Hubs, Bridges, and Switches (oh my!)

good morning everybody and welcome back to next door netadmin today I thought that I would do something a little different rather than talking solely about current items I thought we would go and take a stroll through history Lane as it were this comes out of last week's video where I was talking about the OSI model and all the the various layers that we went through and I realized that there's a lot of terminology that uh Network engineers and admins kind of just throw around because we know what it means but if you're starting out you may not know how this term became a thing what it means what the context is in terms of what it means to everybody who's been around with it for quite a while so that's why I thought I'd just go through this and cover some of our older Technologies they may not be currently useful they may not be in use anywhere and some of them I dearly hope they are not because we have stuff that's so much better now but with if we go through that then hopefully it gives you more insight into the the equipment that we talk about the terms that are in use and it can enhance your knowledge and understanding of what you're working with even in a modern context cool so what we're going to be talking about today is mostly at layers one and layers two that was the physical layer and the data link layer the physical layer all we're going to be talking about is cabled we're not going to be talking about radio today because that's its own thing completely um and Layer Two while we're going to talk about layer 2 kind of in abstract most of what I'm going to be talking about will be ethernet based but did you know ethernet didn't just use to be the cables that we work with today what you would call twisted pair ethernet as a standard started out with coax cables there was thick net and thin net mainly having to do with you know how thick the cable was thick net had more shielding and so it could run for a further length um and thinnet was thinner easier to work with but because it had less shielding it had greater attenuation and the signal would not travel as far this was important because both thin net and thick net with the coax cables ran as a single bus this is something that I mentioned last week as well it used to be a bus architecture and still is fundamentally but you would literally have a cable that ran from one end of your network run all the way to the other that was the network run and then each computer would have essentially a BNC connector that would con to a t-shaped tap that would tap into the line remember that word it'll come back later but when every computer has a tap into the same bus again how how does this work how does how does your traffic signaling work well something that we mentioned last week was csmac CD and I said it wrong in the video but if you read the description you'll know I caught my error afterwards whoopsie but Collision no that's wrong carrier sense multiple access with Collision detection there we go um carrier sense there's a wire there yeah okay multiple access there's multiple comp computers tapped into this line we got that too Collision detection is the important thing when a station transmits onto the line before it transmits it's going to listen first to see if any other station is transmitting on the medium but there's always the risk that two stations try to transmit at the same time when that happens they detect that a Collision has occurred I'm still transmitting and I can hear that somebody else has just started transmitting too and when that happens there's been a collision on the network this is important because early ethernet was also half duplex half duplex essentially means that you can only transmit or receive you can't do both at the same time and so this is a little bit like walkie-talkies if you've ever had one walkie-talkie and and you you talk into it that's great everybody else can hear you if you try to key up two walkie-talkies on the same frequency what do you get you get an awful awful squeal because there's two stations broadcasting at the same time and everybody else is receiving both and it just washes it all out into an awful awful squeal something similar happens on a bus architecture a half duplex bus architecture for computers and so if they detect that a collision has occurred they know that nobody else can understand what's happened and so both stations when they detect a collision will emit a jamming signal a jamming signal just to indicate the this has failed and both of them broadcast so that they can both detect that the jamming signal is being broadcast and therefore a collision has occurred okay that's cool all right so we've got thin net or thick net it is a single bus it is half duplex and everybody has a tap onto this the same bus line okay cool this is where it starts what happens if you want to extend your network both of these have length limitations so what do you do well the simplest thing you can do at layer one is what was the length limitation in the first place the signal got too weak well if you can just increase the strength of the signal then you can keep the the network path going and that's where we get a repeater very basic if you've used ham radio at all amateure radio then you're probably already familiar with the concept of a repeater a network repeater Works in much the same fashion it it hears a signal it increases the signal strength and rebroadcasts it crucially though like I said this is at layer one the repeater does not need to understand anything about ethernet it doesn't have to understand anything about IP it doesn't have to understand anything about HTTP telnet port numbers any of that junk it just worries about I'm receiving a signal I'm putting it out the other end at a stronger signal strength that's it you can do the same thing with fiber optics incidentally you can get Optical repeaters that increase the the light strength and rebroadcast that and that is something that is necessary for example in transatlantic cables because it's it's a long distance and the light won't won't come out the other end at nearly the same strength as it went in so you need Optical repeaters but the repeater is just an extender no more no less okay there's still a maximum distance that you can run with this and the maximum distance total comes down to how long does it take at the minimum frame size for ethernet if a station at one end of the line starts transmitting and a station at the other end of the line starts transmitting at the same time they both need to detect it before they've reached the end of their minimum frame size so that they can emit the jamming signal if they both transmitted at the same time so if you calculate this out at essentially the speed of light the speed of electricity transmission you get a maximum length that you can actually run your network even with repeaters in play and this forms what we call a collision domain a collision domain is a segment where any Collision can be detected and reported correctly Okay cool so you have repeaters and you have a long you know coax cable all right well then we increased our understanding of technology and developed twisted pair of cabling which is what we're still using now there's a couple of different variants by the way you can have shielded twisted pair foil shielded twisted pair um unshielded twisted pair or UTP is the standard that is commonly used so cool but with a twisted pair cable we moved from a single bus architecture to a star topology where we have a bunch of cables going to end points and they all converge at a central point or an extended Star topology where you have multiple you know concentration points and those concentration points are linked a repeater only has one in and one out so this isn't going to be useful in a star topology enabled by twisted pair cabling you now need a multi-port repeater which also became known as a hub hubs also work on layer one it's accept a signal in one port broadcast it out all the others at increased signal strength and you're done that's it so that's cool we can now connect more machines without having to tap a single physical line that has to snake through all the desks back and forth that's cool but would it kind of be cool if we could extend the length of the network which has a fixed size because we have a a a maximum size of our collision domain would it be kind of cool if we could have kind of two networks two different Collision domains how how do how do we join that well we can Bridge them and so a new device is Born the bridge the bridge works on Layer Two why does it work on Layer Two well because it's separating Collision domains Collision domains are a function of ethernet which means that it is a layer two concept and so if you're going to bridge two different Collision domains you're operating at Layer Two a bridge is very simple it listens to traffic going through on one side and when it hears somebody transmitting something for a station that's on the other network it passes it over and if it doesn't have to pass it over it doesn't the rest is just a bus architecture on either side they can just keep talking to each other only the traffic that has to be bridged will flow over the bridge cool can this bridge two different layer 2 Technologies yes it absolutely can there is nothing to say that you couldn't use an IP address that you're talking to somebody because that's at layer three you could be on ethernet the other station might be on token ring it might be on frame relay you don't need to know that all you need to know is a consistent way of addressing it with a TCP port at layer 4 IP address at layer three layer two is determined by your local network because it's got a maximum size that you can practi tactically run so you might have a bridge that is listening to ethernet traffic going back and forth and when it hears something that kind of needs to go over to the frame relay side it will Bridge those two different networks and broadcast it out onto the frame relay side that is something that you can do it's kind of cool when you you think about it most of the time you won't because most networks are ethernet these days it's just how it is okay so you've still got two big Collision domains here if we're talking about you know half duplex you can also get full duplex where your equipment knows that you can transmit and receive at the same time if you can transmit and receive at the same time that's a full duplex connection and it speeds things up dramatically because people can send and receive at the same time but on a bus architecture only one station can be transmitting at a time even if it can receive at the same well I take that back only one pair of stations can be transmitting and receiving at the same time is what I want to say when you came to full duplex communication and cool what happens if you're talking to a station that doesn't understand full duplex well if you're using autoc configuration it's going to default back to the lowest common denominator which is a half duplex connection so you've still got even if you are technically full duplex capable if autoc configuration doesn't work or if you have a device that just doesn't understand full duplex you're still going to fall back to half duplex okay so you've got these two Collision domains where you might have half duplex connections you might have full duplex connection we don't know and when it needs to cross between Collision domains it'll flow through the bridge what if we could shrink the Collision domains even further this is how we get to a switch a switch fundamentally Works in much the same way as a bridge does it accepts traffic in one port but then it's f segmented you're not going through any hubs or repeaters along the way the switch kind of takes the place of all of these and it will if it doesn't know where the destination is on which Port it has to send it out all of those ports because it doesn't know where the destination station is but if it does know where the destination station is located then just like a bridge it will only send it out the destination Network side this means that you can have two stations talking back and forth with each other and none of the rest of the network has to see it which means you've dropped the size of your Collision domains dramatically because the switch also operates on Layer Two And even better than that because the switch is sitting in the middle between these two stations typically speaking your collision domain is actually just the single wire segment because the end station is talking to the switch and then the switch talks back to the station by relaying the traffic and so at layer 2 your collision domain has basically dropped to a single wire and you won't have collisions anymore unless you happen to be operating in half duplex for some reason and if you are then you probably have a problem that you need to fix gigabit gigabit kicked this up by a lot twisted pair cabling has four pairs of cabling two wires each okay cool both 10 megabit and 100 megabit used one pair for transmission and one pair for receiving transmitting receiving and then the other two pair were ground and it was actually designed this way so that you had uh phone lines essentially had two ground and two data that that was it so four wires in a phone cable you could actually rig up your ethernet your eight pair excuse me four pair eight wire um cable so that you could plug a phone line into the middle of it and it would work you could just transmit a phone line over that and then the other two on the sides could be used to create another phone line if you wanted or in the case of ethernet you you'd have a couple of unused pairs that could be used for ground providing power Etc gigabit was enabled by essentially advanced electronics that could make better sense out of what was happening on the wire electrically speaking because with gigabit you're using all four pairs in both d directions at the same time with gigabit each pair of wires is carrying 250 megabits of traffic at the same time full duplex both directions at the same time so the electronics has to be good enough to separate its own signal out from what is being received and understand how those two different electrical waveforms will balance each other and cancel each other out or just increase each other to dramatic levels it has to be continuously error correcting in order to make heads or taals out of what the other side is sending so this was not something that could be done on at least certainly not off the get-go it took quite a bit more development to get to that point but all of this that I've talked about has happened at layer 2 or layer one and I mentioned way back a ways that in the original bus architecture each station would have a tap into the main bus line that terminology is still used today if you look at some of the I want to say openvpn definitely uses this terminology I think Linux uses it sometimes as well but you can see Concepts like a bridge you can also see an openvpn I'm going to use it because it's the perfect example for this one of the available interface methods is tap and when you're using a tap interface you are running at Layer Two so that any layer 2 traffic coming in one side will get sent through and emitted onto the layer 2 Network at the other side that is a tap interface you've created a virtual tap into your bus on the other side so the terminology still exists the concepts are still useful even if we've moved past that architecture to more modern forms it's cool I do plan on continuing with this discussion and we'll see about how you know routers function in more detail the last thing I will leave you with though is a caution because the termin the terminology that I have explained here the device names and everything are what we commonly understand it to be there are some companies that decide to redefine what it means for their own benefit kind of like how Tesla tells you that it's full self-driving and it's not they've redefined it by sticking a trademark symbol on the end of it capitalizing every word and say oh well we don't mean that it's full self-driving it's full self-driving so one of the uh networking appliances I've worked with before is untangle um and this is what I'm familiar with so I'm going to use it as an example but I have no doubt that this happens with other filtering appliances as well one of the modes that untangle offered was a transparent bridge mode it's a if it's a bridge you expect that what you send in one side will flow to the other if necessary on Layer Two and it doesn't care about the layer three whatsoever you could be using IP you could be using ipx SPX you could be us using something completely different it doesn't matter it's a bridge it's only operating on Layer Two but that's not how untangles bridge mode works and it's something that they had to put into their documentation and it would continually trip up newbies myself included at least until I understood this that the way that the filtering worked is it would accept packets or frames in one side destroy them pass the traffic up through the filtering Appliance and then it would come back down on the other side and it would recreate the whole thing and send it out the other side that's not operating at Layer Two and in fact it can't in order to make that architecture work but the key point to this was where a bridge does not care about your layer three addressing untangle very much does because it's going up through the whole layer stack and coming down the other side because of this it had to be emphasized even in bridge mode everything this does is routed it's operating on layer three because it has to even though we've used the term Bridge because that's kind of what it looks like to people who don't know any better it's not actually functioning as a bridge and if you slapped it down and gave it to a network engineer who doesn't know anything about how it's working and say this is in bridge mode you expect it to just pass traffic through and you don't need to tell it where all of your layer 3 networks are if you do that it doesn't work so just be cautious even if you understand what the word means what it's supposed to mean the the equipment that it came from that doesn't mean that everybody's going to use it the same way you got to watch out for uh for that kind of spontaneous redefinition shall we say but like I said I plan to talk more about this in future talk more about layer three and and higher levels because you can get into lots of lots of other details for now though I think that's probably enough to to chew on for one week and next week is the free form so we'll probably pick this up in two weeks from now okay well that's it for today then thank you very much for watching I am your next door netadmin and we'll see you next time

Channel: NextDoorNetAdmin

Views: 27

Keywords:

Id: 5pUSZ7AK26M

Length: 27min 48sec (1668 seconds)

Published: Mon May 20 2024