Empower your operations with customer pipeline management for Operations

How to create outlook signature

hello everyone welcome back to my Channel or a LinkedIn profile if you're watching the LinkedIn live and in this session today I think you guys already know but just you guys know we are going to talk about the aiops road map okay now if you're new to this Channel or my profile feel free to subscribe or follow and do let me know in comments from where you're joining the session if you have any further queries anything that you want to ask me I'll definitely answer in a short time okay now without any further delay I'll just get started with it if you're joining late later on you can just go back to the you know the beginning of the video and can start on okay so um yeah I'll not by the way do let me know from where you're joining that's very important and uh any query that you have okay so that I will know that you guys um have any question I'll clear that okay so now now let's get started without any wasting time okay so okay let me have my um phone also ready cuz I know that somebody will will comment on LinkedIn and my software don't know that part okay all good okay so very first thing uh like you guys U might know or might don't know so I will just start with the very basic like what exactly is AI Ops okay so AI Ops which stand for artificial intelligence for it operations okay and this refers to the art use of artificial intelligence and machine learning Technologies to enhance and automate it operation okay and aiops perform uh platforms leverages data analytics and automation to improve efficiency performance and reliability of ID Services now this is a very vague definition but from Wikipedia and similar kind of resources but now let's try to understand uh it's talking about what it's talking about usage of AI and for what using for the ID operation but now if you I mean you'll just uh think your mind like what are the exact it operation like what are the what is the it operation or what are the kind of it operation that uh we talking about will get handled by the AI okay now let's just try to understand that part okay what are those it operation where we can implement this AI option because without knowing why you cannot learn anything right you cannot learn aimlessly you have to understand what you're learning and why you're learning and where to apply that knowledge right so these are the you can see from the left side of the screen all the use cases okay and I'll just go one by one first one is The Incident Management okay and uh like it automates the detection diagnosis and resolution of incident so what what is this means on the incident there could be a downtime there could be a bug happening there could be side reloading and reloading or loading there could be a number of issues okay we call this incident or issues okay so using AI if you can automatically detect and if you AI can automatically give a resolution so we don't have to even wake up at 2 a.m. or 3:00 a.m. by client to fix that issue right that's a very good uh use of tooling right then performance monitoring which continuously monitor the performance application network and infrastructure and if there is any issue in performances it will automatically debug and fix it okay that's one of the use case of the this aops in the performing in the operation the performance monitoring is a part of the it operation okay similarly there is event um you know correlation analysis which is the root cause analysis then capacity planning and optimization which is the infrastructure part if you want to add more resources instead of manually doing it if AI can do that automatically VI the say past data usage or maybe predictive uh implementation so what happens it will do sampling of for example last 24 hours or maybe last one hour based on your configuration and if it's the spike happening we need to scale up so it will automatically do the scale up okay based on the intelligence metric it will identify okay some attack is happening then it will stop blocking the request of those attack automatically based on the IP or maybe some headers and all that and then it it will just block those malicious traffic and allow only the uh proper normal traffic so this kind of very basic use cases we're talking about okay then the predictive maintenance for example you have multiple Software System or say multiple OS okay there U you see there could be a issue could happen or maybe in nits upgrade or maybe maintenance or patching that also can be done by the help of aiops as a part of the maintenance then similarly there is Security operation configuration management automation of the routine task and lock management analysis all these cases we can this all these cases we talk about are part of the it operation and AI can help us in all this things okay that's the main point here like people know okay there is uh like we there is iops there is I me there is but they don't know where to apply that knowledge okay so that is where this is handy so you know that where to apply that knowledge okay now this is the part we just talk about what exactly is iops okay which is you have it operation what are the it operation these are the example it operation where AI can help us to automate to automatically fix these issues without making us up okay uh if of course if it requires the human intervention it should send us alert SMS or call automatically so those kind of part of the AI of implementation okay now I will not waste your time okay I'll just directly show okay this is the your overview of AI road maps Okay like from thousand fet VI going say it's a very high level overview okay so what what if you're thinking of learning and implementing aops uh in your organization how we should move forward first understand what are the exact it operation which is give you the example I'll just going in a bit deep on that then uh learning the AI basic because man you are handling you're going to handle the AI things right so without the some level of knowledge like zero AI knowledge you cannot survive in this field okay you need some level of AI knowledge to strive in this field okay then the third point is learn about the AI op so of course there is Operation there is AI combined we're going to have some a basic ideas concept how we're going to achieve the AI say implementation so that that we will cover then the uh data collection because see you're going to utilize what you're going to utilize ML and ml what happen on machine learning what happens machine learning uh make decisions do things based on the past data so you need huge huge amount of data based on which you're going to build the Emil model and that Emil model going to predict or give the solution do all this kind of stuff so you need to have huge data and train the model on that to get the results done then you have the model you have the data but uh it's not that easy I mean you need infrastructure to hold that much data process that much data and also build the models then deploy the models as a Form application and all that right so you need a proper infrastructure and tooling then there is the model of course you since you're developing model then you need strategies to deploy the model and of course do the continuous training then implementation I mean you have built all this right then you have to go the implementation and the integration of all these things and then uh the eth the eighth point I'll can go discuss in very detail about the Automation and optimization and Ninth one is the expansion and scanning why I'm coming to the ninth point also I'll just give you a brief because you don't go ahead and you know if you have for example you're running a old system you don't go and just suddenly replace the entire system things will go break and you will be not in a state to fix all the things okay so instead like deploying everything is a once you just go with the gradual implementation like slowly slowly make some changes see how it works get the feedback and make Improvement and then that's how you should be deploying or implementing aops in your current system and I'll come to all that okay don't worry so this is a very basic overview of the AI a road map and I'm now going to explain each and every all this uh points too you guys okay before that let me check if any questions otherwise I'll just quickly move forward no question okay and let me check the YouTube also see okay oh uh okay hey Ashish Guru hey Guru okay so you guys now coming live now let's go a bit more deeper okay so let's go deep so let's understand it operation I given you the incident things right so let's go a bit more into it so if you're going to work in a okay then you need to understand because a stand for the artificial intelligence for it operations right so you need to understand the it operations first what are the things need to be done there first understand the operation management which is a focus is the customer and uh continuous Improvement of course that's the focus of the even if you see the devops also the continuous Improvement was the big part of that right then comes to The Incident Management like use the monitoring tools to detect and log the incident promptly and of course based on that make a decision root cause it and fix it okay then problem management which is the root cause analysis then change management suppose some change happening what could be the potentialities what could be the effed resource and all that okay then um the asset and configuration management because uh as a part of it operations uh they have to or say you might also have to have all this kind of you know resources configuration all that that you have to maintain okay then release and deployment management I think this is one of the big part of the operations team to uh release do the deployment do the roll back if there is issue uh then you know gather the data right that's the manual part if you think about that okay and then need to know these are the it operations are there as a normal um it operations guy have to handle all that okay or oops guy has to do that okay so understanding of the basic it operation because if you don't know what to improve then how you'll improve it right these are the things I covered the basic point that you need to improve using the AI okay then of course since you're talking about the AI Ops you cannot miss the AI part so learn learn Basics uh at least the basics of AI because you you might think okay um for mlops I I particularly suggest learning the basic ml but if you know a bit of it you can still kind of um you know can go you can still go there okay but in particular to aiops without knowing a bit of AI or ml it will be extremely hard okay so understand the basic of AI for example um of course the understand AI implication machine learning Basics okay like use what is the supervised learning unvis reinforced learning all these things you you have to learn then the Deep learning Basics um uh probability degrees and AI okay then natural language processing neural networks all these things okay I'm just giving a video bit over you all this thing to learn I not say extreme deep way like a AI engineer or say as a ml T ml engineer but at least you should have a idea of what what they're doing and how can do that not the implementation part of the AI part but at least how to do that that you need to know okay then comes to the aops basic of course you learn about the basic uh ID of or say a operations then you learn the basic a now let's Le about the I AI Ops okay so in AI Ops basically U you have you have to do two parts two main components here one is the data collection which is a collecting the data like logs metric even data like we have very specialized task there okay which is collecting the system related data application log related data right and then use the a a ml algorithm to analyze this collected data to run uh predictive analysis on this to find the root cost and all those things you have to do okay and what are the benefits of using the aops it automate like if do if you have any kind of Auto like say repeated class you can automate them um and it can reduce the overhead of the IT guy who might be sleeping or maybe not available moment or maybe want to focus on other business logic uh like where you need to focus more this kind of stuff right and uh ProActive Management suppose some issu is going to happen okay if we can catch that issue if we can root cause quickly it's all our time saving it's all we can deliver more value to it right so that's another benefit to it right so if you're going to learn aops of course start with the basic it operations but you need to know if you don't know you need to know about that you need to learn about the AI and then learn okay using AI how you can analyze the uh log data and I will come to how you can do that and how can you do the analysis and uh do all kind of cretive analysis roen and all this kind of things okay so now we're clear on the third point now let's go to the fourth Point like ml aios is nothing without the data and what kind of data specifically talking about say logs data application could be application log be system logs then Matrix data like CPU usage data over time or second basis or maybe minutes or hour basis right all those kind of thing and all this data you have to store right and you have to then integrate and you have to process it okay this is that's why if if you see there's total three step I have mentioned main three step which is data inventory the actual data then the data integration which using the API using the say missing fix the missing data and all that then the processing which is to ultimately make that data mode usable okay because it can be text Data it can be uh you know some GSM data but it has to make it understand machine understandable data where you can run the training right so let's go with that data inventory and what kind of inventory you're talking about we're talking about the logs okay it could be application locks server logs Network logs it could be metrix data which is the CPU usage memory usage response time all the data events like uh for example system up time uh downtime or for example any alerts data okay and of course the main alert data which is the alert generated by the monitoring and alert system all this data you're storing somewhere you are going to utilize that for this kind of AIML a implementation okay and then while you have the inventory also you have to find out what are the Gap maybe you have there could be there are many cases in the past also where uh where we can we lose I mean we lost say 2 hour data 3 hour data 4 Hour data so if those cases how to handle that kind of issue and that will happen I mean you start with it you are very confident is might not happen but soon you'll see this kind of thing happening for various reasons okay could be the system not able to scale could be like the the system that track the data going to store the data the application um code that going to send the data to a permanent storage that itself have a problem so all those cases and and that's very normal to have uh because it's all system and system will fail at certain point of time we just if you can find out earlier you can fix quickly that should be our main goal okay and then comes the data integration which is the for example using different kind of tools say uh aggregation tool like elk stag spun or data legs utilize and aggregate the data from multiple sources then API and connectors which is implementing API and connector to ensure seamless data flow between the different system because you might be uh so you have a see aw you have multiple Services right you need uh same data Maybe different different Service uh places if you have a central way you can just do that it's much easier right then ensure the data quality which is make sure the data is consistent and accessible and uh then come to the data processing where of course you have to uniform like say you have maybe raw data you have some binary data you have to make that understandable data so so that uniform format you have to do then you have to do normalize the data and if you have a missing data you have to need some certain strategy how to kind of handle all those kind of thing then if you have irrelevant data not needed you have to filter those data out all those things you have to do as a part of the data processing so this is about the fourth point the data collection and integration okay then let's come to the fifth Point okay which is the infrastructure and tooling okay so in this particular Point okay uh people if you talk about say if you just ask anybody um be mlops and aops and if you talk them I mean ask them what is the infrastructure part in your implementation they'll say okay we have we running in the gcp as maybe other places or AWS right so they can say like that but it's not not like that it's not that you are you are using AWS or you're using as because you know that no you have to find out the optimal like what is the problem at hand and what is the best tool that can solve the problem not based on the tool you are giving the solution based on the solution you have to choose the tool that's more important okay so that's why tool selection is very important and for example I mean just give you example you have to do anomal detection you have to do predictive analysis we have to do root cause analysis we have to do automation you can develop in-house tools or if you don't have that bandwith if you don't have the skill set if you don't have the budget you can go for existing tool set for example there is uh moft is there dinat is there data dog is there app Dynamics is there Splunk is there big panda is there and all these tools are very good at doing certain things for example uh moft okay and that reduce the noise in event correction provides a realtime analysis animal detection MLB Insight collaboration like that D offers I mean I have personally used for my existing clients uh that and this offers a comprehensive Aid monitoring automated root Cod analysis end to end monitoring Cloud native support and the focusing of application performance data do I think most of people are already using it which provide the unified monitoring analytics with time doogs matrixes these are all the data storage and this have the capability of doing the analysis and root CIS okay just like that I mean you can if you if you see these are kind of tools a tools already available you can utilize or for example you can uh build yourself if you plan to build yourself then uh you need to make sure that you have the infrastructure set up for that for example what the data storage the processing power for that the analysis tools the skillability planning because if you're going to kind of build the models if you're going to uh kind of do all this kind of processing of the data you need a proper infrastructure of course the cloud integration because if you are planning uh to do like locally in certain point of kind of can do that but if you have huge data you need huge processing power then you need the cloud system with kind of infinite scalability and budget to do all this kind of stuff so that's why infrastructure and tooling is important based on the problem at hand you have to find a solution uh based on your um say uh requirement you should be utilizing tools based on the budget you might be using maybe existing tools that's already AV in the market or you might be building your own based on all your requirement you have to make that choice okay so that's about the infrastructure and tooling now model development so if you finally decide to that have your own infrastructure you're going to build own model and you do all this kind of stuffff then come to the model development and training which is first I mean there is three step okay first step is algorithm uh selection then is the model training and then AI modeling and testing okay so all these kind of things you have to do okay then um let's like say there are two kind of learning one is a supervised learning and the unsupervised learning so for supervised learning you need the level so you should have a say particular leveling of the data for answers learning you don't need the leveling and if you think about let's say what are the use case for example if you have you need anomal detection then using the say super learning which is say for example algorithm Superior Vector uh machine sbm neural networks you can utilize that okay for un learning you can use the K means clustering isolation for it principle all these kind of algorithms are there you need to not go and jump right away you go by the flow first learn the basics like all basics of the AI uh basic of the operation then you come to this kind of model development you need to look learn the basics and then all this okay so see this is the algorithm selection where you will be selecting the algorithm then after you select the algorithm then you have to choose like's say build the models and when you are training the models you need to make sure you have that enough data based on which you'll be training the model then you need to decide on how should be the training process where you're going to run all this kind of model training all this you have to make the selection okay and there are a after you build the model you have to make sure you're testing for example you are doing the F F1 score testing you are doing the accuracy testing you doing the precision testing or maybe R testing all this testing you have to perform on the model to make sure the model giving you the proper result it will be not 100% perfect okay but at least it should be bare minimum 60 to 70% final as a final product okay that you have to achieve then the AI modeling and testing which is the say you need to have a pipeline setup because every moment you not like once you're building a model and that's it no every time there is a code change happening for by the ml Engineers or if you're working with ML teams you're making changes any CH at all you need to rebuild the model so there has to be proper pipeline proper testing and all this kind of stuff okay that's why you need to have algorithm selection model training and AI modeling and testing of the like thoroughly test and have the Automation in the place to build the models okay so that's about the sixth St now seventh is the implementation and integration like you have your model ready if youan you have your infrastructure ready you have everything ready but are you going to like suddenly switch from your existing uh monitoring system or existing solution to this no it's about you have some system for example you have your infrastructure running if you have your application running all the things are running you make your this implementation as a new part of the existing system itself you're not going to replace you're going to integrate this new solution with the existing systems okay for example you had you have application log you have system logs if you have developed this kind of model okay or and have the infrastructure to process and utilize the model to do the predictive analysis or do the anomal detection or maybe a root analysis you have to run first on some small projects or Pilot project or maybe a part of the system and you have to observe is it working is it giving us the U say uh the uh desired results is it working at all that kind of decision you have to make make and if that works then gradually you make that part of your entire system slowly slowly 5% 10% of the system you first in like deploy the solution see it's working fine then you make a another kind of an advancement maybe more 20 30% of system you apply this new solution because running all this infra running all this model development it's costly costly scary and if it's not giving you the result you need to think about it might not it might be not the right way you have to make some changes okay so don't directly go and deploy or Implement for everything first you do pilot projects okay so pilot development then Monitor and gather feedback then see that it's working then this then deploy the tools to other workflows okay that's how you should proceed and then the eight is the Automation and optimization okay because see uh there I mean this is one of the there are many use cases I talk right so if you see any opportunity to automate any task or say VI this kind of ai2 I mean AI implementations you should be doing that okay so and specifically the common task you should be converting that first because that common task is something like you do it but sometime you forgot humans will do errors right so if you want something to be happen on that on time and repeatedly automation to be there you can Implement all the solutions right that will be very good then continuous optimization as I said the the solution initially will build the model will be developing it's not the final one it's have it will have many bugs it will have many issues and you have to have a continuous pipeline which will keep fixing keep kind of improving on all that so you need to have all this thing in place like it it involves updating the algorithm uh it involves maybe try out something new some new tooling and all those thing you have to make sure that you do okay now let's come to the uh next point which is the expansion scaling and this is the final point I think I'm going a bit fast and it's good because not many comments so it's fine uh okay then is the expansion and scaling what is about that see at the beginning you will deploy the solution of this AI Ops in a very small subset of your system or maybe 10% system I the beginning then you need to slowly Implement for all other system maybe you can offer this as a part of a solution offering to other companies also right then you need to scale up your system which handle all this break analysis part model training part and all this kind like to make things alive to make things running functioning you are spending a lot you are doing all this right and when the workload increases you need to expand it so in those cases you need the scale up okay and have uh the scale up in terms of say automatic provision of the new servers then automatically running the testing based on the code changes uh it it seems that it it need U additional uh tooling or maybe U there is some security problem happening that automatically detect and report all these things need to be there okay so scale up need in the means that you need to make sure when the requirement G requirement become bigger you need to have a system that can handle that okay so that kind of scaling you should have and of course it means there are many tools already ail Auto scaling there are orchestration tools available cuties for example which can help you scale even faster in more efficient way so you have to utilize all those things okay and uh of course do continues optimization um and you have to make sure it's a culture okay it's not just one person handling everything when you are having an AOS in your company your team everybody has to become a part of that everybody has to understand not very deep but at least have a basic understanding of it and share the knowledge with each and everyone so that they can become part of all this okay otherwise you'll see more issue is than a solution itself so that's that's all about the road map okay uh in the main PDF I have given more details way uh and I didn't want it to take too much time of a Sunday of yours so I will share this particular um AI road map with you guys all in maybe in just two three minutes okay and then yeah then go ahead and if you I mean I don't see any questions so if you have any question after that I will definitely uh maybe reply to that or answer to that as a part of the reply or maybe as a separ video okay so thank you guys for staying half an hour and uh like going through all this map so see you soon in my next video and yeah I'm going to share this uh PDF with you guys okay just now okay bye