Cisco DNA Assurance with Tim Szigeti

so thanks for having me my name is Tim Sogeti I'm a principal engineer in technical marketing I'm the lead architect on DNA assurance normally when somebody's asking to do a 30-minute session they prepare about 15 to 20 slides but I'm gonna try my best to take a page out of the Dave Zacks book I have 62 slides nine demos that I'd like to get to running on six different machines in three different countries and and five of those six machines are not running shipping code we're gonna try to show you everything we possibly can neat cool new stuff as much as I possibly can and the demo gods will cooperate with me my sessions are called DNA assurance the shortest path to network proposal but I know a tech field day you guys like to be a little edgier and so I've renamed this content people pain and path trace so if you look at the CCO today and yesterday the headline page is CCO is about DNA analytics and basically here's the business goal in outcome to free up IT time overwhelmed with burdens on monitoring troubleshooting maintaining your network instead of innovating instead of contributing towards digital transformation so that's what we're here to drive the key here is intent based networking but it's not just enough to express intent and a controller and have the configuration pushed out you need feedback you need the context it has actually delivered on your intent and to complement that context we have learning constant learning machine learning we'll talk about that let me elaborate a little bit on what I mean by context so we have a big data problem we have a series of challenges we have tremendous volumes of data generated by the network up to a terabytes a day they're coming in at multi gig speeds there's a variety of formats streaming telemetry NetFlow SNMP sis logs as well as all these api's from other contextual sources and then the veracity sometimes we have multiple instruments that could report on the same metric like for instance IP SLA could report packet loss but it's not nearly as accurate as interface counters our job here is to take that data and either if you want to use the analogy of you know take crude oil and keep refining it until it's something valuable or another net analogy is like panning for gold you got tons and tons of dirt and you're sifting through it sifting through it processing it until you finally find that which is valuable data an example of data you can give me the timestamps on packets are TP timestamps on voice packets that's data if I start processing that by comparing them doing a delta I now calculated jitter that's a little bit more valuable to you but I can also contribute expertise and saying this jitter value for given applications say WebEx is within range 83 milliseconds or 83 milliseconds per voice is out of range and I can give you a translation and little more information in context regarding that calculated and it processed data and then I can go further and give you insights where is that 83 milliseconds of jitter being injected in your network and why and then ultimately I'll recommend you an action to fix it that's the goal here here's an example I got a user that calls in and said I've had a really terrible WebEx experience well I know who the user is he's identified himself to me so I plug I connect to ice it gives me a little more information about it including the devices he's using and its MAC addresses I say oh which one which device was it on your PC or your iPhone or whatever no he goes by PC okay great I'm gonna then connect to the IP dynamic DHCP system to get an IP address I'm gonna use that as a source key field on all the net flow records and I'll see all the different flows that have streamed from that address but I still don't know which one is the application until I complement this information with application visibility and control now I know which flow I'm looking at I'll also complement that with network topology information where was he connecting to the network which ap and which room also like taking into account geographic location and ultimately taking a look at what the network device is telling me a very very common error here is that hardly anyone deploys QoS on wireless LANs and yet the second most likely place in the enterprise network to induce latency and jitter is the wireless LAN so a lot of issues there I'm going to show you tools that can get right to that root of that problem this very case very quickly very easily the complement context we also want to do machine learning there's so many variables in the equation of things that could go wrong comparing all these variables is simply beyond you know and when these variables are in the hundreds or thousands or tens of thousands it's beyond any humans ability to examine all the permutations and do the analysis much less provide recommended insights and actionable actions however machine learning artificial intelligence algorithms can compare all this sift through it look for patterns and then if they find a pattern is that pattern just a coincidence or possibly even better a correlation or even stronger a causation because if you can identify a causation you could do predictive analytics you can see what's about to happen not only this machine learning could be guided or unsupervised unsupervised or guided unguided whatever terminology you want to use and that way the operator can really help the algorithm say you know what this insight is actually very valuable to me but that that I know doesn't have anything to do with the problem so keep looking for more of these can I ask you about the previous slide how much of that is PowerPoint how much of that is shipping yeah let me show you some demos so I want to get to the demo so I want to get through the powerpoints in like 10 minutes flat and then get right into the demos now the other thing I want to share is that our whole approach to DNA has not just been about technology which has been the traditional Cisco engineering approach for the last I don't know 30 years whereas at the beginning this project all the architecture team was sent back to school to the Stanford design school to learn about human design human centric design and realizing that it's just as important to consider the user interface the human interface as it is the technology and the business cases that you're trying to solve it's a different concept we always thought well if we have the best technology you're done but you're not because the technology is too complex to adopt we were talking earlier about QoS people will not adopt it they'll not leverage it we have to take into very hot top-of-mind account what our users need and what how we can help them and empathize with their pain points that being said we had fired at least for arch types of users from a series dozens like 60 different interviews from the majority of which we're blind and we keep on doing ongoing user studies there's one that's actually happening across the hall there's a team here doing that we do executive technical advisory boards regularly twice at least twice a year and even more and so we're in constantly soliciting this input and feedback from our users what they need what their pain points are so we got about four different arch types of users there's a frontline guy these guys basically they're trained to solve problems by just being thrown into the water that's how they're taught to swim it's like they got nothing they got no training they don't know how to use the tools and they got plenty of pain they get penalized if they don't solve a problem quickly they get penalized if they don't escalate to the proper location making an effective triage decisions they get penalized that they don't know how to use the tools that they weren't trained on they spend about 90 plus percent of their time troubleshooting very little business impact one customer called his NOC he goes these guys are just training monkeys looking for red lights okay keep that in mind I'm going to refer to that this is a very common recurring theme we keep hearing over and over and over again it's not a network problem sixty percent of the time here these guys are saying it's not a network problem but what we're being blamed like it is our basic job only 60 yeah okay great I hear a lot of customers saying you know my prom primary job as a network administrator is mean time to innocence how quickly can I prove it's not my fault the network's fault that is he escalates to tier two tier two he's like oh you're thoroughly I got too many things coming to me that should have been solved at Tier one and I got too much raw data I don't want the raw data I just want the insights don't just allow make me force me to interpret what all the data means he spends about two-thirds or nearly two-thirds or more of his time troubleshooting and his pain points are yeah you got to interpret that data for me just tell me if the events normal depending on how effective the tools we can have we can identify and close remediate more and more issues at the lower tiers if tools are effective the expert these are the guys in the tier 3 these are the guys that can really drive network transformation if they have the time if they're able to work on projects but what's their pain point there's too many problems being escalated to me and a large number of time it's not I've spent a large number of time proving it's not the network broken record this theme these guys have a huge business potential business impact but they're not able to do it they're spending nearly half of their time troubleshooting and half of the time that they're troubleshooting it's again proving it's not a network problem that's very frustrating and then the final use user is the planner who's trying to like you know determine strategic direction of where to go with the network she can't get very far because she's constantly waiting for input from these lower tiers that are constantly firefighting and troubleshooting and proving that it's not the network okay so those are our primary users we're gearing for we want to deliver products that can allow the tier 1 to solve as many cases as possible that's has a trickle-down effect to the higher tiers and everybody then gains incrementally more time to focus on strategic and innovation innovative product projects we did a whole that's a that's a good way to put the problem because what we've been told and what we keep hearing in you know all the marketing presentations is usually the fact that you're trying to make networking easier which in my opinion has the risk of turning everyone into a trained monkey looking for red lights ok cool that's a good point so we don't want to make networking easier per se we want to deliver powerful Network solutions like SDA but make it simple so you know power and simple that's a sweet spot and that's really where that user design needs to come in to abstract the complexity intent-based networking automation and so forth so the goal here was to design a product that could do a single pane of glass design and provisioning assuring a network this is a screenshot I took yesterday you see this is running here at Barcelona 13,000 719 clients peak traffic today what do we got here I'm just going to pull up takes a little what do we got we got five thousand nine hundred and fifty five clients running right now it shows me onboarding times three thirty one thousand on boarding times less than two seconds etc I'm going to go through some of these screens but before I do I just want to provide a little more context and background but the point here is that we wanted to make something we've reviewed the the screens and the mock-ups and what we're going to build with our users before we ever started building a thing and then this front panel UI is just the tip of the iceberg that's a part that's visible but there's so much more going on underneath architectural II that it's just taken for granted so I just want to quickly cover that the long term North Star goal here is to build a self-healing network now that probably sounds like science fiction and it is it does not exist however what would it take to build that let's look at that and you might be surprised where we are in that journey so the first step is instrumentation you cannot analyze learn from report on that which you have it first measured so we got very rich instrumentation in this ASIC how many transistors do you guys know seven point four six billion transistors one for every person on the planet three hundred and eighty four thousand flexible counters I can measure a lot of stuff I can store 128 thousand net flow records all here in the hardware outline rate I not only want to instrument the network I want to instrument beyond the network I want to see beyond my edges to get the whole end-to-end experience I want to instrument using Apple iOS analytics to get information not just from the APS but also from the clients of how they're viewing the world instrument even from synthetic sensors the dedicated sensors that could be sending synthetic traffic or that you could use a flexible radio to act as a sensor whenever you want to not only this but instrument what's happening on the app patience ID whether it's direct API is like I'm gonna try and show a demo on skype for business integration or via agents app dynamics and the power that brings to the whole context that we can see not only this then we recognize it if we have to send 384,000 metrics off the box at line rate that's going to overwhelm everything then the bandwidth of the network the processing of the analytics everything so you want to distribute some messages are more important than others a high CPU is very important to know about more interesting than say the interface how many packets went in and out of an interface you still want to count that but it's not as critical a message the critical messages you want to stream off using streaming telemetry model-based streaming telemetry we can get this we have this on an iOS devices as well as their wireless LAN controllers WL C's they can stream in the the case of air since you guys were talking about a featured in the demo every five seconds it can send updates very very efficient you got to catch all this data so you got to have some significant storage whether it's centralized distributed or cloud-based but then you also don't want to store all the raw data indefinitely and it doesn't make sense to the client calls and said I had a poor experience with WebEx and you go okay when three months ago it's like you seriously want me to troubleshoot that it's like after a while you're going to be compressing the data may be getting averaging it without losing the outliers so on but real-time data or raw data we keep for about thirty days but you could keep it longer now you need a tool to analyze that data what is normal what are the outliers of deviations from that normal baseline these anomalies and then allow you to provide recommended actions just just just a quick on that so you're storing 30 days the raw data and long-term let's say statistics about average and Peaks yeah and what kind of data sets are stored in long-term do you have their and I don't know you stored data point for every hour or what isn't about kind of that you stolen long term we store data points by every hour I'd have to check on the granularity like what happens after thirty days and then it probably be another time where it goes even more course the farther it goes into the back it's just basically a function of if you try and store everything it's going to be economically unfeasible and it's not even going to be giving you a good ROI so at some point you're gonna have to start compressing and aggregating that data rather than just keeping it in its raw form at events like just popped on my mind with baselining for example the super bones next year yeah super balls coming yeah sometimes you see even in the network some significant changes when sports event like this happened and for me is good to have let's say at least for one year the data that you can compare what was happening one year before yeah absolutely I'm just making the point that you're gonna you're gonna need some scalable storage as well as then some methodology that suits your business requirements of the granularity of how far back you keep that data okay some starch windows are happy to sell me as much starch as I want so if I have enough starch it's possible to store longer absolutely that's totally configurable I'm just giving the overall broad architectural requirements at this point not really specifying the the product specifications but I'm gonna show you the product soon as I cover this yeah so then machine learning complements our analytics because we don't even know what we're looking for and then the machines can run all their algorithms to spot interesting correlations and hopefully causations but it's not just enough to get to the root cause of a problem tell me how to fix it so guide my troubleshooting in my remediation and then ultimately I want to get to the point of automating that remediation but this is not just going to be a technical challenge it's not just a matter of hey if we know how to fix an issue we can we're just gonna go fix it this is a challenge of confidence you know after allowing the operators say hey this is what we were in recommend says yes take the action and then show us that he fixed it eventually we want to gain that confidence and this is a real challenge because for instance I'll give an example last year my wife got a Tesla the Tesla is a car that can drive itself and when we get took a test drive it was actually in downtown Manhattan and full busy rush hour the guy just stopped we'll stop touching the gas it was driving itself in like really chaotic conditions and very impressive but in the year since my wife has never ever used the feature she doesn't have the confidence in it and so this is going to be a technical challenge and a confidence challenge so I've laid out eight requirements for the self-healing network as of yesterday we shipped five we're going to be adding machine learning and EFT in March release and then in shipping in July sand that'll be six and then that will keep on going we're not going to be able to solve every problem in the network but we're gonna start with a set and then keep on expanding on that or and then building from there the product subsets are you solving now the top so we look at we interface with TAC and say hey what are the top issues that you're seeing or with customers what are the top issues we're seeing we're gonna start with those and we keep off building a library of issues it's not they I'll show you all the details if you want so I really want to get to the devil so I beg you that just let me let me try and this looks like an architect presenting the idea of Brasilia to us and then it took them 15 years to build the city I'm gonna show you what we built so far just let me lay it out so in the automation platform the main job of this ten minutes okay is to talk to the network the analytics platform is to listen to the network and then close that loop here our software functions within DNA see the network controller the network data platform and then we have applications riding on top of that we have over a hundred actionable insights I'm going to give you a slide that details every one of these and then you can see them for yourself let's get into the demos it's demo time do-do-do-do okay so we've already shown that this is running live on the Cisco live barcelonan network currently there's 59 155 clients the majority of them are having very good onboarding times and a client calls in and let's say it's me I'm like hey what's going on with my stuff I look up my user name and then I'm presented with a user 360 so from a client perspective I got two main devices Honda Network I have my MacBook which I'm presenting from right now and then Jerome has got he's logged in under my credentials and I've got my iphone right here you can see my health over time I can go back and forth in the time last 24 hours or longer or you know zoom in and I can get all the details of how the health score is comprised I can see any issues or trends like long time for onboarding authentication RF issues etc I can get my onboarding details what SSID I've logged into what ap what controller I can run path traces I'll do that a little later application experience it's in beta we don't have that enabled here on this network gives me details of my device and then I can even see very granular and up to date our SSI SNR transmit receive and what's really cool if I go over here to my iPhone my iPhone has one more element and that's that iOS analytics remember I talked about collecting information not just from the network but also from the client now it might take a few seconds to load so I think I had a pre-loaded screen of it well elk okay it'll tell me if I've disassociated and the reasons for disassociating if the device is idle etc and then also let me say I think I took a screenshot of this this was just earlier like just before I started presenting maybe about 20 minutes it shows all the individual APs my iPhone was associated with I'm just going to keep moving in the interest of time beyond this ok I've covered all the client things Klein 360 is it's a 360 degree view or is it the number of the client so that's a really good point so we have different levels of data aggregation we got the global health that shows client and network and we're gonna add application in the March release then we have individual like health pages that show this this is the health page that said ok overall all how are the clients doing it summarizes that particular type of entity huh client network and applications in March and then 360 is the lowest level view for every single entity it's all the raw data that we have and so that's what we mean by a 360 view it says this is the most detailed view that we have and then for every entity whether it's a client network device or an app we have a that low level 360 view it's everything we know about it we're working on this as well this is the application experience section of the client 360 view so for any given application you'll also see the the throughput and SLA is like packet loss latency and even the application delay what do you get this data from its this is being derived from the network devices for instance a router that's in path and it's doing art analysis in this specific example application response time looking at the Delta between TCP syn syn axe syn acts and acts as well as requests and the response that's how we determine the application delay so that's even not a network parameter that's measuring something in the data center but from the TCP we can sure that we can do then Network how this is network health we basically group the network devices according to their roles poor access distribution etc we can display devices either in a geographic view and if there's multiple sites we have a larger circle and if there's some issue that's highlighted you see I got three different health scores in the LA area and it's the Santa Monica that's come to the fore here I can identify the problem devices by the system health score and I can zoom in on anyone because these will also have their own individual 360 views so from their 360 views I could get their issues and say ok the tunnel here is it's flapping verify that the interfaces air freeze the suggested action there's going to be some more information coming up and then you can run you can run commands right from the UI so if you got a recommended action you don't have to now go and SSH and tell it to the box and find that interface and run the command it's all integrated trying to accelerate your and I point something out that you will not never see interface errors on a tunnel interface so maybe that suggestion is a bit off ok very good we'll take that we'll take that input you keep in mind that this is very very early days this is conceptually new it's 1.2 release we're doing our very best we could do out for instance here on this interface we have a whole series of suggested actions we could preview them all and then we can even run them all is just compress happy about this let's just go press happy about this with at once have any more book yeah well because what what you're not what oh there's gonna be books on DNA Center actually we're three quarters of a way through on a book on DNA Center with Dave's axe Matias Matt Faulkner myself and Simoni arena maybe some of those guys were in here before now here's some really I have a quick question you you you said you have the device you so you per device all the information so you get a interfaces serial number inventory all that is in DNA Center yes and that's on pram okay just just for my understanding you had previously in the past a product called Prime yeah so technically DNA is taking over all that was what Prime offered in the past more less yeah I wouldn't say quite that way it will it's a completely different paradigm for management this is prime the device based management template based device base there was no intent there was no abstraction of complexity in business intent and the analytics was it wasn't part of it nor lot of new functions but what I see all that was what was there in the past you also cover yeah yeah well we have to have a way of maintaining software we have to have a way of you know managing inventory and things like that so there's some basic network management functions that also have to be included in the management platform but it's I would not call like DNA assurance a prime replacement or a prime to dot oh because it's a completely different paradigm sure but if I want to have some functions that I were doing in the past was prime I can perfectly use DNA it has more function and more but all that what we have for example inventory I was looking for software version whatever it's all here yeah yeah the software versions and a way of managing the software is golden images and things like that it's all within DNA center and you said you can fix things directly from the DNA Center without having to SSH into the box so you have some sort of user rights management in there so not everybody in DNA probably is allowed to configure everything absolutely and so for instance on the NOC system that I was showing earlier I can't run any commands and I can't even do path trace and things like that oh I didn't show path trace but here let me just show it in slides path trace it's a very effective tool so you can run this from a client from a network device you guys already seen path trace okay cool so then you can see I haven't seen it but it's nice okay well you can as you do a path trace you can hover over a device get details you can even hover an interface and get key OS stats and then you can also if there's an ACL preventing a particular flow it'll call that out very quickly so if you're your trained monkey looking for a red light that's gonna help you that's already shipping now what I really like question is it only doing traditional traces or can you all for example in MPs networks do the tracing function yeah absolutely so trace route is not using I'm sorry it's not using trace route path trace is not using trace route it's using all the network topology from the network information base it also queries devices for when there's an equal cost multi path if there's devices that are not being managed like cloud devices or an MPLS it'll show you a cloud it's like I've said this was what happened when it went in this was the IP when I picked it up and so it'll cover those scenarios now let me show just a couple really cool things because I'm going to be axed for time okay so the triage decision is is a big one so if I'm a network operator the lol one analyst and somebody complains I've got an app issue how many times did we see them complain says it's not the network it's the app and usually this finger-pointing goes back and forth over and over again well we can get insights into what's going on in the data center via API is from app dynamics and that can help us make a much more accurate decision whether it's an app issue or a user issue I'm sorry network issue because they also monitor all the transactions and provide a transaction score card so yeah this is this is now not yet shipping this is something we're just featuring and demonstrating here and and we're working on but I think it's super cool so I take a look at this I see that application health score is a function not just of the network quality but then the application quality if I and this is coming from apt I am extends action score car to one okay so I can click on that and contextually now I cross launch into app dynamics and if it hasn't typed out I won't even be asked for credentials but everything has timed out but it's taking me right to that app so my troubleshooting workflow hasn't been interrupted I'm still staying within context I'm a monkey looking for red lights I'm gonna filter out everything that's normal and I see that this trend this application has got a lot of slow transactions I can browse around if I'm coming from the networking side I might be interested in the network dashboard and what's the biggest read there Oh server health this particular server has got an issue and I click on the red I'm just following these red lights and it's telling me oh the machine is to availability is too low it's detected a problem it's out of resources or here's another case I think this is really cool too so I wouldn't tear one guy I did nothing other than follow the red lights this other app it's like okay what's going on with that why does it have a score of 5 out of 10 please thank you thanks Brianna so this other app what's going on with it I filter out everything that's normal and I see a lot of errors Oh interesting so I can browse around I'm not going to go to all the screens it looks like actually I'm so in this view I don't see anything red I go on to my transaction scorecards and I can click on that errors and say what what errors are you talking about what's causing the errors and then I click just on the red and it's basically telling me if I drill down into it can you guys see or do I need to make this larger that the user experience it's taking a lot of time here and it's a slow SQL call that's the problem I've gotten to the root of the problem without being an expert on anything so that's that's that integration AppDynamics another integration we've done is with Skype for business so for instance the question that was asked how are we measuring some of these stats like latency and jitter and loss while we can measure it in the in the work but if it's in the network it's in the midstream if we take a measurement and then there's packet loss downstream we're not capturing that so it's much better to get that information from the client itself well we can get it via an API for some of these applications for instance you know with Skype for businesses or other apps they can just send us details from the clients from the app servers via api's and we can ingest that and then here's an example of of doing just so now this feature we haven't shipped yet but it's committed engineering committed for March so just a couple months so you you say okay the application score is 1/8 I click on it and I can see the mosque or as reported by the client and to end inclusive of the network inclusive of the final AP to the client any jitter that's occurring on that leg packet loss latency jitter ma scores all of this is taken into account as well as what's happening on the network beyond this somebody asked about machine you were you were showing that and the error was something with SQL yeah it's great but it doesn't tell me anything sure it does it tells me how to triage and escalate that if I'm the tier one guy doesn't okay but where's the problem so there's a latency somewhere you couldn't fix it anyway you just know that it's an application problem it's not on there for problem so you call that guy not bad guy over one where that one where the server had insufficient resources as a tier one guy and no now to escalate that to the app team or to the server team and it's in a rather than this ping-ponging it's a network issue it's an app issue I know where the root of this issue is I have the data points and the other case it was a code issue and again this is not something that the network team is going to resolve but he knows how to escalate it to the appropriate triage decision of network versus app so but but but but I didn't again so they escalated to another team yeah he's a tier one he was a tier one trained monkey he's not an expert this problem needs escalation so he escalates - which team whoever is responsible for this specific app the code the developers okay and he gives all of this so they have the same inside to the ticket I mean if the developer sees deadlock found when trying to get lock okay but they have the same inside so all the teams can go over to this it's helping to give visibility from the network team to the app people in the data center and vice versa because right now they only see to the edge of their domains and this is trying to share information across it that's the overall goal yeah okay I really want to show just a couple more things before I just completely get axed machine-learning somebody asked how real is that so machine learning our cloud-based machine learning is coming in EFT in March and here's some examples of some machine learning insights and basically it's following the same UI as DNA assurance you can click on given issues and then it's applying machine learning rather than individual experts like tMI's or engineers how to how to troubleshoot this issue it's looking the green area is all the predictive it's saying given all your parameters and what we've baselined and we know for the causations etc we can predict that this is the expected range of your throughput in this particular case a throughput issue you're falling out of that range here's a series of things that are commonly associated with that interference retransmissions this and that if you agree you can say was this helpful yes or no if thumbs up this is guiding the machine learning algorithm to look for more of those types of patterns if it's not helpful you remove it and then it's not going to waste its time and cycles on that you can get reports to of individual kpi's over the the your entire network how frequently these are occurring when they're occurring which devices they're occurring which clients are affecting all of this big data being processed for you not only that but then you can compare your data with your peers this is a data coming from a large University that I actually attended hundred thousand students and it's comparing what is your throughput like compared to other clients completely anonymized all the data is you know all the names are one way hash Det cetera and so it's completely secure that way but it's of interest because then it provides baselines how do you how are you doing compared to your peers the final thing I'm going to talk about oh you want to see that screen again okay here let me just okay there you go final thing and I timed out again this time out stuff is absolutely killing me is err sense how this was a tool that was featured in the demo yesterday and it's there's the different levels of tools at least three different levels of troubleshooting tools there are the highest level that always on what I call watchdog tools they're monitoring everything all the time and then you got mid-level tools that will be always on but maybe sampled or sensors that are running specific tests at specific intervals and then you got your low level tool and that's basically to allow you to divide and conquer when there's issues and then finally the lowest level tool is what we're talking about here where you really have to get to that root cause and you know like packet capture but enriched and and really low level insights and so that's what we're featuring here an on-demand tool called air sense if you think about for instance one one use case that air since offers is that it can do packet captures on any wireless segment of your network in real time from any location and if you think about what you would have to do ordinarily if you wanted that you say what's going on I want to take a wireless packet capture you'd either one of two options you've got to send somebody over there that's got you know a network driver that can operate in promiscuous mode and then take a wire short capture that's expensive and time consuming or you shut down one ap from servicing and have it do the packet capture of the other eight now with ear sense you just click on on-demand tools err sense and then here you could start a packet capture on any air P any ap and it's from the client 360 you want to say what's going on with this client I'm not even I don't even have to look up what ap is connected to it's from the client 360 I can trigger this on demand tool I can search through what's been going on I can very easily see errors because again I'm a trained monkey looking for red lights and I can see that there's some key authentication errors and I can not just give a packet capture and have to analyze each packet for output I'll be one minute literally one minute Breanna but it's isn't riched it's helping me to understand and decipher and zero in on what I want to deal with and then finally the final point we've talked about Wireless LAN key OS nobody does it and some people that do it don't do it right so if you want to take a look is it being marked correctly at wmm here's the tool to do it and I can do it very quickly and see what those markings are that's about it so what I want to leave you with is that look everybody's bogged down by operations the more efficient faster effective that we can make troubleshooting especially at the tier one layer it frees up everybody's time we have extensive tools for even the experts machine learning to gain deeper insights capacity planning's baselining comparing with peers application integration a lot of cool stuff still coming it's very exciting we're on that path very with ruff with great velocity towards that goal of that self-healing network we got a long way to go but it's very exciting what we're what we're doing at least I find it so hopefully you did too

Channel: Tech Field Day

Views: 8,060

Rating: 5 out of 5

Keywords: Tech Field Day, TFD, Cisco Live, Cisco Live Europe, Cisco Live Europe 2018, CLEUR, CLEUR18, Cisco, Tim Szigeti, Cisco DNA, network analytics

Id: o0VGIiu0ZPA

Length: 40min 58sec (2458 seconds)

Published: Tue Feb 06 2018