LIVE _ Evolution and Future of Cloud & Edge Computing

sir we are live now please start okay thank you biology hello everybody good evening uh first of all i wish all of you and your families are doing well and uh healthy and safe i i really urge all of you to stay inside and stay isolated and remain safe let me quickly introduce myself uh i'm push gupta i am engineering graduate and certified management in and currently i am employed with ibm system development lab uh acting as a engineering program manager uh you know program director and uh i i have been with the you know it infrastructure industry for almost two decades and mostly mostly my focus area has been storage systems development so in my current role uh i'm the in charge of one of ibm's flagship storage product development group in india i'm a social uh you know social evangelist uh you'll find me very actively particip technical social campaigns so uh after this session if you like to connect me over linkedin or twitter feel free to do that we can learn from each other each other and share some good post on techno technology front uh as a part of this you know while the session is going on you still have an opportunity to post your questions using the link given in the description there is a google form given there you can publish your questions there uh because in first 45 minutes i'm going to cover my my topic and uh i'll try to use last 15 minutes to address some of the questions as much as i can so today i'm going to cover uh today i'm going to cover first about the changes in the technology and you know infrastructure specifically happening uh you know to address some of the today's uh changes in the workload so we look at we'll look at a little bit about it i will provide you some guidance i'll provide you my perspective but but as this is just a 40 minute stock so it's going to be very very limited to to a high level you know a 10 000 feet view that what is actually happening in this in this whole you know area and you know how the landscape is a little bit changing and where are we heading to so so that that's what we'll we'll talk in next uh 45 minutes as it's going to be a monologue so i won't be able to get any feedback during the session but please please please provide your questions in the google form and i will i will read them out and address them as much as i can okay so with that let me just quickly share the presentation i'm gonna uh use to drive my discussion okay for a very long time but the way they are being used in today's architecture and solution is something we need to look at is something we should we should be keeping an eye on that what is happening in this space so what is the objective of this session i briefly touched said that you know a minute back but let me read it out again the idea of this session or objective of this session is to provide a high level view on the need of continuous evolution of computing architectures due to the constant change in application workloads and our special focus is going to be edge and hybrid cloud computing of course because that's our topic but uh but that's that's a very important uh point you also need to know that in in it in the it infrastructure space things are changing very rapidly because there are a lot of changes happening the way uh it is being consumed when i say the way it is being consumed means the workloads the applications the consumer base there's there's a lot of change happening in that at that level so unless we change the infrastructure we won't be able to serve the requirements so we need to really keep evolving the way we have been setting up our infrastructures and solutions to meet the requirement of applications and when i say applications ultimately the users how the users want to consume this this infrastructure and id so that's that's the objective let's see how much we can achieve it but yes just see this if you can take away some of some of this in next 45 minutes so let's step back and see a little bit of past because that context is very much needed to kind of paint a story that why we are talking about hybrid multi-cloud today why are we talking about edge computing so much so for to understand that you need to little bit and you know kind of go back and see what happened few years back so cloud computing is not not something which is very new it's been in play for uh for many years it is a very very successful model of computing we have you know seen a lot of benefits out of that but how did it no what actually made it made it happen so if you look at when cloud computing was speaking up and why people started moving to cloud at that time the focus was to come out of some of the traditional data center limitations and challenges we had so at that time in summary i'll read these four bullets but the summary was that it was because it was mainly driven because of the cost and operations the to come out of the cost and operations challenges the cloud computing became a very very important strategy for all the it infrastructure you know decision makers so at that time when there was no cloud computing when we were into silos we were we were doing the traditional data center based computing at that time data centers were very underutilized the utilization was a big big challenge because the computing was done in silos every application had its own infrastructure had its own running environment so there was no there was no guarantee whether you are leveraging or utilizing your infrastructure to the max capacity and and that's a big deal if you're not utilizing your infrastructure then you're losing out on a lot of compute power which your device can use it or probably you're wasting that that compute so that was a big challenge that how do we make sure that we are fully utilized so that was a problem statement then traditional data centers also had a very overall very high cost of ownership because it was a very high capital intensive you need a huge real estate you need a lot of energy you know to run your machines you need to procure a lot of capital you know a lot of you know assets using your capital so it was a very very high cost of ownership kind of uh affair uh which was which was not easy for everybody to get in right because not everybody had that much of dollar power to go and invest in this this much of you know kind of infrastructure it there was another problem with the high entry and exit cost in traditional infrastructure and the traditional data centers it was very high cost to get in and also very high cost to get out because once you get in you invest in capital and when you want to get out you lose a lot in terms of capital so it was a very difficult entry and exit problem which you know uh the traditional way of id infrastructure was was kind of posing to decision makers and then it was definitely a maintenance nightmare because you know in in that era you needed very specialized skill to maintain these infrastructures and there was a very large lack of automation because one silo was not talking to the other silos and every cylo had its own skills to maintain those silos so it was a it was a big big challenge in terms of maintenance so these were the key you know i would say there were many other reasons too but these were the i can say that were the key drivers to think what can we do to come out with these problems and the solution was cloud computing at that time so cloud computing again this is a picture coming out of the of google it's a public it's a standard picture we generally summarize that what cloud computing offers uh to come out to solve some of those problems so it is actually nothing but leveraging some of the virtualization technologies and a utility-based deployment models to come out with a better computing model to which actually allows you to you know start utilizing your resources and sharing your resources now the silos were broken we broke those silos and you know kind of pulled all those resources together applied a layer of virtualization and involved the whole new model of computing it was a big big boom you know it was a big boost to the technology world it was a big cost savior and it was a very very efficient right so it was a game changer of that time wherein we just you know kind of you know found a way to to reduce the cost of ownership to reduce the overall cost of operations to reduce the you know kind of entry and exit were no kind of a challenge which small organizations were facing because they didn't have too much of dollars to invest in it was a game changer no doubt about that right and and even today if you look at all the amazons and google cloud and you know and azure and you know all these are solving big big problems of past which you know uh which which otherwise was a pain now this is the example now again i gave some examples of public cloud because all the amazons azures google cloud and ali cloud they're all public public clouds because there's somebody else who is setting up a cloud for you a kind of a kind of computing environment for you which you use pay as you go right but cloud computing is not limited to just public cloud uh it is also the private cloud is another very very important concept that the customers the enterprises who are not shifting their workloads to somebody else in data centers or cloud data centers they can still leverage the the advantage of cloud computing technology by deploying the same computing model in-house so that's another very successful model because companies who had big environments they could convert their traditional environments into a private cloud environment which ultimately achieved the similar benefits of over you know appropriate utilization a lot of sharing of resource resources you know you know scaling up scaling down so all these advantages could be achieved by even deploying the same concept on simply similar technologies in-house by big enterprises so ultimately cloud computing is a very successful model and which is still being worked upon right so we are you know there are enough data to suggest that almost every enterprise customer has a foot into cloud right now when i say cloud uh we now in this session we are not going to differentiate cloud between public and private we are talking about cloud as cloud cloud is a concept wherein we are leveraging best of our environment best of our infrastructure and you know have the ability to scale up scale down we are more agile more automated right so so that's what we are referring to so ultimately that solved the problem now so the tagline say cloud computing offered great value over traditional data centers and that's what we have been you know you know reaping from last many years if you look at how things are changing today what's wrong uh there's nothing wrong with cloud there's nothing wrong with public cloud pipeline cloud but their requirements are going beyond which cloud computing or probably current model of public cloud or private cloud can resolve so the problem statement of today are we believe that people want to make decisions better and faster that's a very very very you know broad statement we want to make better decisions when i say better decisions means we want to you know leverage more power you know we want to churn more data analytics to make better decisions the data i think we will we will see that in subsequent charts that how the data is increasing the data every day is multi increasing multi-fold data is of no value until you know it is converted into a into information because data is just bits and bytes this has absolutely no value until you find a way to convert that data into information and that's what we talk about how you apply right amount of tooling applications analytics statistics on top of the data to derive a better decision now the other uh you know the word in this line this sentence is the faster decision it's just not about making right decision now we are in a very very fast fast world things are moving very fast so we don't have enough time to make right decisions we are we need to be very very quick we need to be fast so it's it's about right decisions and also fast decisions so that's an important one of the important requirement and we will see in in the in the future slides that why our current infrastructure deployment models can't can't meet this requirement completely well they are meeting the requirements we are still doing all these things using our current deployments but there are there are you know loopholes there are gaps there are you know scope of improvement now second line says that we believe that you shouldn't move data to cloud that you will never use now now think about that why should you keep pumping data to a cloud now assuming that everything we generate go and get stored somewhere and in this in this context we are talking about we generate a lot of data and put that and save it in cloud but we do we really think that we should be spending keeping every each and every data we are generating to the cloud answer is no because data is being generated in exabytons zeta bytes but not all that data is useful there's we can we can all and we need to really start thinking that how are we going to use this data and how are we going to filter this data so that we don't waste our resources to store that unusable data right or you know use useless data and apply any application or analytics because everything has a cost associated with it so we don't want to worry about the data which we think is not going to be useful so why should we do that so that's where no it's not necessary that everything which gets generated in the form of data should be stored and pumped somewhere which is cloud in this context that's not required now then then what we do i think that's what we will talk about how the edge computing and this hype you know is is coming in played to solve some of these problems now the third statement say we believe that your personal and private information should be protected very important point and you would see that it's a it's a talk of the town everybody's talking about data privacy data safety security it's a big topic it's a huge topic and it's a you know it's a topic which is you know it i discuss at political level at technology level everywhere so data personal information and personal data is very very important to be safeguarded now it's not necessary that you know that you place your personal data at all the all the places where it is not protected so we need to protect that data and how do we do that then the last statement say that we believe that you should be able to continue operate your business even when the networks are down now if you look at we are highly dependent on on on the data the data analytics and making decisions on data think about autonomous car if autonomous car is making decisions to turn right or left or apply a break based on certain decision making based on real-time data think about if there is a network outage and the car stop connect communicating with the with the engine uh the data engine i mean who is making decisions for the car what would happen it will be a disaster the car will go and crash somewhere and it will be a disaster because when when i'm driving my car there's a network between me and my steering and my clutch and brake is that's the only network we have but now we are talking about connecting these devices you know and uh and making decisions over the network so if there's an outage an outage is possible if there's a disaster and it keeps happening yesterday there was a cyclone you know we were we were into cyclone when gujarat was impacted maharashtra was impacted if god forbid if you are running autonomous cars in this country and you are you know making decisions through some some central decision making or producing data processing unit and you lose the network with those machines what will happen either we just shut down those vehicles and just park them aside that is still better but still you are immobilizing them or you there is some kind of mishap happen so ultimately we are trying to say here now the way we are using id it is very important to continue operation even when when the network outages are there because we are highly highly dependent on on on the infrastructure and all the operations are highly highly dependent on some of these applications right so we can't afford to go offline we can't afford to have down times think about an era of last five years how many times you have seen down times on any on any of your banking application or or any reservation application or whatever hardly any because we cannot afford to have down time because it has a real you know financial impact and sometime it's a it has impact on lights because even medical science is using so much of technology which is you know which is based on all this you know communicating between devices because sometimes the surgeries are happening through robotics and if the robot is being is communicating with some other device because that person uh if it has been happening on the network and there's a network outage what what will happen so we need to start thinking about it while it looks all all great that we have such great applications of technology where we are leveraging more data more analytics more artificial intelligence but these are some of the infrastructure challenges we need to deal with so that we have a continuity of service we have a we have continuity of operation and then there's no problem of adoption of the use of those use cases so these are some of the high level you know key drivers which i'll try to you know stitch it to the story of edge computing and how hybrid cloud is is you know kind of becoming prevalent so hope you understood the main crux of the of the slide that these are few challenges which are becoming more and more relevant as we moving forward because the way we are consuming things are very different the way we are consuming id five years ago or ten years ago or fifteen years ago and it is constantly evolving and changing okay so what should we do so when i say these are the sorry when i say these are the four problems we have these are just four problems there are many other problems so what should we do place enterprise application closer to where the data is created and where action need to be taken this is the one line statement which actually define edge computing so what we are trying to say the data is generated from certain somewhere right if i am you know you know driving the autonomous car i have a i have so many sensors in my car it is actually generating data in the car i'm generating lot of data when i'm driving autonomous car now that's the place where we generate the data now what we are trying to say don't take that data because ultimately we are making a lot of decisions based on the data we are producing there right so a lot of decision has to be made based on that data and applying lot of other algorithms on top of that so we are trying to say here that move you know the enterprise application closer to them now you move your you know algorithms closer to where the data is being generated now what what will this you know kind of give you in in well we will look at it in all the slides but ultimately it will allow you to you know kind of mitigate the problem of network outages because you are not going too far with your data and doing the data mining and doing all the you know churning of the data and bring the decision back you are traveling very less distance doing that and bringing back the decision so you are actually mitigating the risk of losing any connection by the time i get the decision back so that is one thing and the other thing is it has also giving you the decision faster because the far you go you take more time the shorter distance you travel and get back your decision to then because ultimately who is looking for the decision the end device if the car is generating data about who is driving in front who is driving the right on the left is there a pothole is there a you know kind of guiding line or whatever i'm generating but ultimately the right left decision also is the car has to take so ultimately the decision has to be passed on back to that device who actually generated the data to make that decision so ultimately decision has to come back very timely otherwise you know otherwise your use cases you may not be able to serve those use cases and you take more time think about it if if my decision making takes probably a few minutes a few seconds or you know few more than 10 seconds or 20 seconds i think i can't apply this technology to autonomous cars because they here i need to take decision on fraction of seconds because a wrong decision in after you know fraction of second may you know cause a cause an accident so it has to be as real time as possible and that's where the speed matters so edge computing is nothing but what we are in a crux in the summary what we are trying to say bring the algorithm the application the decision maker closer to where the data is being generated so that you can make decisions faster and also you don't have a risk of you're losing that connection and you need to retransmit that data to make your decision happen right so this is uh this is probably the very short answer to some of the problems we are facing and again this is we are not talking about all this in complete isolation all this whatever we are talking about it is all supplementing to the great technology enhancement we have already done in past like for example when i say cloud computing today we are doing cloud computing whether private or public or whatever now whatever we are talking here is the next step we are actually supplementing that innovation with some newer innovation we are making it more you know more strong and more robust and more capable to handle our current workload requirements right so it's not that we are getting away or probably you know we are ruling out one thing and bringing something you know we are implementing they're bringing more feathers you know and fills and feathers to what we already have in technology and making it more robust and more applicable to the current use case now this is a very simple term you know graphical representation that we have seen an era of mainframes which you know many many years back then we looked at client server architecture which was a leaner and you know architecture then internet era web application error came in mobile computing came in public cloud is something which is very very prevalent today and we are now looking at building you know edge computing on top of that so i'm not saying that we are getting over from public cloud and bringing it's basically a combination of cloud and edge which is actually the demand of today we need to work together collaborate these two technologies and actually evolve the solutions which are which are useful today so if you read the line decentralize the decentralized data and application processing across hundreds to millions of endpoints residing outside of traditional data center or public cloud is edge computing you are basically decentralizing we are not centralizing the computing we are decentralizing the computing well if you look at you know uh when we talk about cloud computing that kind of was was kind of we were bringing in multiple silos to a centralized computing space right and but but cloud was more about sharing and utilization while we were bringing the infrastructure together but i think centralizing the centralization grid computing you can think about was more of centralizing the compute together and doing all the applications but here we are talking about let's just decentralize it and move back to the edges move back where the data is being being generated and and do that the actual analytic actual application of that data on those edges rather than bring hold the data at one center place do that algorithm do that churn and make decisions and then pass that back it's a very expensive operation it's a very expensive operation and it's not a scalable operation because the today's the biggest challenge of today is the scale right you just look at your own just think about uh your own laptop or your own hard disk a couple of years back how much data were you you had about for your personal requirement you just had few gbs now people are just you know not even you know able to satisfy their requirements with multiple tvs so ultimately we are scaling very fast now how do we deal with the scale is something very very important so with the current centralized approach definitely it's not scalable enough the kind of explosion of you know of data and this all end devices on the application we are kind of witnessing so we need to think about decentralizing it and passing the responsibility to you know outside the data centers and do certain operations in data centers but allow certain operations through the endpoints or probably outside the data center so that data centers doesn't become the overloaded and you know probably you know single point of failure or probably a slow decision making entity so that's what we are trying to achieve out of edge computing okay so we just quickly looked at but the time okay we are already 26 minutes i'll quickly just move on because it's a big topic so i i can just keep talking on and on on this but uh i'll just try to give you just of the whole topic so that you can go back and do more reading and more you know discovery investigation on this so but just just to understand what is actually edge we have been saying edge so what is edge so in this picture you very clearly understand what are we referring to as edge and what is non-edge so we on a high level you look at that two there are two main main main entities one is the edge one is non-edge right and if you look at anything on the you know on the left hand side these three pictures are the we can consider them at edge and all the cloud you know kind of infrastructure is your non-edge which is basically an existing infrastructure sitting somewhere far away and where you know we are pumping all the data today so what we are saying at the edge is your all the end devices like in autonomous car example which i'm saying again and again is my edge device because that's where i'm generating and you know creating the data right and so that's my edge device cameras there are a lot of cctv cameras they are actually capturing the images they are the edge devices machines the robotic machines they are they are these devices who are actually performing jobs on the on on either automobile or any other shop float so these are the edge devices now so edge devices are those who are actually doing the job and also generating the data and then edge clusters is the next level wherein it's a cluster of edge devices you can say and where you have it's the next step is the next step wherein you you actually store and process your data that's the nearest destination where you bring in that data and and you know kind of uh you know process that so take an example in a bank you have hundred 100 cctv cameras in a bank now those cameras are the end devices but they all these cameras are actually wired to a server room in the bank where all these footages are being captured and being stored so that becomes your kind of edge cluster wherein hundreds of these cameras are being you know kind of uh you know plugged in into one entity or one server which is capturing that so that becomes kind of edge cluster and then now hundreds of these branches are being clustered to probably a regional branch now that becomes your metro edge metro is probably one step above before you get into the cloud right so it's basically breaking this whole edge into three small components because components just to kind of explain the concept and it varies from application to application and use case to use case but ultimately we are saying that before we land our data into cloud we have other revenues to to kind of store our data and don't put everything on cloud because that's the most expensive path to travel if i need to probably think about it if my all these ccc cameras are directly feeding into my google cloud or probably on my public public cloud or private cloud then i don't have anything in the middle i'm directly feeding the data i'm traveling i'm capturing that image of multiple gbs or whatever that data is actually thrown you know probably traveling on the pipe on the on the network going to the to the cloud and traveling this much of data is very expensive it's not cheap every every you know kind of gb or you know every bit of data traveling you are paying to some service provider who is giving you this this buyer so you are first of all spending a lot of money to travel that data to the cloud then you are actually spending a lot of resources in cloud to store that data and and you know kind of process that data and also you are adding more to the slowness of the whole process because you are sending the data far away processing and bringing it back again the same distance so you are actually slowing down the whole you know the workflow of the data of this data for decision making so that's what we are trying to get away with and that's where these intermediate this edge devices and intermediate edge gateways and clusters play a very important role in this whole concept of edge computing so in in this picture if i again apply the concept which i said what we are trying to say that edge device become more intelligent like camera today it is capturing data but can can it you know be more intelligent to probably you know filter data to first level maybe yes i would love to have a camera who actually don't throw me a picture which is not clear the camera should be able to make a decision that okay if this picture is not clear at all will not make any sense for anybody let's let me not feed it back to my cluster or the cloud because it's waste of money because they are going to delete it later because they won't be able to make any sense out of this picture so can camera make that decision maybe yes maybe that's the next smart edge device who is actually not filtering at its own level maybe it's a small decision making that okay i'm making a decision not to you know discard this image while sending all the other images up so that the intelligence we can bring in right so so that's the that's kind of a very quick definition of what we are calling as edge and what we are calling as non-edge now this is some statistics if you look at today 90 of workloads are deployed to public or private cloud there is traditional idea all the traditional i.t data centers have evolved themselves into a cloud data centers right by using more and more advanced cloud technologies which allow them to do more sharing of resources more agility in terms of scaling up the capacity is killing down the capacity it allows them to you know kind of transfer applications and move applications seamlessly it also allow them to do lot of things automatically then every manual operations so so 19 more than in fact more than 90 enterprises have already deployed cloud computing in some or the other form and just to tell you now today most of the enterprises are looking at a hybrid approach there is hardly any organization who is either fully public i'm talking about enterprises because the small agencies and small organizations who don't have their own data centers they have no choice than just going public cloud because they they don't want to invest in their own infrastructure so for them public cloud is the best bet but the big enterprise customers who have their own data centers they are also looking at hybrid approach that for certain applications they are leveraging their public private cloud environment certain applications they are using public cloud because today we have evolved in terms of you know kind of integrating public cloud with private cloud today amazon aws azure or google cloud all these cloud providers provide fantastic integration mechanism to integrate their public cloud with my private cloud so that i can move things between these two clouds as in when i want and that's what the concept of hybrid cloud is that's a true hybrid cloud wherein i'm not logged into one thing and i can always you know shift my workload between the cloud and i can also make best use of whichever give me better value because certain applications are better you run and run them in house because that's more cost certain application does make doesn't make any sense running in cloud it's very cheap if you run them in public cloud so that's where you know this hybrid cloud approach plays a big role so 90 plus organizations are already doing it are already doing some other form of cloud deployment now by 2022 this is what gartner is saying by 2022 50 of enterprise data will be created and processed at edge well and we don't have this implementation of probably cons you know this concept of edge solution then everything which gets generated generation is still happening at edge everything gets generated at edge but today there are very few uh solutions wherein they are being processed at edge they are being processed and in in cloud or whether it's public or private but what we are saying here is that by 2022 or 2023 more than 50 percent of enterprise data would not only be generated at edge but also will be processed at edge so look at it it's like you're saving a lot of you know you know kind of investment of your cloud and doing that at the edge but yes it means you need to invest in edge processing that's a it's not that we are going to save a lot of money there but ultimately it will make your user experience very different it will make your uh the scalability requirement you know where it will meet those requirements because you can scale by adding more and more edge computing platform more and more edge level processing platforms right so that's what we are witnessing and this is what gartner is predicting that we are heading towards this kind of a you know kind of composition in 2022 and future so i have not gone through because we are running out of time and you know we just i'm just again checking the time we are already 30 minutes over and i have 10 minutes to wrap up what i want to tell you so uh again this is nothing but kind of explaining this picture if you look at everything till metro edge we are talking about putting all that into edge bucket and anything outside that we are calling it as a data center which is in most cases is cloud now we don't talk about traditional data centers anymore so it's public cloud or private cloud and everything else outside public cloud or in-house in-house data center is actually the edge processing or probably edge you know kind of uh edge device or edge platform right so that's that's what uh it this is also referring to now market trend driving edge computing the there are many reasons i think we kind of covered that briefly that digital business transformation promote decentralization of data because uh every business transformation today is looking at doing things faster doing things more efficiently and you know also bring down the the point of failures right so that's the that's the whole goal of business transformation today how do we bring down the downtime how do we make decisions faster how do we reduce cost as well right so date and and most importantly how do we secure i think i didn't touch base on that when you travel your data for a long distance you are becoming more and more vulnerable if you are not putting your taking your data to a long distance you are also kind of helping reduce your vulnerability you are actually giving less you know tap points for hackers to come in and snooze your sniff your data so i think that's another very important point edge computing is kind of you know handling that you're not exposing your data all your data definitely some data has to travel and go and store in the cloud but you're not exposing all your data by doing localized you know data creation and processing explosion of smart enterprise end points create massive amount of data so scale this is all about scale this amount the amount of data being generated is huge right it is hard to imagine that ub would always be able to pump this much of data to a central location or a central infrastructure like cloud computing and bring that back to the edge because ultimate data has to come back so it is it is a very difficult problem to solve in the current way of computing enterprise generated data processing mic migrating to edge is the same statement now the important one is a 5g accelerate edge use case and 5g depends on edge computing i think 5g is is not something in india is still there but but 5g is not too far so 5g networking is is something which is already deployed in a lot of countries and 5g is bringing a lot of promises to the applications in the space of ai analytics because it's a fast it's a very fast way of you know kind of communicating it's a multi times faster than 4g today we are using 4g today but it's a multi-fold faster communication standard which is coming in but 5g is a high frequency technology which actually cannot go long distance so it actually rely on edge computing because it expect the data to travel a lesser distance to give the value of speed to the end user so ultimately its computing is getting a lot of boost because of the 5g you know standards which are being evolving and very soon it will be a reality right in in in india as well in some countries it's already in production but applications are being built with 5g you know factoring in and if we need to do that we need to help 5g by giving data you know at a you know data to process at a lower distance than taking it very far so i think that's where this 5g is also pushing a lot to kind of deploy edge computing as much as possible h computing enables you to continuous operation as i said it's a continuous operation your your reliance on a lot of networking hawks reduces because you're not traveling multiple networking hops if there is a outage at an upper layer of networking hop your lower level of network is still still intact your operations continue so that's a big advantage then faster insight to an action is the speed this is the i'm saying that by doing edge computing you are actually enhancing the speed you're reducing the latencies so latency is how fast you can get a response how fast you can send our data and that you know receive a response to that that's latency so we are reducing latencies by adopting executing better data control and cost you're you're in better control because nowadays there's a lot of compliance problem also we are dealing with that this country say that my data should not go out of this my geography and this data can go out this data cannot go out etc etc so we have we have a better control in because we now have a flexibility and we have that control that what we want to send across the border what we want to send across the states and what what data we want to just keep local right for the security and privacy and on regulations purpose right so so that is another big advantage which uh which edge computing providers 27 companies are already using or aligned are aligning their business to get advantage from edge computing so this is already being done because you know it's not something the concept is not old we are just trying to leverage it from current current use case uh in next three years it will go to 52 percent so edge computing is an area again i put this chart here because all you listeners see how much opportunity is there in this space when i say opportunity is an opportunity to invest between you know to innovate in this space opportunity to do work in this space because if some some space is growing it will create lot of opportunities because somebody has to develop somebody has to deploy somebody has to solution so as a technical you know as a technical you know kind of you know skill of brains you all need to look out and watch this space because this is going to open up a lot of opportunities in in all aspects now this is a very simple example so that you understand that see if you look at a picture this is a shop floor of automobile manufacturer wherein there's a welding robot who is doing the building and if you look at here there's a camera deployed who's monitoring that what kind of welding is being done now now there is a server here edge server which is in the on the on the factory on the factory floor which actually look at uh which actually look at what kind of weld it was whether it was a good weld or a bad build now if we don't make use of this edge server then this camera has to feed all this footage to the cloud computing or whichever central location you are processing it and then that central processing unit has to take a decision whether this is a good build or a bad weld and by the time you're you're because it's a it's a it's an assembly line right the the floor is moving right so it's a very fast you need to make decisions very fast so in this picture the camera find a find a kind of a belt which is not good which has to be rejected and need to be redone so at this level itself they're making that decision and passing that reject decision to the to the assembly line and while the decision is made here it's very close to the shop floor we are not going to the cloud computing to make the decision we are still going to the cloud computing to pass on the overall report but decision is being made here so that action can be taken quicker and a job can be redone so this is a very very leave an example of what we are trying to say when we say edge computing now you can apply this to you know n number of use cases n number of industry examples now i think i'll just quickly skip a little bit and and i'll just let me just check if i have any questions coming up i think there are a lot of questions coming up i don't know if i'll be able to read all of them and take care of them but let me finish my slides and then i'll start reading some of them and address that so ultimately edge computing enable new business opportunities in every industry in every industry edge computing it has an application so it gives you a better connected experience it gives you you know it gives you distributed it modernization it gives you a better supply chain and asset management so i think i i've now i'll leave it to you guys to go and read more about it but i'll cover a couple of other interesting examples to give you the gist and you know the whole picture the context of what we are talking here now if you look at this is again many industry verticals wherein you can apply this concept of cloud compute computing now think about the first one the financial there are many applications of this i'll take one example of this against ctcctv today what is happening is every bank has cctv cameras in their branch when you go to atm you have a cct camera there you do your transactions they are monitoring what you're doing in most cases what they are doing is they are just capturing and that just feeding that footage to a central server and if i'm i'm a fraud i'm i'm doing a fraud and playing with the machine right so camera is capturing it but actually passing that back to the server and probably a day later or whatever their turnaround time they will you know kind of investigate and start working on that fraud based on that footage after the fraud is done right and by the time they apply any protective measure maybe some some you know kind of uh damage is already done like like i've i have you know kind of placed a camera or probably a sensor to read people's card number and cvvs so by the time the bank take action on that fraud which was caught in the camera probably some of the card you know details were already you know already lost i think they were already locked so can you know if there is an edge computing one application is deployed wherein now in that that branch itself you apply a kind of algorithm wherein if you find any fraudulent kind of action happening on the atm like this you immediately take action and disable some kind of transaction going forward just you know just stop taking cards from there on we will investigate the flood later but take some action so that nobody is allowed to put or insert the card right so that that you know that device make more damn they do more damage so it's basically making decisions on the fly and taking certain actions based on the decision so that's a that's a very good example wherein you cannot wait to look at this footage and make some decisions and then apply some you know some kind of action so that's that's too late in some cases it's too late so we i think that's a that's a very classic example which banks are already doing that then and there itself if camera catch this kind of a you know a legitimate activity take this action and shut it down and just stop any further damage that's one example now the other very interesting example which which is there published in many uh many online you know kind of portals uh in vehicle like this autumn autonomous cars are becoming very very it's a big buzz right everybody is looking looking you know to buy a tesla in india and how can we use more and more autonomous features in that car so yeah there's a lot of enhance you know enhancement and advancement happening there so there's an application we can look at wherein if i'm driving a car and i feel feeling sleepy and there's enough face recognition technologies available in the world today when the car detects that i'm feeling sleepy and immediately the car on the fly make a decision that my gps has been advised that this driver is feeling sleepy give him a route which is having a starbucks on the way right route him through a different different path where there's a starbucks on the on the way and also place an order for this driver uh the favorite coffee because now you know that this this driver like starbucks and this driver like this coffee in starbucks so before the driver stopped by the starbucks the coffee is already ordered and already ready to be served and the payment is already being made so you're not slowing down his drive but you added a complete new experience so all that can only be done if you make decisions on in real time and how can we do can we do that we can do that if we bring the data processing as close to the data you know where the data is being generated as close to the edge devices because car mirror is actually guessing whether i'm sleepy or not sleeping that's what you know that's the data being being gathered and if the decision can be made closer to that device then probably we can make more real-time decisions and bring some great user experiences look at such a great fantastic experience i was feeling sleepy my gps took me to my starbucks and my coffee is already ordered and you know money is already paid and i am fresh and ready for the ride so so that's these are very interesting and these are these cases these are just few i'm talking about we're running out of out of time here but if you go on the web all these you will find many even more interesting ideas but all these ideas can only be applied and implemented if you have the backing of a strong infrastructure solution because you cannot just think about an application and it will run on its own it will run on certain platform and until the platform support that application because this application is nothing but a software unless you can write good logic a good algorithm but if the infrastructure doesn't allow you to run that algorithm fast enough then you won't be able to provide that outcome very quick so edge computing is kind of so you know kind of supplementing that capability of fast decision making by bringing the processing closer to to the edges right some of you already may know some examples of gpus today because gpus are becoming so so popular because of this a lot of image processing uh every every application has some other part of image processing and gpus has a big role to play in it because it has a dedicated you know processor and you know kind of a core to do that so it's ultimately you that's another example of bringing processing close to the data right you're not going to the processor of the system rather you're bringing processor close to the to where the data is being generated so ultimately in a form it is some form of of the concept of bringing processing close to where the data is being generated so so this is a very very crude form of explaining edge computing there's much more to that to do that because it's just not uh that it's very easy to deploy it has big big challenges i didn't get time to go to get to those get to those slides but to do all this there are huge challenges that how are we going to deploy how are we going to manage this it's a huge management challenge and and when we are talking about taking the the processing to the edge we are again not creating the silos we are not saying that you go back to the old model of every branch office having its own machine where they don't talk to anybody and it's a we are not going back to that era we are not talking about going back to the silos we are saying that taking the processing back to those edges but still be the part of the same you know same kind of work space i would call it so that every edge in the world connected to one solution should know what is happening on the other edge right so it's all connected but still allow you to do more decentralized processing and allow you some other benefits of decentralization but still you are overall part of the same name space and same workspace so that we are still have the complete control of the environment and not a siloed one again as we've seen in the past okay so these are some you know some statistics of like uh how many you know devices 21.5 billion connected devices by 2025 this is just a guesstimate no wonder and no surprises that we have more than 21.5 billion devices right so the devices are growing and your scale is growing so that that is going to create a bigger problem data generated i come from the domain of data storage we we are dealing with so much of data and as a part of storage system development organization we are always challenged we are always being given a challenge to do store more and more data in a lower smaller and smaller space the form factors are reducing the requirement to store data is going up so you know we are we are given that challenge to take less space but store more and and just not store about how fast you can read and write how fast you can because now you're now you need to provide the data back to the to the application very quick so it's a it's a very very you know kind of a challenging area because the problems are just getting to the next level and we need to keep working on those problem statements to provide solution uh this we kind of covered yeah so just new challenges it's edge computing is throwing complete new set of challenges four main challenges scale scale you know number one scale definitely so it's gonna put you a huge scale problem how you deal with the scale data device variants now the edge devices are thousands and millions no two devices are same everybody is using a different kind of edge device so how do you deal with so much heterogeneity right how do you you know how you deal with so much of variety so we need some way of you know bringing a common layer to you know deal with so much update their device variants and manage data data is unstructured and managed how you bring them together you know label them and how you play with that and actionable data how do you filter the data ultimately uh as i said not everything is useful there's a lot of junk which is being created also so if we it's sometimes you know it's not 80 20 but it's a kind of you know kind of a say that sometimes it's 80 20. it's always we spend you need to spend 80 percent of time for 20 of priorities right so same with the data there's always very small set of data which need more attention than 100 so you need to have a way to find out what is that important data because it's not very easy to find so we need good way to filter and find that actionable data that what is my most important data and what data i can probably just you know just leave it aside i think scale this is a picture which tells you from cloud computing we are talking about two centers now here we are talking about thousands because the end points are many so your edge edge centers are many so scale is going to be a big problem i think this i'll just skip so that i can take some questions i think i'll skip this probably you'll read it so ultimately this is a picture which is a little bit item-centric so ibm is heavily heavily invested in this whole space and ibm is already made a lot of progress so some you already know that ibm is red hat is an ibm company and red hat has built great products right in the space of cloud computing which actually uh you know solve the problem of a common layer between the cloud and edge so that that's what they call it you know on you know ocp openshift you know container platform so openshift is a cloud manual orchestration platform from red hat so again it's a very high levels you know kind of summary but it's it does it is a great product which allows us to create a layer right on top of my overall solution which include my edge edge clusters my you know metro edges and public cloud and create a common layer on top of it because ultimately i should be able to see my complete environment at some level because it's all belong to me while something is in edge something is not at edge but i should be able to you know see all that together and should be able to action them together at certain layer so it's great great technology and very complex thing wherein some of the cool techies are really doing fantastic anonymously phenomenal job to solve this problem right and which we we as the end user see when we drive a ton of scar or when we you know kind of go to a shop floor i think that's the outcome but behind the scene this is what being done at an infrastructure level to support those use cases okay so in summary uh edge computing improves business outcome in real time with resilience and security by placing enterprise business logic and ai application closer to where data is created so ultimately we are bringing all the business logic and ai and algorithms closer to where data is created so that we can make decisions faster we secure our data we know we we know can be deal with the scale we deal all these problems by bringing this new concept of edge computing okay so i'll just stop here i think we are left with just probably five minutes and if we can go over five minutes we i'm happy to do that now i'll just stop here i'll stop sharing and start reading some of the questions so there's a question from ronak which cloud facility is better cloud is safe or not what can i use what can i use any cloud for build a website what is the cost of it which one is more trustable aws gcp azure none of the above so uh well i think the many questions in one question but ultimately it boils down to your concern of security and cost so see cost is is something which you need to explore what you need right and every cloud provider will have its own way of you know kind of estimating the cost of what resources you need so i will leave it to you you go and do your estimation and these are free estimations you can do your free estimations on all cloud providers and whichever is cheap you can go and go with that now security uh so from security standpoint uh i'm telling you the cloud providers have done a fantastic job by solving that problem because security was always a big deterrent for cloud adoption and today while there is no end to security problem security problem will always be there because the the hackers are smart developers they are nothing but the developers who who are more smarter than people who could anticipate the loopholes because ultimately every developer and architect is actually looking at we don't leave any loopholes any vulnerabilities in the system before they release that system but hackers are those developers who are able to crack it and still get it so they are the smart guys so you will always have some smart guy to break into now the question is how best we can deal with it right how best we can deal with that proactively and reactively both right so cloud providers today they are in my view again everybody uh it's it's a matter of kind of investigation on case to get bases but in my view every public cloud vendor has done a very great job in terms of enhancing the security of public cloud so and especially for your kind of use you want to build a website i don't think so you need to worry about your the security aspect of public cloud they are very secure and and you can probably go and even you can even you know kind of query them that how are they going to secure so they can you'll find enough uh you know content in the on the web to answer that question but yes in summary i don't think so we have a big security concern with public cloud anymore because they have evolved a lot uh then there's another question on security is a major concern about cloud computing uh while security is a concern not only for cloud computing it is also concerned with non-cloud computing is a concern for everyone so yes cloud computing is a is a is a method methodology wherein somebody else is using you know kind of processing your data but i think there are enough checks and balances and there are there's a constant enhancement and evolution of security you know checks and balances which i think which i think is taken care of very well in the public cloud space then devati there is a question how is cloud computing related to financial sector is it something which should be known to finance students also no i think if you are a finance student i think you are the the user of the application you are the user of the technology uh but you know definitely knowing how the technology is deployed behind it it's always interesting and you can read about it but if you are functionally a finance financial student and you are you know starting finance then you should be worried about your application then how it's being used behind it right then you need to leave that to people who are responsible for deploying those applications then there's a question from jatin how the share market will be affected by advancement in cloud and say other ai animal technology that's an interesting one so definitely share market is nothing but the reflection of if i keep the speculation aside so speculation is a share market also runs on speculation and there's no answer to that right uh there are big big stock market uh uh you know kind of uh analysts who who also can't explain you why some day market went up and come came down it is based on speculation but if we talk about the rational uh so stock market is basically work on certain rationals also and uh and rational is nothing but if companies making profit and company has better margins company can show you know kind of a better you know better profitability then the company is good the stock will should go up because ultimately the company is profitable making profit so ultimately any technology any technology if that can bring in the value in the margins and profitability of a company is definitely going to impact it's a it's a image in the market and it should definitely reflect in the stock market as well so yes why not the companies who are deploying good technology right it is actually nothing it's not only about their current operation optimization because they definitely prove a point in the market that we are the company who believe in optimizing our operation using technology that's a great statement it means they are they are saying that we are a futuristic company like if there's a sugar mill if the sugar will start applying ai and ml it gives gives a very strong message right which gives a very strong message that oh they are the orthodox business houses but they are they know that technology is the way to optimize the operations and ultimately operation optimization will show up in the in my you know net profit because i will have more same revenue but if i reduce my cost i will improve my margins i will improve my profitability so ultimately yes answer is yes if the companies are investing in technology to be more optimal operationally to be more profitable then why not it will show up in their performance and their performance will actually show up as a as a stock price again speculation can take a good uh profitable company also down that i can comment on but if we go by rational answer is yes uh the role of cloud computing the question from dr tata that the role of cloud computing present and future very strong role i think cloud computing is playing a very very critical role uh by you know kind of uh bringing a lot of you know agility a lot of you know kind of you know automation and you know you know and scaling up scaling down and it's kind of it's a cloud computing has proved its worth it's already proved its worth so it is a proven kind of concept in technology uh no i'm not cloud computing is not a technology it's a concept so the cloud computing is a proven concept and it is paying tremendous results in the in the in the iit world so that's the reason more than 90 people are already on cloud computing in some other form now in future ideas to build on top of it we have learned a better way of using our infrastructure we have learnt you know kind of better way of managing our infrastructure through cloud computing now we are going a level above to see how do we know kind of solve our next set of problems by building more concept like cloud you know like cloud computing earlier was okay one cloud then we came to hybrid cloud how do we create hybrid cloud then you know then then came a time when now now we're talking about hybrid multi cloud like i am a customer i am running one application from azure one application from amazon one application from private cloud right so i'm using multiple cloud and i can always move my application from one cloud to the other cloud now amazon and azure are not mine they don't mind it they're in fact enabling those features that okay if you like don't like me i don't want to lock you with me i'm not going to lock you if you don't like my don't like my platform tomorrow you go back go to amazon right so they are enabling all that because that's how the future is no one customer customer don't want any any vendor locking anymore so cloud computing to answer a question is is is very important the concept today almost clearly near 100 deployed and kind of practiced and has a very bright future because ultimately this has to evolve we need to keep evolving and make it even more robust and and you know kind of more useful for our applications okay so there are too many questions i don't think so i would be able to more than 100 questions near 100 questions [Music] the question is sir how can a non-engineering student get into cloud and edge computing yeah so certainly engineering student uh like to give you an example uh i know a lot of civil engineers doing very good uh it it design right so they studied civil engineering but they are fantastic designers today because they chose uh that career and they learned over time so i don't think so uh it's that if you are not a hardcore engineer but you definitely need to gain certain academic knowledge which you can definitely find out what you want to really want to do and you can build on top of that if you are interested in this space you can always study and and learn it why not how edge intelligence is different from intelligent edge see intelligent probably see intelligent edge i kind of gave you example uh that can we make our cameras more intelligent uh can we make cameras to you know or edge devices sensors more intelligent like today all the sensors what they do they just sense and pass that the the report to somebody else can we put another you know kind of uh chip into it to start doing one next level of decision right that sense this if you sense this then sense this as well if you don't sense this don't sense this as well so something i'm saying that you know can we make our edge more smarter if we can make our edge more smarter then we solve a problem of not passing on a lot of junk to the edge server because you are applying a filter ultimately what you do in your filter in your filter you take the junk out right so every layer you are ultimately doing some filtering means you are taking junk out at each layer so you as much as filters you apply you can start taking junk out at each layer so you you make more even better use of your you know edge servers or cloud servers for the data you actually actually action rather than spending time and cycles and money to start taking the junk out so by having intelligent edge devices you can probably have another filter in the edge device itself to throw some junk out and only send the relevant or useful you know first level of filtered data to the upper layer so that's probably you can say that that's the main difference uh you can say in between intelligent edge and non intelligent edge okay so we are i think already at seven eight i'm not sure uh if we can take it for the longer biology if you can hear me can we extend the session further should we be stopping here whichever question is relevant please uh you can okay okay okay so i'll probably spend few more minutes because there are some interesting questions which might how can we use and this question is from lalitha mishra how can we use edge devices in integration with 5g for better healthcare solution okay so see healthcare is definitely is an area you know which is evolving by using great technology you know innovations right and if you talk about i think i'm just because you mix two things here 5g healthcare so ultimately healthcare is also a thing wherein you need to make decisions very fast right and now this covered kovit 19 situation has challenged a lot of you know i.t folks to see what else can we do to to kind of uh deal with such pandemic or such situations wherein take an example a very live example uh like today we have there was a time when we had four lakh uh new patients of kovid identified now that if you do that put the kind of profiling of this now not all all of them need there are some asymptomatic ones there are some severe ones there are some very severe ones right so there's a profiling of each patient so what is happening today and that's a ground report there's lot of mortality but a lot of deaths happening because of insufficient medical attention and this is again i'm talking about just related to filtering some junk out is there a way to filter out the junk and probably i in this in this context let's not call it junk but i'm saying that filter out all those patients who don't need a md filter out all those patients who don't need even mbbs maybe they just need a paramedical stuff right so how do we do that is that all that is done manually today right based on the assessments based on their reports there is somebody who is diagnosing and telling them that was you need to get admitted you need this so there's a lot of manual thing happening and ultimately there are some non-severe you know kind of patients landing up to a md doctor who actually is a better you know doctor for somebody who is dying on some other hospital right so so this is a very high level example that we definitely need some automatic way by dynamic there was no time to turn out and find a better solution because you need some time to you know design a solution and apply that whatever as a country we have done in such a short time to deal with this it's amazing why we have seen a lot of damage but i think the country of this this came nobody could have done better than what we had the country have done so but from here whatever we learn i'm sure there would be too many enhancement and a lot of innovation is going to come in medical medical space pandemic and that you know kind of would be solved some of those innovations would be built on top of some of these 5g technologies and edge technologies right how can we so definitely because you can do a lot of things based on sensors with a patient now with those sensors those are edge devices your patient is sitting there patient has all the sensors in the hospital attached to him now can you make a lot of decisions can i collect that and start put applying some artificial intelligence there to a certain degree and pass it on to a doctor only after after reaching a certain threshold so yeah definitely you know there's a lot of scope to do enhance you know innovation in medical especially after seeing this this pandemic and you will see that coming in hospitals will be emitted healthcare will see lot of automation lot of artificial intelligence like they i don't know how many of you know and it's not in recent data it was i know there was a data published by by whom i think manipal uh hospital or somebody i think there was study saying that now the number of cancer patients in the world and the number of oncologists we have in the world the ratio is very very skewed like there are thousands of patients like one doctor in in many thousands of patients it's impossible for for us to deal with this kind of composition it's impossible right so what we do we need to definitely apply a lot of you know artificial intelligence to kind of do groundwork probably the 80 of the work can be done by artificial intelligence and algorithms before you actually go to and and consume oncologist right so so definitely yes medical is a medical field is a area or which has a lot of potential we are we are not one of the best you know kind of uh in the in this space in the world and this pandemic has given big setback to all of us that you know governments you know scientists everybody need to focus on medical system that how to make it more robust because nobody can guarantee that after covert there won't be any other pandemic for the whole you know for the whole life we need to be ready to deal with such pandemic right and there are space there are white spaces which can be which can be addressed using the technology like artificial intelligence machine learning you know analytics and all that would be running on top of infrastructure and that infrastructure is 5g cloud computing edge computing you know offload gpus cpus dpus i think all these are your pieces of infrastructure which make your all this you know great application work the way you want right okay uh what are the best what are the tricks now this is a question from aditya what are the benefits of cloud computing on automation in the development field what are benefits for computing on automation and product development really so automation certainly so cloud computing is it's or it is certainly automation is a big big advantage of cloud computing and how are you able to achieve automation in cloud computing because you are actually you know breaking the silos and you're bringing the whole workspace you know interconnected because cloud computing allows you to interact with different components of the business and apply you know any kind of because if you think about automation you cannot automate in silos because when i say automation what does that mean i say that i've written a script saying that you if you find this condition you do this else you do this if you hit this condition then you go there it's ultimately a kind of script which actually asks you to perform in actions i know there's no certain results and those actions need to be kind of integrated between different functions now if different functions are not integrated together in the single computing environment then how would you automate because then you are actually living in silos and you cannot go out of silos right without manual renewable intervention so cloud computing broke those silos bring all that departments all those components together and allow you know to kind of write a lot of automation and give you all those tools because i'm not talking about tools and technologies here because i'm just talking about the concept but cloud computing then enables you with a lot of tools and and comp you know kind of technologies which help you to do that automation right so so yes automation you in a product development you know kind of in the industry definitely cloud computing is is is very helpful in automating things okay so uh i think we can probably uh drop off for special cloud computing certainly there are jobs cloud computing is in space i think computing if you look at you will find a lot of jobs all around because now nobody is doing a in-house application development today if you develop any application it has to be a cloud application when i say cloud application means you should be able to if you develop this you should be able to deploy that into a cloud environment so cloud computing definitely all the i think every application developer is kind of working the cloud computing space because you are not developing almost any application today which is not cloud deployable okay so so probably we'll stop here but it was a a very different session for me because of the complete monologue i couldn't hear anyone uh and this is very unlike the sessions i've been doing in past but i wish i hope that it was useful for you all and uh don't hesitate to connect with me on my you know twitter handle and linkedin linkedin you know uh connections and we can always network talk offline and uh and also i keep sharing a lot of technology trends and what is happening in this space so watch that space and uh and learn i think i would say that learn as much as you can because this this is just the beginning which i just said and this is a very very high level of concept there's a lot devil lies in the details you just need to go in detail and find out what what what make you know what interests you more and learn learn learn as much as you can so thank you very much and again i wish you all remain safe and you know and healthy so stay at your home don't go out no wait keep some patience let this pandemic go away and then things would be very normal very soon i wish for that thank you very much over to your body

Channel: NPTEL - Special Lecture Series

Views: 8,320

Rating: 4.9469318 out of 5

Keywords:

Id: rifFwHPlOrs

Length: 78min 40sec (4720 seconds)

Published: Thu May 20 2021