CCNA Training - Quality of Service (QoS)

[Music] [Music] [Music] [Music] [Applause] [Music] [Applause] [Music] all right good evening everybody we have a lot to go over tonight so we're trying to dive right into it big pop house hey good to see again good yeah you're already watching the Nuggets for qos good so this will be a great conversation be sure to chime in with any questions you have as we go about it so um yeah first order on the agenda was me picking my drink I couldn't decide between coffee and sparkling water tonight usually it's a sparkling water thing but I was just cold so I got coffee and I got sparkling water so there we go we're ready for whatever this night brings I want to take a quick look at the whoops not the agenda that's not updated haha wrong button there we go we want to take a look at what exactly cisco has in mind for us to learn about QoS for the CCNA and the CCNA I'm not gonna lie this is a lot and we're gonna make it so that we have a lot to cover tonight and yeah we're gonna get as far in as we can we might go a touch over we'll see it just depends how quickly you make it through it so don't hesitate to ask your questions as we go through but we're gonna start talking about just basic what the need of QoS is and then we'll just kind of go systematically through what Cisco wants us to know so we'll talk about marking we'll talk about trusting prioritization shaping policing and this one right here I will say congestion management that looks pretty innocent at the bottom of the list but that is a I mean as I was preparing for this I think I had as many you know a through F sub categories under congestion management as I had under every you know under QoS in general so definitely a lot to talk about let me go ahead and pull up the whiteboard here and I think we're good to go so you know as usual again chime in with your questions kind of got that agenda set that we're gonna go through quality of service here but like I've mentioned several times we kind of got to to roll with this - all right so why QoS and what is QoS let's just start with the absolute basics so imagine I will tell you this a home use perspective there are days that I'm playing my video games online right because I do that and I'd like to compete and I get very competitive and of course inevitably while I'm in the middle of a match I get this lag in this slowdown and I run out to the living room and find out that you know of course not only are my kids streaming but also you know my other child is streaming on a tablet and then maybe my wife is streaming on her tablet her phone and like we're consuming all this bandwidth it's like you know an online game doesn't require that much throughput so if we think about if we have a small amount of data going through a single pipe and then we also have a large amount of data going through and this amount of data right here let's call this time sensitive it absolutely matters how long it takes between the time I push my button and the time that the videogame registers to the server that I pushed that button it's very time sensitive meanwhile something like Netflix which is constantly buffering or any streaming service really that's constantly buffering that you know it doesn't really matter if this little bite right here gets through before all of these bytes do and so quality of service we apply this to the app to the enterprise world and yeah we're not playing video games at least we're not supposed to be playing video games hold on the job usually the biggest place that this comes into effect is what we call voice over IP or voice so this would be I'm picking up my phone on my desk and it's an IP Phone meaning it's going through the network not through the traditional telephone lines realistically at this point probably over 90% of infrastructures are built on IP telephony so in all likelihood if you're at your work and you're at your desk you're picking up your phone you're going to talk now like the video game situation talking on the phone it doesn't require a whole lot of bandwidth in fact it's like kilobits per second it's really miniscule and so when we look at the fact that okay well let's say I'm trying to have a phone call but my neighbor is trying to upload or download a 20 gig file from the network share okay same situation right as the video games in the stream it's the exact same scenario where I've got a small stream again just like this that's time sensitive it doesn't do you any good is you know if I you know if I skip a word you know I said you know with because that packet gets dropped or what-have-you it doesn't do any good to resend that packet it doesn't do any good if eventually it makes it to you it needs to get to you on time otherwise you start garbling like that right no your twitch is not messing up that was just uh just all for demonstration purposes so we have this problem a lot in the enterprise and it's not just voice over IP it could be mission-critical you know traffic in the data center V motion can be a high priority traffic but it can also be a heavy bandwidth you know for those who have data center experience so whatever the situation is we need the network to be smart enough to say hey you're higher priority traffic you should go in front of the low priority traffic so it's this concept of prioritization where we set priority levels and we say okay really mission-critical data maybe is up here and maybe we give it a number of seven and then all the way down or maybe stretching all the way up from seven all the way down to zero this is the highest priority the highest of the high and this is the lowest of the low priorities and so network traffic that's going through the router if the router can only pick one or the other you know because of bandwidth utilization well we wanted to pick the high priority traffic that makes the most sense so if we can configure our networks to be smart enough to do that then boom we're good so here's the problem because that's a very simple concept right choose you know small kilobits per second voice over IP traffic over you know my you know neighbor streaming Netflix on the job whatever is happening executing it is actually very difficult our networks are complex and we have routers and we have switches and we have routers connected to routers and we have routers connected to switches which are connected to other switches which are did two routers and maybe this pipe over here this is huge that's a 10 gig pipe but this LAN circuit over here this is like 500k well probably not in the this day and age but maybe depending on where in the world you are maybe it's more like a 5 mag let's just call that say it's a 5 mag connection that's pretty realistic for a branch site so we've got all of these disparate type connection types we have different architectures because we have routers and we have switches we even have we have Cisco routers and we have juniper routers heaven forbid write down my CCNA draw we've got Cisco switches we've got different platforms we've got Cisco Catalyst switches we've got cisco nexus switches and all of these platforms do QoS differently and that can be one of the biggest frustrations when you're learning QoS really less about learning the concepts fortunately CCNA we just need to learn the concepts of QoS and so we're gonna spend most of our time tonight really all of our time tonight going through all the concepts of QoS however when it comes to configuring QoS okay configuring QoS on a router looks really different from configuring that QoS on a catalyst switch and believe it or not configuring on that catalyst which looks really different from configuring on a nexus switch and that looks even different by the way then the newest catalyst switches they're different than the old catalyst switches and one thing that before we move on here just to help everybody understand that is that quality of service doing these things with QoS it requires tying into the hardware ok there are these Asics we call them application-specific integrated circuits so when i've got a physical switch I have physical chips inside that switch and the way we queue traffic the way we manage traffic queues and such which we're going to talk about tonight the way that all happens is usually whoops built into the hardware meanwhile routers by the way these are software machines so routers the good news about routers they're extraordinarily flexible we can do almost anything we want from a tos perspective on a router but when it comes to switches it's very reliant on those underlying Asics and again the Asics that were developed in old Cisco switches are different from the Asics that are developed for new Cisco switches which are different than the Asics that are developed for the data center for those nexus switches and this is why the implementation of QoS can be such a nightmare for all of us unfortunately so just understand that that when it comes to configuring it there will be some hurdles there but regardless with CCNA fortunately we don't really need to know too much about that we simply need to understand a lot of the concepts so the first concept that we need to understand is when way back I should I should step even further back than that but you know way back in the probably in the 90s I don't really know what what year it was there were two different theories on how to do key OS okay so when we think about it what have we used to have it we primarily have talked about this a little bit right when you look at circuit switched networks versus packet switched networks if anybody took their C C and C sent sorry if anybody took their C sent icnd1 you probably studied that right the concept of a circuit switched phone is like old telephones right where if I get on if I get on the telephone here I have a dedicated circuit to somebody over here and networking the concept of networking was well we could do that like every time a computer needs to talk to another computer I get a dedicated line to that other computer I mean that was modems basically right you know going over the phone lines but as we evolved our concept of exchanging data we realized that a packet switch network works a lot better because in the phone scenario yes this is good because I have dedicated bandwidth but this phone right here wants to call this guy and he's getting something we've none of us have heard for a decade as a busy signal right because we have no available circuit for that third forum telephone meanwhile in a network this concept of packet switching says that hey you know what even though these three you know say these two computers are talking to each other this computer needs to start talking to this computer as well and that can all happen because we're sending packets now and these packets can get interspersed and interlocked interweaved with each other so that the old so that the the destination computer can receive both communications at once okay so this is circuit switch versus packet switched and and a concept that we need to understand is that QoS theories we have two different theories around QoS and we actually had protocols that were developed around both of these the first one resembles the the top diagram with the phones a circus switch concept we had a protocol called the resource reservation protocol which is RSVP and RSVP really what I should have said first it belongs to this theory of integration weight integrated services there we go so we have integrated services and the theory is that ok so this computer needs to talk to this computer and I'm going to reserve so much bandwidth let's say I need 5 Meg of bandwidth well in order to reserve that traffic I go to this router I send that request upstream to the router and it either says yeah or nay yes you've got 5 mega syv got that much bandwidth or not and then it proceeds to send it to the next router and of course it gives a yay or nay and it sends to the next router and if indeed I get approval through the whole process I can get 5 Meg of guaranteed traffic through this network they these routers will lock down that 5 Meg and give it to nobody other than that original computer or speaker the source that's the concept of integrated services then we have a switchboard operators yeah you gotta take you back to that concept right I mean that's what the routers are doing right it's like ok clear clear clear meanwhile we have this other concept and this will probably sound a little bit familiar it's called differentiated services and the reason it might sound familiar is because we're going to talk about something tonight called the differentiated services code point or dscp and anybody who's spent a little bit of I'm studying qos is going to know at least what you know the acronym DF CP is referring to which is just to say that it's referring to Q less differentiated services is similar to this packet switching concept which is simply to say that I'm going to provide bandwidth guarantees that any packet flow can take advantage of so I kind of set these up already throughout the course of the network and I say that let's just say voice over IP can get five Meg of throughput throughout this network so by guaranteeing that VoIP as five Meg now my phones can take advantage of this even though this phone only needs I mean again we're talking kilobits per second but let's just say it's like 0.5 Meg or something like that well in theory I could have 10 of these phones now because if each one is taking up 0.5 Meg I can continue to attach phones and and they'll just consume what they need versus think about that RSVP situation you know what happens if I add a second computer in here can and this computer let's say it's only using 1 Meg of that throughput well it's reserved all 5 so in order for those routers to guarantee that it gets 5 Meg of that throughput it can't share that 5 Meg with anybody so if we back up and we kind of take a look at this which one of these two scenarios is better for the individual you know machine well circuit switching is way better for the individual and RSVP integrated services that is the same situation right it's great for the first computer to get that reservation which is why by the way we have this protocol and we believe it or not you can still configure this protocol in a lot of cases it's not really used just fair warning but we don't use it anymore but we do have a protocol and the whole architecture built around this because every now and again we have this same concept of yeah big poppa yeah you got it circus switch this is the same concept of just needing to guarantee bandwidth but what is better for the overall utilization of the network and that would be both the packet-switching which of course is why we landed with packet-switching and differentiated services is going to be the same way it's better for the overall network it's you know not one computer one speaker is going to win out over everything so these are the two concepts that we're going to need to understand that on some level I would expect a question on this with the CCNA you just never know for sure what what they're gonna ask I would say that if you're gonna take the CCNA after the new date in February it's less likely the last about RSVP but on the current CCNA I would say it's fair game for sure alright it's hard to say that with the exam blueprint being as vague as it is but just this general concept of integrated services versus differentiated services is sort of a QoS 101 conversation so let's dig in what are we digging into actually what are we digging into let me check my notes here we're going to talk about marking that's right okay so section a of the QoS conversation is marking now we just gave a scenario hello there we go we just gave a scenario where we have a series of routers and we have some device that's trying to communicate to the rest of the world and these routers are going to need to apply policy to that traffic and so the question is how does that router when it receives the traffic how does it know how to treat it okay the concept is that we have to mark those packets in such a way that the router can receive the packet tell what it basically tell how it's supposed to treat it all right so we can mark packets in certain ways usually with a tag of some kind so if I've got two different computers here for example and they're both communicating we have two different streams going and let's say I tagged this packet with a five and the higher number is the better in this case so we'll say okay so this is a five that's a pretty good priority let's say it's out of seven and you'll see why we're talking these these examples because realistically zero to seven is a pretty common range for QoS values and let's say the computer up here has a value of two so there's a there's a lot of different concepts we're gonna get to when we drill more into the you know cue us as we go on tonight such as for example let's say this five stream is actually a pretty big stream you know and and now we've flipped it in backwards right now Netflix is the high priority and the video game is the low priority even though it doesn't need much so if we're constantly sending that traffic of five what do we do with the traffic of to do we drop it do we occasionally let some traffic through on to I mean that's what we'll talk about kind of our options around that but that doesn't have anything to do with marking fortunately that the concept is simply we're going to eventually configure policy against something where we configuring it against well we're going to configure that policy against the marked packets so we've got a lot of different ways of marking packets unfortunately we we truly have three different ways of marking packets that we have to go through tonight and this is where QoS gets to be a pretty big conversation all right I like that flip a coin if well just just random right just randomly do it believe it or not you're not far off from the truth we're gonna talk about some random weighted random early detect later and and yeah we definitely get into some coin flipping so to speak so all right so yes okay IP precedence IP precedence is the first method that we have for marking package so this was actually chronologically also the very first way that we had to to mark packets as well this is taking a section of the IP header specifically one byte when IP was developed we had what we called the type of service byte that type of service bytes a byte is eight bits so we have eight bits and the concept was this IP precedence we said okay we're gonna take you know let me um I'm just going to clear the screen on this because I think we're gonna need all of this real estate choose a more appealing color than yellow all right so IP precedence the IP header we took a bite we called it the type of service byte which again is eight bits now what IP precedence did was it really took the first three bits so let's just say let's just do this so one two three four five six seven eight okay so we've got eight different bits within this byte and IP precedence took the first three and said we're going to categorize our traffic this way all right we're three bits and those can be all zeros through all ones so with three bits two to the third power gives us eight values this is why we've been using zero to seven as the example because if these are all zero zero zero zero that's an IP precedence value of zero if they're all ones binary 1 1 1 well that would be 7 so 0 2 7 is our max min and max range so IP precedence gives us this range of 0 through 7 and it allows us to mark our packets now how who's marking the packets usually it's a mix of things so like an IP phone for example it's gonna mark its own packets if I wanted to I could download firmware to my computer and I could you know usually built into windows there's there's not a whole lot of options there for example but if I download custom firmware then I can do whatever I want and so I could as on my machine I could set my own IP president's values usually not a good idea because then if I'm marking everything with like let's say a 7 well then that's telling you that my Netflix traffic is more important than those routers trying to communicate with one another that's bad news so we don't we'll talk about trusting here in a little bit as well whether we want to trust those values as they come in so yes the end device can definitely set its own value most often the best place to to tag a packet to mark a packet is so let's say let's say we have a machine and maybe even a phone connected to the switch the best place to configure that marker that marking is right here on the switch port so we typically want to get on to the edge switch as close to the client as we can and mark the traffic there so this does take intention so what kinds of traffic do we want to tag I mentioned foam voice a few times video is another really good one by the way there's I mean when you get into it like in the data center again I spend a lot of time in the data center I scuzziest storage traffic we absolutely want to prioritize I scuzzy I mentioned V motion as another one I mean there's it just depends on your environment typically there's not a lot that you need to mark outside of voice and video unless you're inside the data center so those are definitely the two main ones you can actually go a step further and say okay well we've got video voice and then video and then best effort best effort we call it which is kind of everybody else and then maybe create a scavenger class that scavenger class would be like you don't even get as good priority as best effort but the kinds of things that you really you don't need that much control and so if you find out that hey we've got this mission critical application it absolutely has to communicate regardless or anything else well then you could throw a you know go to that server right let me it's a server communications you go to that server switch port and you say hey everything on that server let's say we have a server hanging off over here so everything on the server let's mark that if if let's say if voice is a five and video is a four and believe it or not voice actually uses three as well typically so voice what we call voice control traffic uses three well we've got two umpty so let's make that our application whatever that application is is going to get to and so we tagged all of that traffic with a two as it comes in so we want to we want to tag or mark that traffic as close to the edge as possible that is certainly something they could ask on the CCNA they might show a network and say where should the tagging occur and you know if it's sourced from this server over here the answer would be the the nearest switchport ideally is typically the answer for that so IP precedence was good we decided though in some environments a values isn't good enough and so that became a problem i need to erase this because i'm going to use that space will say IP presidents come on there we go zero through seven as all we could get with IP president so we got eight values by the way even further to reduce what we have six and seven are reserved in most cases seven every time a router sends a packet like an OSPF packet EIGRP some kind of control plane traffic it's usually marking it with a six or a seven and so I believe I believe routing I did not look that up I can't remember now if it's six or seven but either way yeah let me look that up when you're studying for your CCNA but we can't use six and seven in a network in most cases so really we're drawing about six values and six might not be enough because just me rattling some random things off here a lot of environments are gonna have those things and there's four of our really four of our five right because that best-effort class is going to be zero so we're already using zero we're already using three four and five we need more values than this so this is where the concept of dscp comes in so DHCP was an expansion of IP precedence DHCP I already mentioned it but it stands for differentiated services code point and it's it it simply built to give us more bits so we could mark things more granularly so now that we have six bits while two to the sixth is 64 so now we actually want the SCP we can go from 0 to 63 that would be 64 total numbers right pick up a pop house if I am Shankar I flew from earlier the closer to the end device you can start the marking the better off it is yes thank you for the clarification you want to mark traffic as close to the source as possible generally speaking all right so dscp so that's great now we have 64 different classes now we almost have the opposite problem now I've got so many numbers I don't even know what to do with them all and so the industry came up with some helpful you call it helpful ways of providing some standardized values for us to follow and even though it's helpful it's also kind of obnoxious because now we studying for the CCI CCNA need to really like study all of that and try to keep up with it all and it can be a big pain but fortunately it's not it's not too bad so what we but we do need to walk through this so first of all what they do is they take specific numeric values and they assign letters to them call it assured forwarding is the word that I'm is the word that they used so let's say dscp values let's just make one up here let's say we have a dscp value of 10 all right so as dscp dscp value of 10 would be 1 actually let me change colors here to make it more clear and you know what I'm gonna make this a new layer so we can maybe very easily wipe and really erase these one zero that's not very clear is it hold up and try different color there we go that's clear one zero one zero okay that's a 10 so this concept is taking the IP precedence value here as one number and the dscp values as another number so if I look at it from a binary perspective this right here 0 0 1 is simply binary for one decimal value right and 1 is 1 that's good all right so we're calling this assured forwarding af1 and then what's the dscp value well that's uh that's a 2 that's just a one bit here two bits here four fours there so we have one in the two spot so we'll put a two down okay um actually you know what hold up I mouth was moving faster than my brain on that one we actually are only looking at these two bits I apologize for that so we're talking about assured forwarding for whatever reason and believe me I've looked all over to try to figure out like at what point it became this way for whatever reason this bit right here is always zero I think it might even been I don't think it was in the original RFC so if anybody has some really cool insight into why that was that's always a zero then let us know but either way it's always a zero so really when we're talking about dscp we can only change five bits now our values do still go from 0 to 63 because technically 62 right I mean if that with that being a zero we effectively have all of the even numbers available to us and none of the odds because we can't have it all be 0 and then a 1 at the end right we can have it all be zeros and then as soon as we put a 1 here into this spot well that now it's 2 and so that's as that's as granular as we can get so we technically only have 32 values 0 to 62 with all the evens I know it's weird and I wish it was more straightforward than that but the reality is that this is what we have to know and so the Cir shirt forwarding actually comes into play because when we were talking just a moment ago about okay I'm gonna call this AF 1 because IP precedence is set to 1 but I'm not looking at all three bits I really am only looking at those two bits and those two bits if I just take them straight as they are well that's a that's a 1 0 1 is 1 so this is what we call assured forwarding 1 1 or assured forwarding 11 however you want to say it so AF 11 is supposed to mean something but doesn't really mean much yet let's talk let's let's do a few other examples let's say that we have the same IP president's value but now this bit becomes a 1 there we go 1 all right so what's that well again we take the same process assured forwarding the IP precedence is 1 and the other 2 bits 1 1 in binary is 3 so that's the shirt forwarding 1 3 so now we've talked about two separate dscp values a f11 you remember that what the decimal value was for that we said it was 0 0 1 0 1 0 that's binary for 10 so I'd be a DHCP value of 10 AF 13 we added what bit did we change right we changed this one right there so that's the force so we technically just added 4 to that value so AF 13 is 14 at the end of the day this is only a dscp decimal value I mean if I just told you hey this this application is using a dscp value of 10 well you could go into your Cisco device and configure certain policies against a dscp value of 10 keep in mind this is all this is all part of the markings conversation we're simply talking about different ways of marking our packets I mean ultimately the router is just gonna look at all the ones and zeros this is all just from a human perspective so well looks at the ones and zeros ultimately it sees a decimal value of 10 versus a decimal value of 14 and if I can figure that versus configuring the cisco router and I can I can get on to the cisco router and say a f11 it will know what that means it's just enough it's just like a human word for 10 so what's this mean why are why are we even talking about all this AF stuff well the reason for that is because again they tried to put the industry tried to put some organization around dscp values so as they laid this out it's ok well we're gonna have a f11 AF actually you know what I'll just write it out this way a f12 and AF 13 and then AF because the IP precedents now you can go up we're gonna have a 21 a AF 22 in AF 23 now could they have done an AF 2 0 or an AF 1 0 yes they could and they did but they call it something else which we'll talk about in a moment I know it just keeps going on and on so as we do this there are actually three different levels of this AF wait did I say 3 or 4 I can't do two things at once there's actually four levels of this and I'm writing them all out here and again as obnoxious as this is this is something that we absolutely must know for the CCNA so they're the twelve different AF values that just wrote out these all correspond to decimal numbers again AF 11 corresponds to 10 like we talked about 13 responds to 14 we could probably guess that 12 responds there corresponds to 12 yeah actually it does and so as we move down the path we can actually calculate out which one of you know which which values are assigned to which numbers now there is a way to do this I'm gonna go ahead and tell you right now it's this call it an algebra equation right it's 8 X plus 2y okay so if you have AF if you're given a f X Y which again fair game on a Cisco exam right if they say here's a F 32 what decimal value what dscp value corresponds to a F 32 well memorize this algebra equation because it makes it very straightforward 1/2 32 I plug the 3 in here so it's 8 times 3 plus 32 two times so those would be 2 times 2 so that's 24 oh man ok I just remembered I have whoops all right this is what I try to get every way I get for doing this live I'm trying to I tried to move my head shot so sorry about this I am writing underneath let me just um I think I can go fullscreen here for a second I think I can I think I can I think I can I never made that okay so I never made a scene for that alright let me just move myself over here I know it's looking weird there we go all right now I am over here alright I apologize for that okay so hopefully it's updated let me know if you can see now yeah Shira I apologize for that and it looks like you guys have been warning me for a while now so sorry okay this is good now okay so I'll I'll quickly rehash what I was saying so this let me change colors here there we go okay so this algebra equation right here is what you're gonna want to memorize eight X plus two y because Cisco on the exam might give you this they might say a f32 is one of the example I was trying to give right so if they say a f32 what decimal value does this correspond to you can very quickly figure it out if you have to see algebra equation memorized otherwise you're going to have to kind of reassemble this you know from a bit perspective and be like oh man a f/32 I know that that's a 0 1 1 and then a 2 is you don't have to memorize and take it all back to binary but if you have that equation memorize you can just skip that now the reason this binary equation works is because this is the eights bit right here so we're taking our IP precedence value what do you say 3 right this 3 and we're multiplying it by the 8 bit and then the same thing here this is the twos bit are the twos column or however you want to describe it so again think binary math right we have 1 2 4 8 16 and 32 we're just making the 8 and the 2 the significant part of this equation so that's why that's where the algebra comes from I don't want you to simply assume that is like I don't want you to memorize an algebra equation that know where it comes from there's method to the madness here and so what I started to write there at the bottom is that ok so let's say we're given again a F 32 I would simply take three times eight which is 24 and then I would do 2 times 2 which is 4 so it should be 28 28 should be the answer all right so I highly recommend again just committing that to memory and understanding how to bring the AF values back to a decimal value I actually have a let me see here I've got a valley a separate PDF document one to share so this is a Cisco document that honestly just Google I googled searched for it Cisco and I think AF values or what have you because I just wanted to see where Cisco laid them out because they usually do somewhere right and so as we scroll down here we start to see actually let's double check our number what do we say 32 is 28 all right we got it right so here's the binary over here and you see that they've got the three first three bits in bold the second three bits would be the DFC p-values here's the decimal value here's the AF and and then from there just you know the these are you know the believe it or not I did skip this part a lot of IP precedent all the I per precedents values have a value also associated with them for example IP precedence of 0 has the you know it's called routine traffic and so it just it kind of works its way up the list we don't necessarily need to worry about that but you know just add it to your study list I suppose alright so I wanted to share that with you just to show that this is not something I'm making up this is something that is very industry standard and we need to understand this like I said routers can actually you can tell a router to pay attention to AF 11 traffic as a marking and it will look for dscp values of 10 because it's smart enough to do that now we did mention I did mention earlier you know let me just go ahead and flip back mentioned earlier that okay so why is there not an AF 10 why is there not in a f 20 a 2-0 well those are actually special values as well and let me let me clear let me clear the screen here get back to this alright so when we're talking about I would I not give myself a new there we go okay I'll figure out maybe there's a better art program out there alright so let's say that we've got the value of 0 1 1 and then zero all zeros okay so this in theory if I if I could just kind of construct the AF value myself I could say okay so this is an AF IP precedence is a 3d FCP is a zero so this would be AF 30 which sounds good and that would technically be correct except they give it a different name we call this and I believe it stands for common services I'd have to look up look up what CS means but we call this cs3 and I mean I don't know why they did it that way but they did it that way and ultimately what this is saying is guess what this is the exact same thing as saying IP precedence 3 because my IP precedence bits are set to 3 my DHCP values are irrelevant they're all 0 and so understand that when we're dealing with AI dscp CS 3 is equal to simply using those first three bits it would be lovely if IP precedence being legacy was so far gone that we wouldn't need to worry about IP precedence anymore but unfortunately just simply isn't the case especially comes at Cisco exams they really want us to understand the difference between those three dscp bits and those three IP precedence bits okay so up next we're going to oh you know I almost skipped over a very important part all right there is one actually two things all right two teams two things two more things with this one of these things I'm gonna bring back the old here all right so one thing we need to understand about the shirt forwarding concept as we go down darn it I didn't pick my layer there we go as we go down this list so as we go from a shirt forwarding 1 to 2 to 3 to 4 the priority goes up and you'd think if you're like me cuz I used to think this we think that as we go this way it also goes up so AF 11 is lower prioritized than a f12 is lower prioritized than AF 13 but unfortunately when they designed the spec and the whole reason why you Kenny they kind of got away from this but was originally designed it was that this bit right here and I was getting busy but that fifth bit was supposed to refer to like how how likely it was to drop and so that when when they translated that over to dscp these assured values essentially as they go up it's the more likely that they will drop that they will get dropped so in a weird way you actually are going down whoops there you go so we're actually lowering if I can draw this backwards we're going more priority to the left and more priority down it's really bizarre and it takes a little while to wrap your brain around that understood but just just take it for what it's worth now do routers class selector oh thanks for looking that up Oh GI T Keith hey thanks for coming Keith good to see you thank you for chiming in with that all right so the yeah so so as the number goes up that that's more likely for them to get dropped so let's just keep that in mind oh I remember now I got distracted by class selector what I was saying do routers automatically make it so that they're going to more be more likely to drop a of 13 traffic over AF 11 the answer is no all of this is sort of how the industry views things and they want to standardize things so you know if you as an IT person you start off with one company and then you go to another company and you you know you're bringing your knowledge with you they kind of want the industry to be standardized around these same concepts but at the end of the day when you've got you know three routers or what have you in a row every router is going to be treating this traffic as you locally configure that router it's one of the other pain points of QoS is simply that you've got to have your entire network be cohesively configured to be deploying this you know to be applying the same policy to your traffic so when it comes to the AF concept there's nothing that I could configure my router to make AF 13 the highest priority traffic and to drop everything AF 41 as soon as something AF 13 comes in I could do that I have that control because I can configure that but obviously that goes against the way the industry kind of wants us to and so just be aware of that the other concept that I almost skipped out on I want to go ahead and pull the yeah I will just do one more layer here is this concept of expedited forwarding okay so we have one very special value it's called EF expedited forwarding if it had an AF equivalent it would be AF 53 because what we're going to do is we're going to put a five in here one zero one and then we're gonna put a three in there one one and if my math is right this is going to equal 46 it's a decimal value of 46 what this is is the it's supposed to be the highest priority traffic that we can figure within a network I already mentioned that you know routers speak at that you know CS value of 6 or 7 or what-have-you I mean they're speaking at a higher level there that we have to have those reservations but within a network I want to make 46 the absolute highest priority of everything which is interesting because well now I guess I'll just leave that alone all right so so when we configure our routers we have the option to apply a a marking to packet types and this 46 is usually synonymous with again that voice over IP because there aren't a lot of networks out there that have anything on the network that's more mission critical than voice over IP traffic Voith is very very important it's a very high priority lives could be at stake literally right like if you need to pick up a phone and call 9-1-1 those phones have to work they have to work well and and also even though you know yeah on the one hand they're very very you know they're tied to the human health it's also the first type of traffic to if you you experience even a little blip then it's going to be painful I mean I was on a phone call earlier today and the person on the other end of the line was they were I mean I know I'm at home so I've got great Wi-Fi sir at risk my phone's on the Wi-Fi wherever they were was miserable and the funny thing is is when I think about it I was probably getting about 80 to 90% of what they were saying and 80 per to eighty ninety percent seems like it's pretty good but I tell you what again that 10 percent or 20 point whatever I was losing was was enough that I could not understand what they were saying and eventually the call was dropped and so that's why I say it's like you know even just a little hiccup can make it so we can't communicate now we mentioned video as well is video or voice more important because you think well video you know I mean that that that's very important to if you're having a video call you have to be able to see each other but I tell you what if the video goes out a little bit and the voice stays does it still work yes I can can even though it's the quality is low and I'm annoyed that the quality is low we are still communicating what if it's the opposite what if the voice is a little hiccup II and the video is immaculate it's perfect well I can't communicate with them so video actually is a lower priority than voice for that reason alright moving on I think now to remember we talked about three different types of markings so if comfortable - we've covered IP precedence and we've covered dscp I'm looking if there are any questions I don't see any questions jumping in here so if you do have anything about IP presidents dscp any of the assured forwarding expedited forwarding stuff be sure to chime in whether the chat will try to address it live now we get to have the conversation for the last value or the last type form of marking that we have and that is called class of service class of service takes us away from IP so everything we were just talking about was in the IP header if I have traffic that doesn't have an IP header for whatever reason it's pretty rare in this day and age that we have any non IP traffic but every now and again we can have that especially in the data center and such then we can't use those values right we can't we can't do IP precedence we can't do DSD because we don't have an IP header well that is a rare situation the more common situation is that we cannot use class of service and here's why class of service happens in the ethernet header and unfortunately when Ethernet was devised it was not given this concept of a marking we had no way of marking it and that created two problems one sure class of service was was certainly problem but another problem that had to be solved was hey we've got all of these networks we want to kind of virtually segment them virtually segment the lands right virtual LAN ringing a bells so we have the ethernet header we had no way of marking for class of service and we had no way of marking for VLANs and so as we know when we're talking about because you know we're staying for icnd2 and icnd1 it was definitely was covered we have two switches and let's say I have a host in VLAN 10 and a host in VLAN 20 and these two hosts need to go across that same link so when this switch receives those packets how does it know which VLAN the traffic I'm receiving is well kind of in the same way we're talking about QoS we have to mark that traffic and the way we market is with a VLAN tag but again Ethernet header doesn't have spot for VLAN tags so what we did what they did I wasn't around but what the industry did was we carved out a section or not it carved out I'm sorry that's the wrong way of saying it we actually injected a larger not a larger I'm tried we injected a section of additional header that gets added into the header on certain links and so what we did was we added the ability to do the VLAN tag as part of that that's cool and we also got three bits for what we call class of service so the VLANs they got I got the bulk of it right there's like 12 bits and the VLAN the you know tag we can have up to 4096 VLANs it definitely got the lion's share of the bits but the class of service we got three bits and guess why because IP precedence has three bits as well so we can start to align our layer to quality of service which is what Big Poppa House said right class of service is layer two we can align our layer to QoS with our layer three now it does take a little while to read it's like MTU if you ever had on MTU conversation it's like oh there's layer two MTU and there's layer three MTU and it takes a while to understand the differences and when one comes into plain when the other one comes into play same thing with this unfortunately it'll take a little while to kind of see into this concept understand this but one of the biggest things we need to understand is that the three bits for class of service only show up when that VLAN tag shows up so here's the pop quiz for you where and the network are my VLAN tags what types of link I missed I'm gonna take a sip of coffee while you think about that good I haven't even had a sip of my coffee yet and as expected it's uh it's not the hottest anymore all right I'm seeing some some options coming your trunks trunks yes absolutely trunks is where those VLAN tags exist so if I were to set up a network that let's say to two switches and we're not trunking we're not trunking here okay so this is what would call an access port and I bring my traffic in and I do what I'm supposed to do and I tag this I tag this traffic with a cos of three or whatever value I want and that packet ends up getting sent across this link that cos tag of three was just lost okay so my end and QoS policy is going to be fragmented because as soon as I hit a link that doesn't have the ability to tag traffic I'm sorry the ability to tag VLANs I also lose my ability to tag TOS wait for the most part for the most part we rely on layer 3 and home said MTU where we've rely on layer 3 QoS because it's end-to-end IP you know the IP header is for the most part going to stay the same even though yes it gets torn down and reconstructed actually no it shouldn't get torn down and reconstructed as it goes from hop top but that's the point where it should persist all the way and so that tag should be kept all the way down scratch my tags I'll scratch hers and and so one fun thing by the way bot Cisco about Cisco tagging and such on switches it's very intelligent around some of this so it can receive a COS value of 3 on the Ethernet frame and it'll recognize that and they'll actually translate that over and put it into the IP header for us and vice versa and so it just depends on the model again this is that's one of those diving into the weeds conversations not expected for us to understand that on the CCNA but understand class of service layer 2 3 bits which means again 0 through 7 right let's run there go 0 through 7 that's that's what we can do with class of service layer 2 QoS all right so yes generally speaking run a line with the layer 3 but generally doesn't matter so the question though is when we do when we define our policy what are we applying that policy to because I can match against IP precedence I can match against the SCP and I can match against cos and so it's whatever I'm telling my device to match against that that that's what's gonna pay attention to and so some a lot of layer 2 switches by the way can't match on IP precedence because it never looks at the IP header so keep that in mind when it so if if we need it to be able to tag on in order to apply policy to a COS value well then we need Theo's values exist and that were trunking everywhere we can okay so um that is cos all right so we are done with nearly done with the first hour we've got 55 minutes in and we are done with section 8 but I new section a and Section F are definitely the long ones and we are going to burn through the rest a little bit quicker so let's go ahead and dive into Section B which is actually device trust so this is going to be the fastest section because it's not too complicated but we do need to understand it for the CCNA so this concept is this we have the example I gave earlier I'm sitting here on my machine I'm getting frustrated with my Netflix or whatever it's like it's downloading too slow and buffering and it's granulated pixelated and so I'm a you know I do some google searching and I download my code I like hey you know if I set my own dscp value to EF then the network is going to prioritize my traffic all right so that would be a bad actor on our network there they're going to do something malicious so even though it's not like they're not trying to take your company down but they are trying to mess with you a little bit because I've got other phones on the network that are talking at EF and so if this switch gets traffic at EF from two different sources what's it going to do well in most cases it's I mean it's gonna put it into the same queue and my Netflix traffic is now going to get you know highly prioritized along with the rest of the voice traffic so that's that's no good we don't want to do that so how do we how do we fix this problem well this is concept of a trust boundary and the trust boundary says this let's say and I really need a larger network here do-do-do-do-do there we go alright so my trust boundary would say I'm going to trust this port because there's a phone there and I'm not going to whoops Lendl is just keep talking about this I'm going to trust these ports because they're connected to switches and I know that I've got those two just Confed configured and so my trust boundary at this point kind of looks like this whereas I do not want to trust my users I don't want to trust anything down below my line right so maybe there's some other users hanging off of here that's that's about the extent of the trust conversation if I don't trust you guess what I'm gonna do I'm gonna wipe your tag as soon as I get it so this EF gets wiped down to a zero because that's what I've configured that switch port for anything coming in from the trusted interface it were good now one quick further conversation might be what wait a second Jeff I I know I remember now I'm on the bottom left is my drawing okay could my drawings good all right well wait a second Jeff I know I've got a PC hanging off of my phone at work so is that a loophole well believe it or not no because the phones by default are have an untrusted interface on like their downstream connection so by default it's gonna wipe everything that comes into it and this okay try to keep color consistency here here this switch port right here can actually be configured with a command that trusts and then there's this additional parameter called Cisco phone so if we have a phone on that interface that will trust it but if again I'm a bad actor right if I'm like you know what forget this I'm gonna unplug my phone and I plug myself right into that port oh man well now I'm on a trusted port that's that's not gonna work because the switch is gonna recognize hey there's no Cisco phone on this port anymore therefore I don't trust it okay so Cisco fortunately has put some thought into this and made it so that we can tackle most of these scenarios all right so that's about it from a trust perspective that's so it's my fifty five minutes on section a we spent four minutes on section B but that sometimes the way it goes any questions on that let me know but yeah configuring your trust boundary is something that you'll want to consider whenever you're deploying QoS so on to section C and Section C is prioritization so we already talked about this will be really quick too literally all they want in fact let me just pull up the the topics again literally all they say here prioritization is voice video and data and we already kind of talked through this but let's go ahead and talk through it one more time here just to just to be sure so when we're talking about voice voice traffic is absolutely mission-critical so we're going to assign that with EF traffic that's again DF CP value of 46 you're going to want to memorize that that will definitely come up at some point on a Cisco exam video should oh you know and one other thing too from a IP precedence perspective and a COS value perspective these would both be set to five all right IP precedence is five cos is five all right so that is that is voice video is usually going to be associated with cs4 which you know again we can think about that same concept right of with that eight X plus two Y where in this case Y is just zero so cs4 would be four times eight so that would be a value of 32 and that you know we already said common selector for CS for maps to IP precedents for an IP presidents and class of servers tend to map to each other so this is typically what we're going to see what was the what was the word that they used for it was data all right so they're calling it data I tend to call it best effort a lot of people do so best effort this is where we're going to see CS 0 which you know do the basic math right that that decimal value would be 0 which would associate with IP precedence 0 and class of service 0 this is this is what they put on the on the blueprint so I would have these memorized either jot it down or grab a screenshot or what have you this is this is the I mean the the the priority value goes up this way for sure now for bonus points I did mention that voice control VoIP control traffic is in here oops voice control so this is like if you've ever had a if you've been involved with a Cisco phone systems call manager or Communications Manager that that traffic it has a value of 3 so like when it's sending the dial tone for example when it's hook connecting phones to each other telling them how to get to you know the other phone they're ringing up this would be a value of 3 so cs3 IP president's 3 cos 3 all the way across the board Deb you know decimal number of 24 at that point so yes thanks Papa house absolutely cos of 3 so that's a bonus points one another bonus points one could potentially be in here some places and this this is kind of common way I mentioned it earlier it's called scavenger traffic scavenger traffic even though we typically assigned this with a CS value of 1 which you know again fill this hour right eight IP precedence one cos of one scavenger tracking traffic even though it's rated higher from a number of perspective then best-effort the the purpose of it is to give us something below best effort because best effort is zero and literally as low as we can go if I've got traffic on the network that I want to be D prioritized like I want that traffic let's safe like research or test dev or something that is like could impact the network and I would rather have Susie and Karl be able to surf the Internet because they're probably doing their jobs and surfing the internet at least that's my hope I'd rather than surf the Internet then that traffic be prioritized over them and so I could actually apply that cs1 tag to them to that traffic whatever that traffic is and then go on to the routers and again because routers 1 and 0 and 5 and 4 they mean nothing to the router until we tell it what to do with those values and so if I get on the router and I tell it hey one is worth less than zero the routers good with that so there we go those would be the bonus points ones the other ones would probably be the routing control traffic and again I didn't bring that information with me so be sure to go Google that and make sure that hey you understand what network control traffic looks like a quick search will tell you what six and seven do if you've got the official study guide it'll they'll fill you in on that as well I'm sure alright so there we go a few more minutes and we're done with prioritization shaping and policing these I'm gonna I'm not gonna spend a ton of time on this just because I do want to get to there we go because I do want to get to congestion management because it's uh it's a lot all right so we're gonna cover these both at the same time so we have shaping and we have policing these are two sides of the same coin okay similar concepts both attacking the same scenario the scenario is this I'm a router and I have a bunch of traffic coming in let's say I have a hundred Meg coming in and I have a weigh-in circuit over here that's ten Meg okay what am i to do in this scenario and if we want to be a little more realistic I mean I could probably pull this out I mean 100 Meg to ten I mean that's just gonna be bad news right what if I have what if I have fifteen Meg coming in what if I have 12 Meg I'll just say 12 Meg all right and now I've got ten out all right do I have to drop that 12 Meg of traffic or add extra 2 Meg of traffic basic math says yeah I mean it's got nowhere to go all right I mean you think about water if it's if it's a thick pipe and a thin pipe then that water is just gonna back out and it's got nowhere to go but the routers do have some storage space in there I mean it's it's called buffers right when when the packet lands with the router it can hang onto things inside the buffer for a brief time and so the concept of shaping says I'm going to hold some of those packets back in the queue without dropping them and I'm going to release them in hopes that I can get it all out without without dropping any traffic alright so shaping it truly is sort of imagine this triangle concept where it's I'm shaping it down you only shape it down you don't you don't shape it out shape it up so that's that's the concept of shaping the big reason why we want this concept is because of policing because inevitably what's on the other side of this 10 Meg link and usually a router and that router is owned by who my service provider Internet service provider or LAN service provider whomever it is so I am sending this traffic to the ISP now what's the ISP going to do if I send more traffic than it let's say this is actually a 100 Meg Ethernet pipe circuit this is very common right it's a 100 Meg Ethernet connection with a 10 Meg circuit so they're telling me that I can only send 10 Meg at a time but hey they gave me an Ethernet handoff it's it's a gig it's a hundred bag whatever it is so it's like big papa house so it's like a funnel creating the most efficient stream of data yes now we might self to drop traffic and we're going to come back to that in a moment because in the event that I send too much information to the ISP what's the ISP going to do it's gonna cut it off now somebody's piece will have burst rates and things like that that gets a little complicated they might say hey you can burst up to 15 Meg is long you know that's just saying that that 12 Meg would get through as long as it's not sustained because if the 12 makes days it's no longer a burst it's sustained so there this concept is policing the concept of policing is is more of think of a police officer and I think of a traffic stop you know officer or whatever you know with the whistle in their mouth and putting the hand out like you know fleet so so we've got a big stop sign here that says okay ten Meg I'm sorry 12 Meg is coming in too good all right my face isn't in the way victory all right 12 meg is coming in two meg is going to get dropped and ten Meg is going to get through okay that's policing that's this concept of policing policing is usually an inbound concept shaping is usually an outbound concept and like I said there are two sides of the same coin because of this reason so let's go back to the shaping conversation for a minute let's say that okay I've done what I can but I have to drop traffic I have to drop two Meg of traffic so does it matter I'm gonna I'm gonna drink my coffee here in a moment so I'll give you a few seconds to think about it doesn't matter or why would it matter which router drops the traffic doesn't matter does router if router a drops the traffic or router B drops the traffic which one would it be better to drop the traffic which one would we not want to drop of that traffic here's my coffee break think about it that's getting colder and colder I should probably just open up my soda water okay so I see one a come in I assume that's that we want a to drop the traffic and that is the correct answer panda so the reason why we want the router a to drop the traffic is because I'm the one configuring it and if I'm going configuring it I get to choose what traffic gets dropped so if I've got voice and Internet traffic going across the same circuit I get to choose hey I'm just gonna drop two Meg of that Internet traffic this might be a live and Meg of Internet traffic and one Meg of voice okay let's just say that that's the scenario here so 1 Meg of Internet traffic 1 I'm sorry 11 Megan Internet traffic 1 mega voice now if I'm in control we're gonna drop 2 Meg of Internet traffic and we're not going to drop a bit of that voice stream the ISP is the ISP care no they just see 12 Meg of traffic they're going to drop 10 Meg of traffic I'm sorry - Meg of traffic so if all things are created equal they're going to drop roughly boy now my head's breaking with the math they're gonna drop roughly like 90% of the traffic they drop will be Internet traffic but 10% will be voiced something like that so this is why Roger let's see a Papa House says router a because it would be less resources upstream yeah it's less about the resources and more about getting to choose the traffic type because you know that ISP is gonna potentially drop my void stream so this is again this that is the biggest reason for shaping yeah we want to not drop traffic if we can but the bigger issue is I get to choose which traffic gets dropped okay so shaping and policing policing just think of that police officer just think of that traffic guy you know guard with the with the whistle and the handout or the or the stop sign or what have you that's policing inbound you you you are out of space and then the shaping again think of that as that triangular shape right shaping that traffic down one when you if you go on to CCNP and you use especially if you go on to CCIE you're expected to understand how shaping works and that is very complicated very very complicated but you know for a CCNA we just need to understand what the concept is which is we'll start there the two Meg alright so Shara has a question the two Meg is it dropped completely or hold in buffer for later so if it has the flexibility if it has the luxury of well when we talk about congestion management this will make more sense but we have cues and packets will line up in the queue ready to go and so all we're doing in this shaping conversation is it's this guy's turn to go let's say we have three cues here it's his turn to go but we don't actually want to send that because we've exceeded our amount so we're not going to send a packet and then we come down here and can we send it now still no can we send it now still no can we send it now oh yes all right so we send that packet that's the concept is we're simply not clearing it out so when do we when do we drop traffic only if a packet comes in and we don't have room for it in the queue then that packet or another packet would get dropped in favor of that new packet coming in so as long as the queues don't fill up as long as the buffers don't fill up we're we're gonna be fine but as soon as the buffer to fill up that's when we started losing traffic out the backside all right so I believe I believe that takes us to congestion management and we have 15 minutes oh okay actually that's more than I thought we'd have all right so congestion management this is a big topic and actually that question about dropping was was a good lead-in because we need to start talking about queues change the color here just to have some fun what's a good color purple we haven't done purple yet tonight I don't think okay so how does a router man again CCNA the the blueprint truly just says congestion management and leaves it at that this is like the heart of how QoS works I don't expect that Cisco is going to want us to be an expert at any one of these topics but we should at least have a basic understanding of what we're talking about here at the end of the day what congestion management is saying is a kind of what we just had with that scenario we're getting a bunch of packets coming in and again we've got this concept of queues so we've got these we've got these queues so called output queues they happen usually on the output the external facing interface whatever that is and so let me draw them slightly differently so as I have these queues lining up here I start filling in packets and start filling it up with packets so let's say I have three queues how am i I'm getting a little bit ahead of myself actually let me let me not let me not do multiple queues yet let's just do one queue all right let's say that I have a router here's my scenario I've got a router I've got a land connection here which again is usually on the higher side and I've got some kind of way and circuit here which we'll just say is in this case 50 Meg something like that okay that's gonna go over to the other router all right so I'm getting traffic in and I'm going to start sending it out that way in circuit now if this is literally a physical 50 Meg circuit I'm not really great way of doing that I suppose so it's a bad example but you know if the router is uncapable sending more than 50 Meg then then what this is this is the kind of situation we want to talk about so and the reason I say that by the way and this my brain goes a few different directions on that is that in order to get 50 Meg usually you're gonna get again that gig Ethernet handoff and then you're you're gonna be responsible for making sure you don't send more than 50 across that gig but if the router for example has like a a t1 or t3 coming into it or an e1 e3 whatever you know that circuit is coming in if it's hard bound to a very specific bitrate then we need to use these cute we're forced to do these queuing mechanisms whereas the other router I mean with it's a gig pipe is just gonna send it and that's where we need to do the shaping all right so that was needed here and are there let's just talk about this so we have a queue let's just say we have one queue to start with queue fancy word for a line right and so where our packets are gonna start lining up and by the way we're going to have you know this is the elephant and mouse conversation that you sometimes see out of you know study guides and things like that we have these small packets let's say these are the void packets again we'll call these the mouse packets and then you've got maybe somebody's downloading something from the internet and this is a big packet so this is what we call the elephant packet so so we've got large packets and small packets in this queue and okay that's good well and good and we're going to send this packet there's something called a scheduler here and this scheduler from within the router is going to pick and choose I didn't spell that right there we go there's going to pick and choose when to send those packets it's gonna send them as fast as it can but it's limited by what's going on here in our case it's a 50-megaton only send packets so fast right it's gonna send at 50 Meg so the question of course is always going to be what happens when this queue gets full so I already kind of exceeded the queue let's say the queue truly expands out to here all right so we're talking about so many kilobytes or megabytes of space usually megabytes etc so we've got so many packets will be queued up here and then a packet comes in okay so now this packet comes in and we don't have room for it by default the common the the default behavior we want to call it is called tail drop tail drop simply means oops sorry you didn't make it in you know there's no room in the queue right it's kind of Christmas season right there's no room left in the end there there is there's no room for you so where you gonna go the router is gonna drop that packet that's tail drop that's kind of the default behavior there is a slightly better solution that we call well technically first of all it's called random early detect read all right this as the concept isn't really used anymore we use a concept called weighted random early detect which is with a W let's just talk about - the W for a moment here random early detect says that okay whoops I've got actually okay I'm just gonna have to erase a few of these do-do-do-do-do all right so random early detect says let's just look at those words it says early detect okay so early detect says wait a second here we're starting to get a little full so as this packet comes in a really detect write the edie part early detection says oh I'm starting to get full I'm going to randomly decide when the next packet comes in whether it gets dropped preemptively or not so the next packet comes in and let's say it's an elephant I don't know and randomly remember we talked about the coin flips earlier randomly we're going to decide whether that packet gets to stay if it is if it stays it stays if it goes it goes so maybe maybe we had to drop that one and then another one comes in and it's good and then another one comes in and that one's bad and it gets dropped and then another one comes in so why do we care about it well random early detect actually does smooth out internet traffic or network traffic so there's reason in and of itself to be to use it but where it really comes into benefit is when we add that W that weighted random early detect because now we're going to give weight to the traffic type so we might have these prioritized with our tags and these might be those voice packets tagged with v and our elephants are tagged with zero and now a random weighted or weighted random early detect or read with a W is going to more what be more likely to drop zero packets right the the elephant is gonna be more likely to drop those than it would be to drop voice packets which again is kind of the goal right we want to keep the high priority traffic flowing so this assumes a single queue fortunately we don't have to worry about single queue very often this is well it's not even technically to default behavior which we're gonna go into here in a moment so Cisco I'm doing the wrong thing here there we go let's go wait there we go change color again so cisco said okay that's not a good way of doing this really you know we don't we don't want the uh we don't want to have just one queue because it doesn't give you any control over the traffic right I mean we saw several elephants for example in front of VoIP traffic well VoIP traffic okay it's good that it made it in the queue it's not gonna get dropped that's awesome unfortunately it's not so great when that router has to sit there and turn out even though we're talking about microseconds and nanosecond and such I mean we understand that it's a small timeframe from from a human perspective you mean from a nonhuman / from a router perspective it takes a long time to grind out a 1500 byte packet versus a maybe a fifty by voice packet however big it is so it would be great if we could send those small packets in front of the large packets and with a single cue accuse a line right I mean if there's five elephants in front of one voice packet by the time it gets out it might be too late and that's no good so what we have here is some Cisco called this weighted fair queuing okay it was kind of the early entry into more sophisticated quality of service the concept is this Cisco is going to whether when their routers are receiving traffic they bring traffic in and traffic gets classified so here's our traffic and its first of all gets classified into something we call flows flows are the idea of a flow would be if I'm sending a file that would be one flow if I'm downstream if I'm streaming Netflix that would be another flow if I'm playing a video game that would be another flow and from a technical perspective we're looking at source and destination IP address source and destination port you know very variety of other facets that would define a single flow because in theory if I'm it's supposed to be me and you know one computer and another machine whatever when talking to each other that's a that would be a flow that would be how we decide which traffic belongs to which flow so first I've always put them in two flows then we split them into two different types of flows we split them into large flows those elephants and small flows now Cisco says that most flows will be deemed small I don't they don't give away their secret sauce so I don't know what the threshold is also you know if my flow is a bunch sending traffic and it's you know here's a tip you know five 1500 byte packets and then here's a seven by packet and then here's another a thousand byte packet you know I mean Cisco is going to somehow weigh that out and average it out and figure out whether it's a large or small flow again we don't really see the secret sauce of how exactly they do this and then these get split put into two different queues so now we have two queues which is better and the small traffic gets prioritized over the large traffic so if there's anything in the small traffic queue it's going to get sent first so those wishes are sending that's a that's a very meaningful sound effect those those packets are getting sent out that interface before the the large queue even gets to send one packet so this is what we call a strict priority queue strict priority queues always go if I have a packet inside this queue it's going to go before anything else does a strict priority queue is not always the best idea because you could actually completely starve the other queue if I don't know how it happened but if we got 10 Meg of video been our voice bandwidth at the same time somehow we got up to 10 Meg on a 15 megatons grind to a halt so you have to be very careful about using strict priority queues but weighted fair queuing does a pretty good job again this is why Cisco says most traffic gets put into the small category because you don't want to risk starving out everything else that's classified as large all right we've got about 4 minutes left so we're gonna go fast on this all right so we mentioned strict priority queue we'll come back to strict priority queue the concept is important oops the concept is important we're gonna come back to it a moment here so now let's talk about the evolution of that now we have class-based weighted fair queuing so we see we've got the wfq in there for weighted fair queuing but now it's class-based so what's up with that well class-based says I'm going to create pretty much any number of queues I darn well please I can create as many Q's as I want because remember I mentioned that routers are software machines and software is it's not bound by the underlying hardware certain switches they are like it's a lot of especially like the some popular switches the older switches like the 35 60s and 37 50s though any those had exactly four output queues so I couldn't configure as many as I wanted Niger for out here just because I ran out of space but I could have eight I get however many queues I want and then I can assign this traffic it can match it based on whatever I want now what am i basing it on well now we've come full circle back to the marking category right so I might have a queue for EF traffic oops let me just write this like over here so now I've got EF traffic here I've got a f11 traffic here I've got AF two one and AF three one alright so those are my four queues as a router receives packets in and it looks at those markings so again this is keep in mind this is the output queue okay the input that you know you can't see that okay the input is when it gets marked right it's our and not when it gets marked it's when it reads the markings so it checks the marking says okay you are an AF 11 so as I put you into the output under the output interface I'm going to put you into the AF 11 Q okay so now we have multiple queues and these queues are starting to build up and class base weighted fair queuing essentially before we get into the wfq part the class base simply says I'm going to do remember this concept of the scheduler I mentioned so the scheduler is going to send these packets in a specific at a specific rate and typically I'm going to divvy up the bandwidth in some way so I might say that I'm going to give five Meg to the EF traffic I'm going to give 10 Meg to this traffic 20 Meg I don't know 30 Meg something like that okay so I'm going to give so much bandwidth in the configuration to these queues and the schedule is going to make sure that we do that that we accomplish it right that EF gets no more than 5 Meg in the event of congestion if if there's no packets and any of the Q's none of this really matters right we're only talking this is why this concept of congestion management is so huge is because I mean that once there's congestion all of this opens up and all of this is at play so we take our all of our same concepts and we start to apply them if a queue gets full we apply weighted random early detect because we want to make sure that we're dropping traffic that comes in based on these based on these values so I've only got so much room in the overall queue so as the overall queue fills up I'm going to want to drop af3 traffic before I drop EF traffic I definitely don't want to drop EF traffic okay so weighted fairway weighted random early detected comes into play we don't want a tail drop because if we're tail dropping now okay now that should come back to come to life a little bit because remember before we only had one queue well with random early detect I actually keep my buffer open if my overall buffer because all of these are part of the same buffer right my overall buffer is filled up to here I will always have space to receive a new packet and so as long as I always have space to receive a new packet I can make a decision as to whether to keep that packet so if a new packet comes in its EF I'm keeping that packet if that packet comes in and it's a f31 well how am I is this the last packet in the queue because there's a high percent chance I'm just gonna drop that to keep my queue open if I'm only halfway full there's a low percent chance I'm going to drop that that's this concept of keep my I'm early detect right I want to keep my queue empty so I can always receive and choose it's kind of like the policing concept I want to make sure that if EF traffic comes in I can keep it I don't ever want a tail drop because when I tail drop it's just like that policing concept again if I tail drop it's outside of my control I want everything to be inside of my control as much as possible okay and last but not least there is a I'll try squeeze it on the bottom here there is a default class that I don't have to configure anything that doesn't match my criteria for the 4qs will get dropped off into this default class and interestingly if I don't use default if I don't create my own default class it's still there and it's going to do when it weighted fair queuing the this the same concept is going to happen within that one queue splitting up into large and small x' but if I create it then I have to manually specify to use weighted fair queuing otherwise it's not going to do it so it won't do this concept of splitting into large and small I have to type the word if you've ever seen in router configuration fair - queue that's the command fair queue alright so I threw that at ya hopefully that makes sense the last concept I believe we're on the last concept let's see here llq yeah all right oops I did the wrong thing there we go all right last thing I know we're a little bit over two minutes but again this is very important and probably this is potentially the most important because this is what Cisco likes us to deploy then we have this concept of low latency queuing llq takes this strict priority concept except it chops off the strict and gives me a priority queue that we call the low latency queue and of all of my queues guess which one I'm gonna make the low latency queue that EF q should always be my low latency queue I think most routers most Cisco devices allow me to create two low latency queues but then there's really no reason to there's nothing that should be as mission-critical as voice they simply give us that option assuming that that's still valid because that's that was on the ISR gentoos I believe maybe even the gen threes either way the low latency queue is a priority traffic in that it will always be serviced first however this bandwidth limitation will always be held and it'll always be held in check by it so if I have a packet if I get four packets in here remember my scheduler concept was that I would be taking a packet from here and a packet from here and a packet from here and I'm going to do that at the rates of those bandwidths so it's gonna keep track of how much how many bandwidth is a it's like speed like you never have a speed at any moment in time it's always an average you know it's kind of a confusing concept I like my speedometer says I'm going 50 but that's that's an average over the last you know it took me X amount of time to get from point A to point B so that tells me I went 50 miles an hour same thing with same thing with bandwidth right so taking an average over time it's looking at that saying okay I've sent it about three meg for Meg whatever and that schedule is gonna be going through and moving packets from the queue and sending those packets well the EF q when when it becomes a low latency queue and and that by the way there's a prime there's a keyword in my configuration the keyword is right here it's priority so when I configure that bandwidth with the priority keyword any time a packet shows up in that queue it's going to scrap its plan whatever its plan was like well next I'm sending in a f31 packet and EF packages came in I'm gonna go right back up there I'm gonna send that and then I'm gonna come back down and send the AF 31 it's a priority queue in that sense but it will always do one quick check say okay five packets just arrived in the EF Q so instead of sending that AF 31 I'm gonna go up I'm gonna clear that out one two three oh hey we've hit five Meg you two have to wait so now I'm gonna come back here send that AF 31 and then by then probably go up and clear that Q out so it's it's very important because the it matters a lot in that the strict priority queue could possibly starve other queues and we do not want that low latency queuing is a great option because it gives us eight and again it's still a priority queue it's considered a priority queue but it is not considered a strict priority queue so it will always get serviced first as long as it hasn't exceeded our bandwidth limitations okay I cannot believe that we actually got through all of that in an hour and 35 minutes so [Laughter] I understand that that was a lot of information and good thank you for especially for those who started at eight o'clock and have digested this for an hour and a half wherever you came in thank you for joining I understand that that was a lot of information this video will be archived as all these videos are archived number one it will go to my youtube channel for a permanent archival but it will be available here on Twitch for the next sixty days so don't hesitate to go back and just kind of go back through some of those concepts CCNA this is an important concept and we absolutely need to understand this if we're expecting to pass the icnd2 so thank you very much for joining hey quick quick advertisement tomorrow at 3:00 p.m. Eastern for those who are available I am right now going through the CBT Nuggets dev net content made by a fellow trainer of mine Knox Hutchinson so that content is phenomenal it's a super exciting I just started this on Monday it's kind of a learn along concept like here I'm teaching concepts that I know and I understand this dev net automation world is something that I'm trying to get more involved in I highly encourage everybody you included to get involved and so if you happen to be free come out and watch that with me we're watching videos offering as much commentary as we can I'm gonna be doing the labs it's just kind of this learn along concept so come out on Mondays and Wednesdays at 3 p.m. Eastern for that meanwhile we'll be back here next week to wrap up icnd2 the last section which I believe is infrastructure maintenance that's right so have a great night everyone and we'll see you on the next stream [Music] [Music] [Music] [Music] [Applause] [Music] [Applause]

Channel: KishSquared

Views: 2,503

Rating: 5 out of 5

Keywords: ccna, qos, quality of service, cisco

Id: axDhwHWuTs0

Length: 103min 43sec (6223 seconds)

Published: Tue Jan 14 2020