AIOps Provides a Path To Fully Autonomous Networks

welcome to z cast everybody i'm zeus caraval from zk research and i'm back as your host for another one of my thought leadership videos i'm joined today by ajay pandaya the director of product management at masergy now a comcast company so congratulations on that recent acquisition we're going to be talking aiops today but before we do that ajay why don't you say hello and introduce yourself and tell us a little bit about what you do and who mr g is thanks for having me over at zk uh i am a director of product management here at mesa g now comcast business so thanks a lot for your wishes mr g is a world's leading secure cloud networking platform so we have been in the business for over 20 years we provide managed services for networking managed security collaboration and we've been helping our customers out in any of their telecom id needs 2019 this was the first time we launched ai ops uh as a network network assistance tool for it teams and since then we have not looked back we have added a lot of enhancements and uh this year we added another major major feature drop in our ai option that's i guess we are talking about that today before we get into that though i do want to give a shout out to e-week my immediate partner and all z casts are done in conjunction with e-week so in addition to watching here on youtube you can also read the summary on an upcoming ewik article now ajay um we're here to discuss as i mentioned zk research my research firm just ran an aiops study in partnership with masergy so uh thanks for funding that i know masoji's been an early adopter in aiops and just you know give me an idea why masergy wanted to study this topic uh well we really wanted to understand uh what is the benefit that ai provides to identities right so we we totally understand that id teams are totally overwhelmed today um the one of the study findings tells us that iit teams spend over 50 percent of the time just troubleshooting and figuring figuring out customer issues so it's like essentially i mean think of it as a workman or a painter just washing his paint brushes while uh the actual paint job is still on so we are everyone understands that ai has benefits it's just a question of how do we take advantage of this benefits and how do customers see what ai can provide in in optimizing or whatever they do with the network day-to-day jobs what can we do or what can ai do to make their lives better so we know that ai can revolutionize it's being used in so many other applications apart from it services but tell us we wanted to understand what is it that we can or what are it leaders thinking that we can help out with our ai of solutions so the idea was to understand ai ops the attitude the usage the business value as well as the journey where the where this technology is taking them in the future that was the main idea yeah and it's a good thing to study i'm uh before i was an analyst in full disclosure i was actually a network engineer uh i i'd say i spent probably a very good chunk of my time at least uh 30 40 of my time doing nothing but troubleshooting uh problems and so you know the and even back then i had a lot more visibility in the network operations and most network engineers have today because the rise of the internet things like that so uh but before we go deeper into the study uh let's uh levels up the audience and and define a app so as the name suggests ai ops is artificial intelligence infused into iot operations pretty simple concept uh very difficult to actually uh do it comprises machine learning behavior analytics and predictive analytics and so you know there's massive amounts of telemetry information and data being generated by network devices i suppose if you had spock running your network you might not need aiops but not everybody has a vulcan as part of their knock team so uh you know for that we need something else so it makes um i think a good way to think about is it makes anti infrastructure smarter more responsive and more secure it and one of the keys to apps is it doesn't just give insights there's lots of inside platforms but it actually gives you recommend gives you recommendations on how to repair that problem and so we've had good alerting tools for a long time the question is what do you do about it and um you know finding out what the source the problem is the recommendation to repair is actually where a lot of the troubleshooting time is taken up i also think it's a path to fully autonomous networks so uh if we're really going to get to this point where we can automate network operations we do need ai for that and i think it's a you know a good way to think about it is you may there's a lot of good automation tools out there and they work pretty well in static environments but we don't really have static environments in there right everything's always changing and so how do we how do we keep those automation tools uh up to date with what's going on we do that through ai so it's a hot topic with both vendors and buyers today it's it leads to some confusion that i'll talk about later um i think um you know it's something that i've noticed a big change in in the attitude of network engineers a couple years ago you mentioned ai and they've been go you know i don't want anything to do with that but now they're very open to it so audrey give me some examples some real world examples of how mesa app tool works well we call our ai of stool as the friendly 24 7 network monitoring agent or friend that is always on the lookout for what is happening in the network so things like alerting earlier when things might go wrong for example figuring out if there's going to be an excessive packet loss now this is something that is proactive as compared to react right typically we get stats when something starts going wrong by the time things get resolved uh the ship has sailed so we use a available data that includes things like traffic patterns we we look at people's usage patterns and based on that using algorithms figure out what would be happening based on certain preconditions that are met and this automatically gets enhanced as and when the system goes through further learning so things like for example as i said excessive packet loss or even things like video applications getting routed over the most optimum path that that would be one of one of the things which would completely seamless to the user under the hood so that the user gets the most optimum video or audio audio conferencing experience also things like uh configuration errors for example idea ip addresses overlapping addresses could be something that are endemic to the to the system and that would be automatically detected and also notified to the user or ask the user will be recommended to reach out to the service provider right so many different things that would eventually result in a not ticket or a support ticket would be automatically detected in some cases resolved or at least notified to the service provider so that they can work on it to resolve it yeah i think i'm liking the apps to you know be the tool to help you that find that needle in the stack of needles right there's so much data being generated today even the best network engineers can't look at the data correlated manually and figure out what's going on and i know a lot of really good network engineers and uh but you know obviously the the volume and data has gone up so uh before we dive into study though i i will let's go through some of the demographics it was uh 500 u.s based companies a little over 500 uh it was all i t decision makers and hires so we had some c-level executives in there we studied seven verticals technology manufacturing retail healthcare finance media media and communications and pro services we only studied the enterprise class company so they 250 million in revenue and up to 10 billion i think below that the networks get pretty simplistic and uh so you've seen the data i've seen the data what's the most surprising thing that you learned from the findings i see that there's a lot of excitement i mean um if i if i put myself maybe five years uh ago then uh i could have easily seen people treating the this whole ai thing is like i don't want to lose control over the operations of my network right now there is this certain level of trust that has come in so i mean the data one of the things like things that i really found interesting was that 64 of iit leaders are saying they are already using ai ops over 50 of users i believe 55 users say that they are using it for both networking and security together and almost everyone like about 90 percent id leaders said that it is important for ai ops to play some role in management of the network going forward so that was that was really cool i mean again this is survey results right uh it's not the reality uh but i know that you also have something to share so do share what what this study also tells you yeah you know the number that the 64 number of companies saying they use apps i think is both a good news bad news number i i think on the positive side the majority of companies almost two-thirds are looking for an ai op solution to help them manage their network and i and i and i think that's that's good because they actually recognize that there's a problem the downside is i don't really believe 64 of companies are using aiops i i think one of the challenges for companies is that um uh you know there's a lot of there's a lot of solutions out there called aiops but not all solutions are are created equal some are simply rules-based systems and i think this is where i t buyers need to actually you know be careful and so machine learning is the key does the solution get better over time uh in fact a lot of app solutions like i know i've talked to you know mesa geo but theirs it wasn't you know perfect day one it made some errors but over time because ai ai is a closed loop system it takes the incorrect data feeds it back into the system and that becomes part of the training set and then it becomes smarter and smarter over time and so that's an important thing for buyers to look for does the solution get better over time and this is this is why due diligence is critical though when you're choosing an app's partner so uh it leaders are you know from positive side again media apps with open arms and an essential path forward but they do have to make sure they're making the right decision uh the study also shows that companies are using and benefiting from aiops i think one of the interesting couple of data points is the top use cases for respondents include cloud application analytics performance improvement network service optimization faster threat detection and response in fact 64 percent of users measure app success based on it operational assistance efficiencies and a number 54 another 54 on improved network or app performance and this to me that what this reveals ajay is that there's real business value in an app's investment both from making ai more efficient but then also improving uh business operations and so how are you seeing masergy clients benefit from aiops absolutely so i mean think of it this way one of the findings that we were really surprised with was that it takes on average i.t departments over 30 minutes to fix and resolve a customer facing issue 30 minutes imagine like uh your service down for 30 minutes so that is like that is fundamentally wrong and what we also observed is that with ai ops video conferencing critical application performance is significantly improved um in just even addition to ai optic even we recently also talked about performance edge one of the other solutions that we put out where uh critical uh traffic is assured even on best effort services like broadband right now again ai ops does play a role there as well but overall is basically your critical applications that are always observed by this machine learning system so that the traffic patterns are understood under the hood and and the network responds to those needs dynamically yeah um some of the data points you gave or the storage spell i think is a good lesson here right if you're improving video improving app performance uh that actually helps the business improve productivity and so i think the lesson here is that while aiops can indeed lower the cost of running it to me that's not the end game of aiops it's about the transformation of the iot organization and the transformation of the company and so i've always thought if you do an i.t project simply to lower cost there's better ways to do that right and so ai ops is another example of that where do it for the right reasons which is business and i t transformation so another thing we learned from the study is that the prep preparation is key in the step is a key step in the journey in fact the thought process of you know measure twice cut once is you know is probably appropriate here but more than two-thirds of companies 73 to be exact identified sd or software-defined network modernization and visualization as a top investment priorities as a prerequisite for ai ops and so um you know i think as uh masergy has been both an aiops pioneer and one on sd-wan and so what's your take on why there's such a strong linkage between ai operations and sd-wan well ai i mean anything is only as smart as the data that it works from right so if you take a non-software defined network it is basically not centrally managed centrally orchestrated uh and the information essentially is siloed like it's siloed in many ways like it is siloed across an ip and optical network for example or siloed in control plane a data plane so now what i'm trying to say is that because of that centralization that sd-wan provides because of the centralization that's sd-wan with uh sassy for networking and security provides that data is now available which is much more enriched and that data is now able to give us a lot more information out that our ai algorithms can operate much efficient like efficiently as well as more optimally right um the study also showed us that over 70 percent of users found that ai ops performs really well in a sassy architecture yeah if you think about how the network is going to evolve with the distributed workforce and and and thin branches sas is going to be uh a lot more entrenched in in enterprise networks in the next two to three years and that's where ai ops is going to come into picture so i mean i just feel that uh this is the right approach for a software defined anything and as a result the new age networks so new and enterprise networks are really uh very good candidates for ai ops and this is where also messages yeah i i think an easy way to think about it right is that um uh when you there's a lot of people ask me how to define software to find and there's a lot of technical definitions about separation and control playing the data plane and stuff nobody understands but i i think when you when you when you do something in software it's programmable so if i want to use ai to be able to help reconfigure my network it needs to be programmable and so legacy networks weren't right where software-defined ones are and that's why i think that that linkage is there i'm not even sure how you would manage you you'd have to create a lot of middleware if you were actually trying to use legacy gear and so that would be non-optimal as well so uh but you know like you mentioned the beginning there's a lot of confidence in pressing aops today and the study also uh revealed some um uh interesting data points on where it's been taken in fusion so in the future so 84 see ai operations is a path to a fully autonomous network um 86 uh expect to have it a fully automated network in the next five years which i thought was aggressive but you know we'll see and 97 are confident that ai ops can be trusted to act alone and so those numbers are really kind of fascinating right so it you know the majority of companies 97 that's almost the entire respondent base now feels ai is at a position to be trusted alone and i think part of that is just that people make a lot of mistakes and so once ai gets over that threshold of people making mistakes it's now better than people right and i think that's people and buyers don't understand that and the 86 that feel that they can get to a fully automated network in the next five years um that's great from an industry perspective i don't really know if we'll get there in five years or not it's a little bit like the journey to a fully autonomous car is it gonna happen in five years it'll be in that time frame though and that's going to be kind of fascinating to watch and certainly there's going to be an area for a lot of investments and so the investments that masergy's made you know should pay off here so but why do you think creating a fully automated network is so important today aj and and how is masergy aiops helping his customers get there so just just to kind of reiterate on that earlier point um this is a this is a journey as compared to a destination so having a fully automated car might not be something that we will see in the very near future but there will be a lot of functions in a working of a car that will become automated same thing about networking right uh we might not end up creating the next skynet but at the same time there will be areas of of network management that will take benefit of automation so i mean there are lots of things that are happening with the pandemic for example hybrid workforces is essentially now a norm right uh people are working from anywhere people are working from their home people are working while being remote coming into the office every once in a while uh lot of iot proliferation uh and as a result the amount of attack surface that security vulnerabilities open up is is just like uh something that is going to be very difficult to marry to manage so as a result uh as applications move towards cloud um as there is this whole digital transformation that is going to take place in the networks there's going to be more and more use of automation in different aspects for example in troubleshooting or in monitoring or in in analytics and ai ops is a tool that is that covers all that with its uh algorithms that get better day by day and as a result we see this increasing trend that uh it leaders are starting to put trust into these tools that we are we are now okay that at least some sort of automation is available that is smart enough to help us in our day-to-day operations reduce our costs and also make us more efficient right so we already laid the groundwork for this in some way in the next few years you will see uh products coming out of of messergy that will have things like closed loop automation and that is going to essentially what i would call self-healing self-managed self-organizing networks right so the idea is that aiops puts us in the right step to achieve that goal all right and uh uh you know i've been looking at the clock going a little long here so i'm going to wrap this up but i want to leave the audience with some key takeaways here and i think the you know the key takeaways from the study to me were um the era of aiops is here right and so you need a strategy uh the majority of companies are looking at it if you are not you need to start the investigative process now because if you don't you're going to be left behind and so take that as a word of warning and uh in this fast food area you don't want to be left behind second you need the right platform so ai ops requires both sd-wan and sassy and so you want to use a provider that can bring you all those together it's one place for network and security data one place for automation and you know the 97 percent of people uh you know do do trust aiops to uh to run it to run the network itself but autonomous networks fully autonomous networks it's impossible to get there until you move your software and security or sorry your network and security into software and that requires sd-wan and sassy we are moving into a period of prolonged work from home or hybrid work right if you just look at the trends i see you're on the office uh you know but i'm sure you're working in some sort of hybrid mode as is everybody else and the new recipe to solve this this work from home and hybrid work problem is plus sassy and so you get better uh predictable application performance you need that or your users can be calling your help desk all the time smarter security for remote work right we have to be able to give corporate great security to everybody at home and the only way to do this through sassy you want one platform if ever we've needed that single pane of glass for ease of management it's today it's impossible to correlate information across management platforms and so work towards that and uh and like i said apps is that path to autonomous networking it's been you know i think we've looked at that as sort of a nirvana state for networking for a long time but it's right around the corner and so normally i get the last word in badaji i'm going to give you the last word here so if you have any closure remarks anything else you want to say so we are particularly excited about ai ops as the path for especially automation right so we build uh the best in class networks we just feel that for any of our customers success their i.t network has to be cutting edge and we really strive very hard to provide the best experience for them to to use the network and our products right and the idea is that that network providing the maximum value uh for the investment that they put in with this kind of smartness is something that we just feel uh very excited about and i hope that that our customers also feel the same way i i welcome everyone to go to maistrage.com check out ai ops we have a lot of content over there and apart from that thanks a lot for having me over all right well aj pandya from uh misha thanks for joining me if you're interested in that survey uh it's on both the zk research website and masergy website so uh feel free to go download uh download and learn more so again thanks for joining me aj thanks for watching don't forget to click subscribe and i'll see you next time on zcast

This transcript was generated automatically from the video's captions and may contain errors.