Net Talk #3 - IP SLA

[Music] hey welcome back to the channel everybody this is Kevin and in this week's video we're gonna take a look at yet another topic on Cisco's encore exam let's exam 350 - 4 0 1 and this is gonna look a little bit different than previous videos because it's a replay of a webcast I did recently periodically I'll go live on YouTube and we'll have what I call a net talk we had a net talk covering Jiri over IPSec configuration we had a net talk covering diem VPNs well this one is on IP SLA SLA standing for service level agreement and with IP SLA we can configure a router to send a probe out into the network and we can have another router configured as a responder that will send us back statistical information about how the network treated that packet we're also gonna see an example of how we can make a routing decision based on delay that IPS at Laeken measure for us and if you really enjoy this video do me a favor click the like button below it helps the channel and also please click on subscribe so you don't miss any of our weekly content now let's dive into this net talk on IP SLA let's begin with a look at IP SLA theory and again SLA stands for service level agreement and a couple of things are a couple of primary things that's going to do for us it can measure network performance one of the reasons that that I started using this back around the year 2000 was for quality of service I was really into our still kind of am into voice over IP and they color they call that collaboration these days but I would set up a quality of service configuration in my voice network and I had two ways to get from one office to another office one way was over the LAN I mean that I'm not paying any toll charges to the phone company if I get to go over the LAN but the LAN might get saturated it might not give us really good conditions we might be dropping packets there might be excessive delay so we can use IP SLA to send these probes out into the network to kind of test the dip our toe in the waters and see if it's a if it's a safe environment how beat up does our probe get and when the receiving device got it it would give us statistics like here's how much jitter you experienced here's how many drop packets you experienced so that's one application of IP SLA another one and by the way we're going to demonstrate both of these live today another one is to influence routing decisions we have two paths to get from one router to another router and the way we're gonna do it later is we've got OSPF running as the dynamic routing protocol and OSPF is all about cost that's how it calculates its metric it's gonna say the lowest cost wins and the fastest link the link with the highest bandwidth that's gonna be the lowest cost that's what we're gonna try to use but if it gets saturated if the delay is too great over that high bandwidth link IP SLA can say I know it's got a better metric but it's it doesn't meet my criteria so we're gonna go an alternate way I cannot wait to show you how that works it is super super cool but going back to the voice option for just a moment here's what I used to play with and we'll set this up today as well we're going to use a probe that looks like real-time transport protocol or RTP and that is what carries voice and video streaming media and I can say that what we're trying to measure with IPS la is jitter now let me define jitter for you cuz we're gonna be working with that in a few moments jitter is the gap typically measured in milliseconds between the time that different voice packets arrive I mean here comes voice packet number one maybe behind that we have some sort of web packet and it was delaying the next voice packet and it comes in well what was the what was the duration between the time packet one was received the first voice packet and the time we receive the second voice packet maybe it was 30 milliseconds well we could set a threshold to say if the if the jitter gets too great then let's route to call some other way and we can also say how often do we want to sin these probes we set a schedule and I'll show you how to do that in a few moments we could say for example we're gonna be sending out packets every five minutes I mean we don't want to flood the network with all of our probes and maybe it only makes sense to do that during business hours so that's what IP SLA can do for us I want to take you out and show you an example now so let's go out to our our demo gear and I've got a fairly basic topology and by the way if you're taking some notes on this IP SLA has gone through a lot of name changes I don't know of any technology where cisco has changed the name of it so many times this feature that we're demonstrating used to be called the response time reporter or RTR the response time reporter but there was an older version of that which was also called RTR it used to be called the real time responder and again I first worked with that and that's when it was called the RTR back in the early 2000s now it's called IP SLA for service level agreement but they're all the same you can use those terms interchangeably response time reporter real-time responder IP SLA and when we configure IP SLA today we're still gonna see some shadows we're gonna hear some echoes from the old response time reporter configuration we still might have the occasional RTR command that we enter now in our first example if you take a look at this topology on your screen we want to have router r1 send probes down to router r2 and those probes essentially our packet set that put on a disguise they they do a little cosplay action and they put on a disguise to look like different traffic types such as voice packets and Winn router r2 at the bottom receives those packets it's going to send statistical information back to our one saying here's how this packet was treated over the network here's how much delay we experienced here's how much jitter we experienced that's why we call our to the IP SLA responder because it's respond to the source of those probes with statistical information and to begin our configuration let's go over to router r2 that's gonna be our it's going to be our responder and we'll go into global configuration mode and I'll say IP SLA and give some context sensitive help and you'll see that one of the options is responder and that's what we want to do first we want to enable the the IP SLA responder so we say IP SLA responder and let's confirm that it is indeed active let's do a show IP SLA responder and we can see that general IP SLA responder is enabled that's what we're looking for now by the way an IP SLA configuration doesn't necessarily have to have a responder I don't want you to think that it's always two parts there's always somebody sending the probes and always somebody responding to the probes it doesn't have to be that way in all cases in fact in our next demonstration we're not gonna have a responder we're gonna be sending ping packets out and when we receive the ping responses back we're just measuring a round-trip delay so we're not even going to have a responder so please keep that in mind it doesn't necessarily have to have a responder now let's go over and that was simple was it on our - we just gave one command boom we're done let's go back over to r1 now and let's configure it to send those probes out into the network and let's make those probes look like voice packets shall we and we'll go into global configuration mode again and I'll say IPS IPS LA and you can give a locally significant instance number we can have lots of IP SOA processes running at once so I'm just gonna say one I'm only gonna have one but I'll say one and that takes me into IP SLA configuration mode and if I do some can do some context-sensitive help here notice we can emulate we can put on disguises for lots of different packet types the one we're focused on is UDP jitter because voice packets RTP it uses UDP user Datagram protocol and we want measure jitter so I'm gonna say UDP jitter and then we can say alright where are we sending these probes to and that's gonna be the IP address of the responder and if you look on the topology you'll see that that is at an IP address of two zero three zero dot one 13.1 uh I'll see you excuse me that wouldn't have worked yeah it's a dot two it's that top interface that top interface on router r2 we're not done yet when we send voice packets in the Cisco world they typically use a UDP port number in the range of 30 to 7 are excused me 16 384 through 32 7 67 so it says all right what port number you want to use and I'll say I'll just use the first one that is often used for voice it's gonna be 16 384 by the way different vendors have different port numbers that are used for RTP but that's what Cisco uses and now again to continually continue making it look like a voice packet we can say what sort of codec is it going to use when we set up Cisco IP phones we can configure what codec what code or decoder they're gonna be using for example if I were using either the first top 2g dot 7 11 those don't do compression they give you a little better voice quality but they take up more bandwidth I mean payload only they take up 64 kilobits per second of payload and what you might see on a link that might be subject to being congested we might see the G dot 729 a codec it only has 8,000 bits per second of a payload of course we have to have have header but the payload is one-eighth the size of G dot 7 11 secondly choose that one I'll say G 729 729 a and we still get to be really specific about what this voice packet looks like we can say how big is the payload what is my codex size in bytes the number of bytes in the payload well by default on a lot of Cisco gear G dot 729 packets they have 20 bytes of payload oh by the way they also by default have 40 bytes of header can you imagine that the header is twice the size of the payload that's a quality of service thing that we deal with in in another discussion but we can we can get around that but we're done we've been very specific haven't we about what our probes look like their UDP just like voice they're gonna measure jitter just like we need to know about voice we're specifying our destination we're specifying a voice port of voice codec and the codec size great so we have now defined what our probes are going to look like now let's set in fact if I give some context sensitive help one other thing I like to set is that is the toss byte the toss byte that's the type of service byte that we have in an IP header and now a byte is eight bits and I'm just debating on how deep to go into this if you've heard me teach quality of service before and get into quality of service markings you're gonna understand what I'm talking about we typically use a marking called dscp differentiated services code point in fact pop quiz for anybody that knows what dscp value should we use to mark our voice packets chat and in if you would there's a best practice recommendation how should I mark my voice packets with dscp quench at that in if you would not cs5 voice payload packets yeah yeah Adams got it excellent Java Adam Adam says it's 46 which is called expedited forwarding you know we had a guest yeah if you're signed in as a guest in chat roll and really appreciate it if you put in your first name that way I can I feel like we're having more of a one-on-one conversation if I could use your first name but yeah it's 46 this is a little bit weird and it's a big binary discussion and that's really not the point of today but for those of you that are already familiar with with dscp markings you know that it uses these six bits to give us that 46 well if you stuck on two extra zeros to round out the full byte at the end of those six bits where we had a total of eight bits and you you calculated the decimal equivalent then yeah it's it's not gonna be 46 it's gonna be 184 and oh I need to specify TOS 184 so I'm giving this the same quality of service marking that a voice packet would have now that we've defined the probe that's gonna be sent out we need to set up a schedule to say when are we going to send it how often are we going to send out this probe and let's say that we want to sit but and I'm gonna do it pretty aggressively because I want us in class to to get some meaningful statistics I don't want to have to wait for a couple of minutes to get one probe so I'm gonna say I want to set the frequency to every five seconds now let's say that we want to schedule those probes to be sent out starting now and I want them to go forever so here's how we create a schedule let me back at one configuration level and I'll say IP SLA schedule and I'll just give this a locally significant number schedule one and I'll say the start time is now I want to start right now and we'll say that the the lifetime is forever until I shut this down manually and until I change the configuration then it's going to just keep running and every five seconds Bam Bam Bam it's shooting out those packets that look a whole lot like voice packets and we're done with our schedule so we said who are we pointing to we set up the how our probe is gonna look and act we said how frequently are we gonna be sending out those probes and we said go if we said start right now so what we're going to do right now is we're gonna take our first cue and a break because I need to kill some time while those probes are being sent out every five seconds so if anybody has an in fact I'll go back on camera for this I feel like we can talk better if I'm actually sort of seeing you face to face but if you've got questions I'm looking at both the YouTube chat and the and the chat role if you've got questions on the new enterprise exam or the encore exam specific interpretation or if you just want to ask just whatever if you've got questions I've got opinions and I can share those opinions with you and sometimes the answer is I have no idea but I'll be happy to to give it a shot let's see do you actually have any video specifically for quality of service yeah in fact I've got a couple or three of them out on YouTube in fact one of my most popular my most popular youtube video of all time is is quality of service on Cisco Catalyst whichis so if you just search for if you search for catalyst QoS it's Kevin Wallis you'll find it and I've got another video on YouTube I think it's about half an hour long that goes over the fundamentals of QoS I think I might have another one that goes over the markings so yeah just check out my youtube channel my channel is Kay Wallace CCIE by the way Kay Wallace CCA what is the best on CCP Encore book to get well the one that Cisco press just recently came out with is the is the best that I've personally used I haven't full disclaimer I haven't seen any other encore books out there to compare it with but I mean it's a solid effort it's a it's a good book that came out with for sure why what exam do you recommend to renew the NP cert this one I would recommend encore for that let's see do you have any security modules to prepare for encore well yeah in the encore video course and that's what we're doing here today we're celebrating the launch of our encore complete video course we cover every single topic on the blueprint I'll be showing you each of those topics I'll be showing you all of our content videos I'll just list them out it's page after page after page yeah we cover everything you need to know about security let's take a look at YouTube over here let's see will your new video course to be available on Ohio that's a great question and the answer to that is is no it's not going to be available on O'Reilly previously previously a lot of the courses I did went through Cisco press as as the distribution partner and that's not happening with with Encore there's a lot of behind the scenes stuff and I won't get into but it's it's it's good it's all good and I'm happy the way things worked out but now the bottom line is this is not gonna be available on encore excuse me on Safari so if you think I've got a safari membership I'll just watch it on there it's not gonna be available there why did Cisco decide to remove yeah GRP from encore that's a great question and here's my again total guess I've not heard anything official from Cisco but what that really reminds me of is when Cisco used to be a huge proponent of ISL interests which link as their trunking protocol even though there was this great industry standard called dot1q cisco really pushed theirs I think it's kind of the same with ERP and OSPF EIGRP was developed by Cisco even though they opened it up to the industry not many people took him up on that and OSPF by far based on my surveys of my students OSPF by far is the most popular routing protocol inside our organizations of course BGP between organizations but OSPF within organizations I think they did just because OSPF is is kind of leading the game right now now not to denigrate ERP when I worked at Disney we had over 500 routers all of them ran yeah ERP I love ya GRP and just because it's not on encore doesn't mean you're not gonna need to know what for your NP because it does show up in a big way in what's gonna be what I'm guessing for most of you will be your specialization exam that's called an RC M that exam is it's sort of like a troubleshooting course but you have to know a lot about OSPF and a lot about AI GRP it's just not in the Ankur course all right we had tons of question we've got hundreds and hundreds of people joining us here today so please don't take offense if I don't answer your specific questions I'm doing my best to scroll both QA screens and try to pick out things that I think are relevant but I think we've spent enough time now and take my picture back off the screen I think we've spent enough time now to to get some meaningful statistics so let's say show IP SLA statistics and it says what scrolled off the screen it says the type of operation that we're doing is UDP jitter and the latest round-trip time in other words the time it took to get from r1 to r2 and and back again that was 2 milliseconds we've also got the number of packets that we sent out it looks like it was about a thousand that we sent out and we see minimum and maximum and average round-trip times but let's see surah here's the jitter yeah here's what I was looking for the jitter it is averaging 3 milliseconds and that is fantastic Cisco says at maximum you should not go over a hundred oh well actually the industry says you should not go over 150 milliseconds we're at 3 so this is fantastic and when you get in the voice world they actually give a score to the quality of voice for example that codec G dot seven eleven that did not do compression it it sounds better than G dot 729 a that does do compression but it doesn't sound much better you have there's actually a way of scoring this and it's called the MOS the mean opinion score the mean opinion score and it is for point zero six in our case and I'm telling you for G dot 729 a that's really good gene at 7-eleven typically comes in at about 4.1 so personally if just human speech I wouldn't be able to tell the difference we've got a network that is behaving very very well probably because there's nobody on it it's just set up it is only set up for a lab environment and that's the first IPS LED Emma I wanted to share with you today our next actually our next demo that's coming up and that's how we're gonna be making routing decisions using IP SLA and I'll go back to my apology actually I've been a different topology a little bit more complex topology see here take a look at PC of the PC at the top and it's trying to get down to the server and it's gonna go down to r1 and then we have a choice we could go from r1 to be b2 for backbone to or we could go from r1 to be be one either path would get us to the server but notice the bandwidth of those links if I go down to BB - that's a gig link if I go down to be b1 that's only 100 Meg link so by the way I'm running OSPF which is only considering bein with OSPF says you're gonna go to be b2 because that's a that's a lower cost and yes I did adjust the the the reference bandwidth which is a whole OSPF discussion to make this work but let's just confirm that that's really what's happening right now let's go to our PC and let's do let's do a trace route to the server let's do a trace route to 190 8.51 dot 100.3 and notice the first top obviously is r1 that's our default gateway but the next hop ends in a dot v that's BB - if it ended Ana dot 1 that would be BB 1 so that's no surprise that was that's what we predicted OSPF would do what we want to do now is we want to set up IP SLA to say if the delay between r1 and b b2 is too great then we want to fail over to this backup round we want to start going to be b1 yeah the bandwidth is higher but if it's saturated that the more bandwidth doesn't really help us and what we're gonna do to check that delay is we're going to configure IP SLA to do a ping test we're gonna we're gonna ping from r1 to BB to and we're gonna measure we're gonna measure that round-trip delay time the first step if you're taking notes the first step we do and we're going to do all this on r1 of course because that's where the the packets are coming in from the PC on r1 on r1 we're gonna go into global configuration mode and I'm going to create for your notes and IP SLA operation I'll say IP SLA and I'll just give it a number locally a significant number and I'm now an IP SLA configuration mode and before remember we used UDP jitter I'm just gonna do a ping that's ICMP that's an ICMP echo packet oh yeah that go frame I should say no that's right it's a packet it's an ICMP echo packet a ping packet and we're going to send a packet out and it's gonna be destined for gigabit 0/1 on bb2 and we're going to source it we're gonna source it from the incoming interface on r1 so we're coming in on gigabit 0/1 that's where we want to source this ping from so let's set up the characteristics of our probe I'll say ICMP - echo and I'm gonna say - 03.01.2012 B let's see it looks like 190 2.0 2.1 and that's all I'm gonna say about the probe it's just a basic ping packet coming from that top interface on r1 going to that top interface on BB - now we set up the frequency like we did before how often am I going to be sending out these probes let's say the frequency again I want some meaningful things to happen quickly so I'm going to say the frequencies every 5 seconds now here's where we get into that past selection logic I'm gonna say I want to say that if the delay is greater than the threshold then I want to fall over to my backup route that threshold is going to be measured in milliseconds now right now I don't want there to be an issue let's set it to 100 milliseconds surely there's no traffic there's no production traffic on this network surely the delay between R 1 and B B 2 is not going to exceed 100 milliseconds if it did we want to fail over to be b2 but we'll come back later and set it to something much lower but this is more or real-world I'll say that my threshold is 100 milliseconds the next thing we do is we determine the schedule that that this thing runs with it's gonna be every 5 seconds but when are we gonna start it when are we going to end it and we're going to do just like we did in the previous example we're gonna start it now and we're gonna run it forever so I'll say IPS la schedule 1 the life time is forever the start time is now what we need to do now this is where it gets a little bit trickier everything up to now is pretty much been identical to our previous example but what we want to do now is dissociate this IPS le operation here's a term for your notes we want to separate it we I want to separate it with mmm are not separate excuse me I want to associate it with a tracking object now a tracking object can keep an eye on something and it can give us a state of up or down well this tracking object is going to be measuring whether or not these ping response times are greater than the threshold or not if we exceed that threshold then this tracking object is going to be down if we don't exceed the threshold then the tracking object is going to be up so let's create a tracking object I'll say track and I'll just give it a number and what I'm tracking here is going to be based on my IPS oa1 operation which says sending these ICMP echo packets and my threshold is 100 milliseconds and I'm going to say delay down 10 up 10 let me explain what I'm doing here the reason I'm saying delay is if I just briefly go over 100 milliseconds I mean I'm like 85 milliseconds 92 milliseconds and just for a split second something happened and my delay got up to 105 milliseconds but it dropped right back down I don't think that's sufficient enough to reroute everything all of a sudden I want to make sure that this is a condition that seems to be ongoing so what I'm gonna do is what I'm going to do is say don't go to a Down State unless this condition of exceeding the threshold has been going on for 10 seconds and by the same token if it gets better if it drops below 100 it gets down to 98 but then it goes right back to 105 it was it was good just for a split second that's not enough reason to come back into an upstate so I say up ten I want to make sure that things are good for 10 seconds now as I mentioned earlier OSPF is is influencing our routing selection but remember OSPF in fact you tell me oh SPF has an administrative distance of what who can who can be the first to tell me what's the administrative distance are the believability of OSPF what is the believability of OSPF the the ad anybody know by the way there's like a 15 second by the time between the time I ask something and the time you hear it that's why I'm not immediately responding to your answers Wow almost every person saying you guys are doing awesome guys and gals almost everybody's saying 110 exactly okay well how can I inject your route that's more believable than 110 I can use a static route check this out I want to create a static route that says to get from r1 to this network at the bottom that 190 8.51 dot 100 0/24 network the network sits the servers on I'm gonna say to get to that network I want to use the next hop address of B B - just like OSPF told us to do under normal conditions i wanted to use b v2 as my next hop but I'm gonna put a qualifier on that I'm only gonna say I only want to use that as my next hop if and only if my tracking object is in the Upstate in other words my round trip delay has not exceeded 100 milliseconds for at least 10 milliseconds so let's do this let's set up a static route I'm gonna say IP IP route and if I'm going to 190 8.51 dot 100 0 with a 24-bit subnet mask I want my next hop to be that incoming interface on BB - I want it to be - 03.01.2012 if tracking object number one is up Oh excellently presto slice there only want to do that if tracking object number one is up okay great now at this point we're gonna be using a static wrap because a static rat has an administrative distance of a1 that's a lot less that's a lot more believable than than OSPF so we're gonna be using that it's going to say go to be v2 if tracking option 1 is up what if tracking object number two our number one is down what do we do then well right now it would just fail over to OSPF because that's the next most believable route so I need to create another static round and here's the key follow me on this here's the key I need to make this other static graph that's going to point to B b1 I need to make it less authoritative than the static graph that goes to be b2 but I want to make it more authoritative than the route learned by OSPF in other words it's only gonna kick in if that static rat going to be b2 goes down because I've got too much delay so I'm going to create a second static route and I'm going to give it a higher administrative distance to do that I'll say IP route 190 8.51 100 zero 24-bit subnet mask what's your next stop to be to 0-3 0.1 13.2 that's maybe 1 and here I'm going to give the number to remember a static route by default hazard administrative distance of a1 that's the one the one I just gave had an ad of a1 so I didn't even specify it now I'm saying I want you to have an administrative distance of a 2 so it's only gonna be used if that first static rack goes away oh I typed in the wrong address excuse me uh what did I meant to type in a what did I type in IP route to 0-3 0.1 13 oh yeah I typed in pardon me I topped in dot 2 I meant to say dot 1 there we go all right just double-checking yeah that looks good let's see what route we're using right now let's do a show IP route and right now to get to this 190 8.51 dot 100.0 size 24 Network I'm going to this next hop it ends in a dot 5 that's bb2 okay it looks like we're using BB - that's because the tracking object is up we can verify the tracking object is that by saying short - a show track 1 and the state is up and let's see the latest delay was only 1 millisecond that I was I was gonna set the delay to 1 millisecond to try to break it but we might not be able to break it we'll try though but here's how we one other command I want to show you a quick show IP SLA statistics show IP SLA statistics is another way to make sure that your SLA operation is running and is configured like you think it is now what I want to do now is I want to adjust the threshold from 100 seconds all the way down to one millisecond in hopes it might not but in hopes that our tracking object is going to go down now it's interesting that once I create this SLA operation and I start it I'm not able to edit it that is a heartbreaker out there I have to actually get rid of it and then put it back so I'll not give a lot of commentary here because we've already gone through these commands but over the next minute I'm gonna be saying very little and topping very much let's get into it I'm gonna get rid of what I did before I'll say no IPS cell a1 and I'll say I want to create IPS cell a1 all over again I'll be using an ICMP echo message going to two 03.01.2016 my threshold is gonna be one I'm gonna set the schedule to have a lifetime of forever and I want to start it now oh okay if we check the state of our tracking object now we might see that it's down let's see let's see show track 1 ah it is that look at this yeah we've been it looks like we've been having some to mobile second round trip delays so the state is down right now okay let's do a show IP route based on that down that state what is our next top that we're using now it's 2-0 3.0 don't one 13.1 it's bb1 how about that in fact let's double confirm that let's go back over to our PC and let's do a trace route to the server 190 8.51 dot 100.3 notice our next hop after the first router it's a dot one it's going to be v1 and that is a look at IP SLA again I guess until Cisco changes the name for a fourth time but that's a look at IPs [Music]

Channel: Kevin Wallace Training, LLC

Views: 8,554

Rating: 4.9899497 out of 5

Keywords: ipsla, ip sla, cisco, cisco cert, ccnp, encor, 350-401, ccna, ccie, #kwtrain

Id: HceUeV9llFg

Length: 37min 57sec (2277 seconds)

Published: Wed Mar 04 2020