment free is also an essentially obsolete technique, as well as cut-through switching. These methods made sense with stop-and-wait transport protocols, and when the media speeds were equal on both sides of the switch. In general, however, far more efficiency comes from aggregating 10 Mbps traffic onto 100 Mpbs trunks, 100 Mbps onto Gigabit, etc. Hcberkowitz 23:36, 25 June 2007 (UTC)

Cut-through is not obsolete, it has come back. It is claimed as a feature by at least one manufacturer of 10 Gb/s Ethernet switch chips.

I'd want to see a very careful analysis of why that manufacture thinks it's a useful feature. I could see some scenarios where it might be of use with both 10 Gbps in and out, and also with the traffic mix being primarily jumboframes (i.e., 9Kbytes or so; it's not a rigorously defined term). Hcberkowitz 02:44, 26 June 2007 (UTC)

Obviously cut-through only applies to same speed in and out links. As for useful, you didn't restrict it that way in your previous comment when you said it's obsolete. Instead, it appeared in 100 Mb/s Ethernet, disappeared with 1G Ethernet, and recently came back. I believe it' s useful, but it's too early to argue about commercial success; instead we can describe what exists, and it does. By the way, as I understand it, jumbo frames are not standardized for 1G Ethernet, but are standardized for 10G Ethernet. Paul Koning 14:18, 26 June 2007 (UTC)

You'd be surprised how many people do not realize that cut-through requires equal ingress and egress speeds. It's obvious only after doing the calculations. Hcberkowitz 16:03, 26 June 2007 (UTC)

As for "fragment free" -- I've never seen that term, but if the description above is correct, it is something that every correctly designed switch does. Maybe someone thought it had value as a marketing buzzword, but it doesn't seem to have caught on anywhere, for obvious reasons. Paul Koning 01:10, 26 June 2007 (UTC)

A properly designed switch that does not try to absolutely minimize latency -- when latency is worth the added complexity -- doesn't necessarily do either fragment free or cut-through. Runts are most likely when you have true shared media and collisions, which is fairly rare given full-duplex 802.3 and port-per-device providing collision avoidance. Hcberkowitz 02:44, 26 June 2007 (UTC)

My point is that an 802.1d compliant switch cannot forward runts (fragments). Obviously, the 802.3 MAC layer will generate runts on half duplex links if a collision occurs. I took "fragment free" to mean "this switch doesn't forward runts". Well, duh, of course not -- that's true for all correctly built switches. In other words, "fragment free" is not a real category. Paul Koning 14:18, 26 June 2007 (UTC)

In certain Cisco switches -- I won't even try to track how many "switch" technology companies they acquired -- the original cut-through would try to start forwarding as soon as it read the preamble, start delimiter, and destination address. The fragment free would wait until it saw a full valid header, but, as opposed to store and forward, would not wait to see and validate the frame check sequence at the end. You are correct that a runt can't be forwarded, but fragment free could reduce the number of aborted frames that came out the egress port. Hcberkowitz 16:03, 26 June 2007 (UTC)

This is unfortunately fairly common -- start with a design error, fix the error, and call the bugfix a "feature". Paul Koning 16:22, 26 June 2007 (UTC)

== Need Architecture Discuss

This

topic discusses the basic operation of a switch, but it does not provide a "frame" flow diagram. I came to this topic looking for a basic flow and to learn what the base module and child module is really called. —The preceding [[Wikipedia:Sign

your posts on talk pages|unsigned]] comment was added by Rbmcnutt (talk • contribs) .

Called by who? why do you assume there is a standard name? Plugwash 01:37, 6 March 2007 (UTC)

Manufactures

Why not add the names of some companies that produce switches. Linksys, Cisco, Netgear, and the dozen or so others. —The preceding unsigned comment was added by 64.128.46.234 (talk • contribs) .

• Because of the large number of companies that make switches I don't think it is necesary to add each one of them to this article. If it were only a handfull that did so, that might be different. Thoughts? N. Harmon 15:19, 28 May 2006 (UTC)
  • My thoughts exactly, take a look at the router article for example, it's just out of control really. Adding a list of manufacturers opens up the article to abuse aswell (advertising). Sure a couple of companies could be mentioned in the article, but it should never get like the router article. --Bruce 08:59, 29 May 2006 (UTC)

Picture

Is it relevant to have a picture of a monster rack of switches, rather than a simple 8 or 16-port switch? Compare Ethernet hub. aditsu 22:06, 22 February 2006 (UTC)

Agreed if you have a picture of a simple switch handy put it here and shunt that monster rack off to multilayer switch. Plugwash 18:05, 25 February 2006 (UTC)

That "monster rack" is not a multilayer switch. It's a Cabletron Smartswitch 6000, which is an enterprise level piece of hardware but definitely not a multilayer switch. I assume the request is to put up a picture of a dumb switch which might be found in the typical home. I don't have an issue with this but I don't see that it's any more relevant to the topic than this picture of an enterprise level switch. —The preceding unsigned comment was added by 138.162.128.42 (talk • contribs) .

Ok i was mistaken, my searches and in particular [3] seem to confirm that the switch pictured isn't a multilayer one (it seems that it needs a separate router module which doesn't appear to be present in that photo). Plugwash 23:42, 26 May 2006 (UTC)

i have a switch similar formfactor as the equipment shown in the hub picture (it has 8 ports not 4 but wth). if someone could explain to me how to upload a pic i'll do it. 76.200.101.39 09:06, 4 May 2007 (UTC)

D-link switch is very simple and you should mention this photo for easy understanding by every normal persons and now I am using in my office of 5 port switch and it is very easy to intall and very very easy.

added photos as desired. --akc9000 ^{(talk • contribs • count)} 02:35, 30 June 2007 (UTC)

Bridge vs Switch

Maybe the article could highlight the differences between "Switch" and "Bridge"? —Preceding unsigned comment added by 84.63.10.249 (talk • contribs)

There isn't one. "Switch" is a term invented for marketing purposes by a bridge vendor a number of years ago.

Well, except for Fibre Channel switches, which aren't at all like LAN switches and probably should be described separately.

Paul Koning 18:37, 15 April 2007 (UTC)

At least one of the problems with this article is that the evolution of the local area network is assumed rather than explained. I suggest something along these lines: Ethernet was designed as a mechanism to allow two or more computers connected to specially constructed cable to exchange data. Ethernet was a specific implementation of Carrier Sense Multiple Access with Collision Detection (CSMA/CD). Multiple Access indicates that all of the LAN connected computers are connected to the same physical communications path. Carrier Sense indicates that a computer with data to transmit first listened to see if the communications path was already being used by another computer. Collision Detection indicates that computers that are transmitting data also listen to the communications path to determine if another computer had started transmitting at the same time causing a collision. The laws of Physics and the protocol specification put a limit on the maximum length of the communications path, the minimum distance between the physical connections, and the number of computers that could be connected to an Ethernet LAN. One of the first steps in the evolution of the Ethernet LAN was the creation of a mechanism to extend the length of the communications medium. The device was call an Ethernet Repeater, it connected two physical Ethernet LANs by accepting messages from one, cleaning it up electrically (regenerating something called the preamble) and sending it into the second connected LAN. Repeaters did not differentiate between traffic that needed to appear on the other LAN and traffic that was between two computers that were on the same LAN and that had no need to be repeated. The next step in the evolution of the Ethernet LAN was a variation on the repeater that could determine if traffic needed to be repeated and only passed that traffic to the second LAN segment. These devices were known as Bridges and came in several flavors. Early repeaters and bridges were fairly large and expensive but they made LAN technology practical. As time went by, both bridges and repeaters became smaller and less expensive. This allowed the creation of hubs, a collection of repeaters connected to a common interconnection mechanism. It also enabled the creation of multiport bridges with a set of bridges connected to a common interconnection mechanism. Unfortunately multiport bridges had a well-deserved reputation for being unreliable and hard to manage. About this time routers made their appearance and they allowed networks (local area and others) to interconnect. Back on the farm, bridge technology continued to advance but the old bad press made them hard to sell. Enter the Mickey Marketers and enter the Ethernet switch AKA the micro segmentation bridge.

With this as background we can go on the discuss things like collision domains, QOS, VLANs, and the rest of the ideas good and bad. Docdave (talk) 02:07, 30 September 2011 (UTC)

Heterogenous switch - fact or fiction?

The article currently claims that a switch can switch between token ring, FDDI and Ethernet. Those 3 nets have different MTUs, for one thing. Is anyone really doing this in practice, or do they all use routers? --Alvestrand 18:02, 26 May 2006 (UTC)

First, very few people use FDDI or Token Ring any more. Yes, I have used such things as source route translational bridging on devices called routers but configured to do layer 2-ish things. Ethernet to Token Ring, at Layer 2, is exceptionally ugly. It's not just the MTU difference, but that you have source route explorers learning Token Ring topology while you have Spanning Tree BPDUs doing Ethernet topology learning. I can only say that it was not a question of how well the technique worked, but that worked at all. Any sane network architect, who wants to go between two significantly different media, routes rather than bridges. When hardware was much slower, bridges offered some cost-performance advantages, but that really is no longer an issue. Hcberkowitz 02:44, 26 June 2007 (UTC)

I thought I'd mention a couple articles I came across while joining you on your quest. Take a look a the following documents [4] [5] which I found on cisco.com. I also quote the CCNA Companion Guide: "Bridges can also be used to connect dissimilar protocols and media". I however can't provide you with a well thought trough answer, perhaps I'll ask one of my professors. --Bruce 20:03, 26 May 2006 (UTC)

Unless the CCNA Guide in question is referring to things like SR-TLB, which I'm not sure Cisco even supports any longer, I'd call that statement wrong.

Well 802.11 also technically supports larger packets than ethernet, its just not usually set up to use them so it can be bridged to ethernet. I'd imagine similar constraints would apply to interoperation between nodes on token ring and ethernet. Plugwash 23:32, 26 May 2006 (UTC)

I picked up a hint in a rfc i was perusing that 802.1 may be relavent to this discussion but i haven't had a chance to read it yet. anyone feel like doing so? Plugwash 00:38, 6 March 2007 (UTC)

802.1 does deal with bridge architecture, but I don't see any good reason to force bridges to do what routers do better. Hcberkowitz 02:44, 26 June 2007 (UTC)

I'm not sure, but certainly this is done as mentioned with 802.11. I think this is usually referred to as protocol conversion. this is a part of the article which needs improvement.--Boscobiscotti 02:04, 22 May 2007 (UTC)

There are a couple of cases here. (1) same frame format but different MTU: regular Ethernet to jumbo frame Ethernet, or Ethernet to jumbo frame 802.11. That's a matter of tossing oversized frames. Upper layers avoid the issue by using the smaller MTU. (2) same forwarding approach (transparent bridging) but different frame format. This applies to Ethernet/FDDI bridges. Those definitely existed and worked well. The MTU difference was the biggest difficulty. The frame format didn't matter much since most protocols used Ethernet framing. A few oddballs, like Appletalk and Novell's nonconforming "raw 802.3" were additional hassles. (3) different forwarding approach: Ethernet to 802.5 bridge. 802.5 (normally) used source routing, not transparent bridging, while everyone else uses transparent bridging. Heroics were invented to cope with this. I think they occasionally worked somewhat but not well enough to be commercially successful. This may be one of the reasons why 802.5 is dead (though certainly not the only one).

Yes, hardware speeds are one reason why bridging was used where nowadays you might prefer routing. But there are others: bridges (ignoring 802.5 SR) are transparent and plug-and-play. Just plug them in, they work. Routers require configuration. Not necessarily all that much, but it's clearly more work and more planning. They have benefits too, of course. Paul Koning 14:25, 26 June 2007 (UTC)

Request move

• Network switch → Switch (networking). Rationale: I'd like to be bold and simply move the article but I feel a consensus would perhaps be beneficial. I believe adding a category indication to the end of the article's title would improve clarity, it would also be much easier to organize other articles this way. --Bruce 23:56, 26 May 2006 (UTC)

Voting

Add *Support or *Oppose followed by an optional one sentence explanation, then sign your vote with ~~~~

• Oppose, bracketed disambiguation is an ugly last resort, network switch is a perfectly good title. Plugwash 01:04, 27 May 2006 (UTC)
• Support: in the context of networking it's always referred to as a "switch" and sticking with the common name (per WP:NC). After picking a dozen links or so and seeing how it is used in context then it's clear that the most common use is "switch"; look at the articles that use it: [[network switch|switch]]. —The preceding unsigned comment was added by Cburnett (talk • contribs) .
• Oppose. There are many switches (electrical switches, railway switches) that are not network switches - so when "switch" is used alone to mean "network switch", it's because it's in a networking context. --Alvestrand 21:22, 27 May 2006 (UTC)
• Abstain - The Router article is listed under just "Router". Hub takes you to a disambiguation page where you can click on Ethernet hub. The lack of consistency here might suggest you go ahead with your move. N. Harmon 15:21, 28 May 2006 (UTC)
• Abstain - for now. Look at the Switches (Disambiguation) page. The computing / electro-mechanical switches generally dont use this form, while the music/media "switches" do. If you change this one, you ought to change some of the other switches too. Are there any guidelines for this? I do suggest this page should link back to Switches (Disambiguation). IanB 14:58, 5 June 2006 (UTC)
• Oppose makes no sense at all to move it. just create a redirect to here. --akc9000 ^{(talk • contribs • count)} 02:39, 30 June 2007 (UTC)

Created redirect from Switch (Network) to here. I would think this would solve the issue. --akc9000 ^{(talk • contribs • count)} 02:45, 30 June 2007 (UTC)

Discussion

Add any additional comments

Well so far we have 2 supports, 2 oppose and 2 abstain (who do feel that the issue should be looked into). I'm quite busy next week, but as soon as I get some spare time on my hands I'm going to start looking into the situation and try and figure out what can be done. --Bruce 07:23, 24 June 2006 (UTC)

Bruce I dont see the issue at all. I just created the redirect. I hope this solves your concern. --akc9000 ^{(talk • contribs • count)} 02:46, 30 June 2007 (UTC)

Calculation about switch

How to calculate the width of the memory data bus, how to calculate the minimum clock frequency in turn. what is the input queue architecture, how it affect the performance of switch?

These concepts really don't apply to high-end multiprocessor switches, or even lower range ones that use specialized ASICs. Hcberkowitz 23:33, 25 June 2007 (UTC)

weird sentence

"Performance specs? - Switch Fabric (definition needed) - MAC table size - RAM buffer size - Network Protocol and Standards - optional ports (fiber, SFP Expansion Slots, etc) - auto port speed detection, configuration" - that part looks funny... not really in a good way though ;) --bb 22:00, 15 December 2006 (UTC)

Alert--wrong link

The term "repeater" is linked--however, to something else, from electronics. It seems like it should be linked to the article called "Multiport repeater", by all logic. Also, the article "Multiport repeater" should be added to the disambuguation page for "Repeater". --Sukkoth 10:50, 26 January 2007 (UTC)

I made the suggested correction. I did not change the repeater article but note that multiport repeater is listed in the See also section of that article. ✤ JonHarder ^talk 14:05, 26 January 2007 (UTC)

removed from "types of network"

this part

A multiport repeater is the simplest multi-port device in use. However, its technology has been considered outdated since a hub is a "dumb device", as it resends every datagram it receives to every port except the original incoming. With multiple computers, the speed quickly slows down, and collisions start occurring, making the connection even slower. However, with the advent of the network switch, this problem has been solved.

Because this article is about "network switch" and this is a description of a hub.--Boscobiscotti 06:58, 12 May 2007 (UTC)

removed this section

from section on switch monitoring:.

Other methods (which could be classified as attacks) have been devised to allow snooping on another computer on the network without the cooperation of the switch:

• ARP spoofing - fooling the target computer into using your own MAC address for the network gateway, or alternatively getting it to use the broadcast MAC. This tricks the switch into sending your computer another computer's data.
• MAC flooding - overloading the switch with a large number of MAC addresses, so that it drops into a "failopen mode".

the two pages are previously mentioned in the article. #1 is not so relevant to switches anyway, #2 is, but has already been mentioned. --Boscobiscotti 02:03, 22 May 2007 (UTC)

Vast majority

" ...plays an integral part in the vast majority of current home and office Ethernet local area networks" Umm, don't most home Ethernets, thus the majority of all Ethernet LANs, merely consist of a router connected to a CATV or DSL modem and one or a few computers? Without, that is to say, a switch? Jim.henderson 03:22, 22 May 2007 (UTC)

yeah, well I was thinking that the majority of home ethernet (other than wireless that is) "routers" are really a 4-port switch glued to a a router. But your point is well taken. it will undoubtedly confuse people. I took out the home reference.

see the linksys "basic networking" page: http://www.linksys.com/servlet/Satellite?c=L_Product_C1&childpagename=US%2FLayout&cid=1118334622279&pagename=Linksys%2FCommon%2FVisitorWrapper&lid=2227922279L01 looking at that page, we can see that a number of home networking devices, are now a vpn firewall glued to a router, glued to a switch :)--Boscobiscotti 04:53, 22 May 2007 (UTC)

Hmm could say something like "many home DSL and Cable routers include a small embedded ethernet switch" ??--Boscobiscotti 04:57, 22 May 2007 (UTC)

All right; with "integral part in the vast majority of current office Ethernet local area networks. In addition many home DSL routers or Gateway devices contain a small (4 - 8 port) embedded Ethernet switch" it's an improvement, but it still fails to think small. Think of all the dental offices, small elementary schools, auto body shops and other small businesses with a few computers Ethernetted together with no other network box than a router and DSL modem. Those may not make a big blip in a survey of business computer fleets, but surely they are at least a big minority of LANs. Jim.henderson 14:19, 24 May 2007 (UTC)

If more than one computer is "ethernetted together" even in teeny office There has got to be something doing that - I'd give you 20 to 1 odds that the lan device embedded in the router or dsl modem or "gateway" box is a switch. or there is a small switch connected. But we could certainly add "soho network if that helps, and also enterprise.--Boscobiscotti 04:22, 25 May 2007 (UTC)

Yes, that's an improvement but maybe we can dream up a way both simpler and clearer to say it isn't usually a box or other thing someone should buy or worry about. Rather, LAN switching is usually just an invisible feature of another box, namely a router, making what was called a "Brouter" back in the day when this particular product convergence was a novelty and "Bridge" had not been dropped in favor of "Switch". Nowadays LAN switching happens, but almost everyone who sets up an Ethernet (the few who administer a big one won't be looking in an encyclopedia for advice) doesn't have to know it is happening. For them, it's nice to know that once upon a time the topic was worth thinking about. Jim.henderson 19:07, 25 May 2007 (UTC)

Form of power source

has "standard" and POE -I think this is a bit confusing. AFAIK switches typically provide POE to low-power devices, like APs and IP phones. - but its rare for a switch to use POE. so maybe this heading should be changed - I would call this a switch feature. comments?--Boscobiscotti 05:21, 22 May 2007 (UTC)

Expert tag

This article needs help from an expert. I am spread a little thin so I am tagging this article. Please do not remove the tag unless an expert re-vamps this article. This article should have data in it so a general reading can understand what a network switch is. It should talk about or at least show little consumer based switches (not hubs) before addressing the monsters. --akc9000 ^{(talk • contribs • count)} 04:47, 26 June 2007 (UTC)

Time for a different approach?

I wonder if rewriting/editing is the best solution.

"Network switch" is a marketing term. It has no precise meaning. Over time, it has been used for LAN bridges. It also has been used for routers ("Layer 3 switch", sic). It has had adjectives attached to it to create distinctions (real, spurious, whatever) from "other" switches/bridges/routers -- like "smart".

Maybe we should just have a disambiguation page, or something like it with a just a bit more text in it. Something like:

Network switch is a term used for various layer 2 or layer 3 network devices. See network bridge for the layer 2 device, router for the layer 3 device. Some of these devices combine the layer 2 and layer 3 functions, in which case they act partly as a bridge and partly as a router.

Paul Koning 14:50, 26 June 2007 (UTC)

I just saw your suggestion - I had just made a similar one below. An article on "lan switch" (as cisco calls it) or "ethernet switch" (another term I have heard) to refer to layer-2 only devices could be referenced by a general "network switch" article. I hesitate to completely equate a bridge and a "lan switch" or layer-2 switch, since the managed variety of layer-2 switches allows configuration of VLANS, STP, etc. But then my only experience with bridges is with 802.11 bridges. I understand the forwarding occurs similarly, but I am not sure what features a "typical bridge" might include vs a typical "lan switch"--Boscobiscotti (talk) 23:01, 8 January 2008 (UTC)

Can we replace the hub with the switch?

Can we just replace the the hub with switch from a working network without changing the network configuration?

Often, but not always. One reason can be that undersized servers, previously connected to hubs, may have been working at the limits of their capacity, and full-duplex traffic may overload them. I prefer to change from hubs to switches in a controlled test, monitoring server performance carefully.

Is this not a somewhat acedemic concern? For one, you could force the switch port to half-duplex on many switches. And even when replacing a switch by another switch some monitoring should take place. Pgallert (talk) 12:30, 19 March 2008 (UTC)

Unless you can make the switch turn into a multiport repeater, no. If the purpose is to be able to see everything on the network, then you could setup a span port to mirror all traffic through that port. - Tbsdy lives (formerly Ta bu shi da yu) ^talk 11:57, 10 June 2009 (UTC)

Other reasons hubs may do something different from switches is in network management. Typical switches have a single mirroring port. In applications such as sinkholes, where more than one monitoring device is needed, I've used hubs as effective fan-out devices. Howard C. Berkowitz 13:27, 6 July 2007 (UTC)

But the network would still be working, so the answer to the question of this paragraph is "yes". You could still connect the now obsolete hub to the switch mirroring port and attach multiple monitoring devices, couldn't you? Pgallert (talk) 12:30, 19 March 2008 (UTC)

I happened to run across this page from a google search. I read a few paragraphs and thought the sentences sounded stilted, sort of like you might write if it was an ad.

So I googled a couple of the sentences. You might find it instructive to do so yourselves.Ilbob 11:46, 19 August 2007 (UTC)

Could you clarify what sounds stilted, and what you were trying to find with Google? Howard C. Berkowitz 13:05, 19 August 2007 (UTC)

I had the same feeling, not with this discussion page (which seems to be profund) but with the article on LAN Switching. See for instance the paragraph about history; that one is utter nonsense.Pgallert (talk) 12:30, 19 March 2008 (UTC)