 The title, yeah, it's a little grand, but I think the intention was a little sarcastic because I really don't think it's new. It's been very, very old. It's been there since many, many years and I think everyone, it's not just making a big deal out of it, but I think now people are really doing it the right way. And that's really the difference. And so let's just, the whole intention of this today's talk is really to give you a brief overview of not what Microsoft is doing. To me, there is nothing in this that's going to be Microsoft. The whole idea is I wanted to make it academic. I've based a lot of my talk on Berkeley's paper that is above the clouds. I hope you folks have read it. It's a brilliant paper and it's not that dated, but I think it's a little bit way back in 2009. So it's a very general talk giving you an idea of what we're doing. So this is what we're gonna talk about. We're just gonna give you a brief overview of what you mean by the cloud and then go on and see the economics and what is the driving force behind why people are really following this, why companies are really putting billions of dollars and really investing in that technology. And then let's, I've just kept a couple of scenarios that are real today, that one of them is still in research. We're working with a few partners, including one of my customers like Eric represents TCS here, we're doing some work with them in that space as well. So I'll just delve something on that and just to give you some food for thought that what you folks can really think about in your research projects is ever you want to go and do work in this, right? So let's just, I mean, looking at what cloud computing very, very simply put is nothing but applications over the internet, which we typically term as SaaS software as a service. And the second bit is really utility computing, which is really the all end and all end umbrella of where we really do computing on. So SaaS is typically based and utilizes the utility computing infrastructure space, which is nothing else but pay as you go for the hardware and for the software that we give you. The interesting part was that nowadays there's some three new aspects that we believe have been introduced in the area of cloud computing. One is the whole illusion of having infinite amount of resources, okay? The whole point is that if you have a private data center in a company or in your organization or whatever, you know that there is some finite amount of resources, you want another server, you need to have a requisition and get those servers filled in, installed in whatever and configured. But in our space and in the cloud computing space, the whole illusion that the end user has is that I have any amount of computing power that I would want. All I need to do is go and provision it and a matter of minutes and hours, depending on which technology you're using and which vendor you're dealing with, you really have that technology at a disposal. The second one literally what comes in is very, very interesting and I think this is one of the most important drivers of the adoption of cloud computing today in the commercial space, is the elimination of no upfront commitment. If you ever get into working with enterprise customers in your careers later on, you will realize that there is a lot of hesitation, a lot of apprehension when I say that I need to get, install a big, huge data center which costs huge amount of money. And literally I don't know what the demand is gonna be like. So it's a lot of commitment, it's a lot of money that gets spent on this infrastructure and we attempt to take this away onto the cloud, de-risking you from that investment and that's really one of the driving factors. And the third is literally to have the ability to pay for computing as and when you need it. It's like eating food and paying for the food when you're hungry. It's as simple as that and that really is changing the game now, it's really getting the demand going on. Like I said, personally I believe that economies of large scale or what drives the, it's a compelling need for people to move. By no way, if anybody says I'm gonna move everything to the cloud, it's gonna be possible. It's not realistic as of right now, there are a lot of issues and we'll be looking at some obstacles later on which prevent that from happening. But there are definitely some workloads that we are using and moving to the cloud today and some customers, right? So some things like, for example, what really drives them? Like for example, Microsoft, the kind of data centers we have. We put it right next to where electricity is cheap. If I put a data center in Bombay for Azure, it's not going to be serving the purpose because it's not cheap power, right? So there are some locations in the US and in Europe and wherever that we put data centers, the power is cheap. It's next to hydroelectric power. It's next to the internet pipes that are going out. So the internet, the network bandwidth is cheap, okay? You have automated systems that provision and manage those systems which make it very economical because you don't need people to manage those data centers. For example, we have an average of two people to manage 96,000 servers in the container data centers, right? So it's a lot of servers and it's very few amount of people because that's all automation that's been built in. And that's where we are doing cloud computing and I think everybody, all the companies that we're talking about, we Google, we at Amazon, they're all doing it in a very different way. Even though cloud computing is an old concept, but the way we're implementing it right now is very innovative, right? So that's really how we do it. And the most important part also on the second part for cloud computing which we call as platform as a service, right? Where we give you a platform, you can create applications on top of that platform and really go and do stuff. They have some three different app models that we have today. I put a fourth box there because personally, I believe that there is a lot of innovation going on. So I don't know who will come with some crazy idea, probably one of you guys, right? And we create another app model. So the first app model that you see is, it's like a spectrum. One end is a very low level app model wherein you have closer to the steel virtual machines that are running. And we give you access to those virtual machines. You have everything that you need to do full control of those virtual machines. And on the other end of the spectrum, you have this high level abstraction. And typically, a Google App Engine is a very good example of that. A very, very robust internet application-based platform which gives you a platform to develop internet apps. And you have to write it in Python. If you know Python, it's good. And if you know the Google API, it's extremely good. And it's very, very robust. It gives you a very good scale when it comes to the constraints that are there in place. So the important thing is, more the constraints, more the automation, more the availability, because you cannot fool around that much with that environment. And on the left-hand side, the constraints are very less. You're doing everything. The only thing that will provide you is basically a platform for VMs and provisioning and so on and so forth. So it's more flexible, but it's less automated, right? So you remember that. More constraint, more automation, and no control. And there are reverse also. Now, where do we play? Now, this is a very, very niche segment. So the Azure platform from Microsoft really comes in between. I'm not talking about Azure because I want to, but it's really very important that these different app models should be very clearly articulated to you folks. So what we do is the platform will give you access to .NET to be support PHP, MySQL, everything, Python, and so on and so forth. We give you that model, and you are not that constrained, but you are not very free as well. So it's really in the middle where we tell you, here is an XML file. You put your constraints in this XML file, and then we'll do auto provisioning, auto scaling, and so on and so forth for you. So these are the three app models that are coming, and the fourth one, let's see if there are anybody else who's doing any innovation in the space, because there are a lot of people who are doing a lot of work. Now, what really drives this? So very, very simple things. If there is demand that is varying over time, it makes people very uncomfortable. I deal with a lot of enterprise customers. I manage Tata Group as an account for Microsoft. There's a lot of group companies in Tata Group. I talk to a lot of people. There is a lot of investment that goes into these data centers when you talk about workloads. Now typically, a workload is very seasonal. You don't know when it's going to spike up. And when you don't know when it's going to spike up, you really don't know how much money to put in the bag. And the second point is you really don't know when, what is the amount of demand you're looking at. You can just get a spike, and you are not provisioning your data center, and then everything goes down. That's a typical example. When you're in for a startup, you're starting up a new company, and it's a web-based startup. You have this beautiful site. Imagine Facebook starting up, and suddenly it gets so many users, and it doesn't know what to do because it has its own data center. So it could not anticipate demand. And does it mean that you should not be successful? That's the question. So the point is for you to be successful, and technology not limiting you, it's a very good platform for you to move to the cloud because it's cheap. Not cheap in the sense cheap. It's cheap because you will pay as you go, and you will not commit a lot of capital expenditure beforehand. Now this example is from the Berkeley paper. And I typically like this example a lot. And my slide was based on that. In the way back days, when we used to be making chips, a lot of innovation was going into making circuitry. You needed integrated circuit foundries to really process that and get that thing into production. It used to cost a lot of money. And as and when the chip starts becoming smaller and smaller and smaller with a lot of innovation, these foundries started costing more than $1 billion. And only companies like Intel and the likes would really be able to afford such foundries. So what people started doing, and we saw a lot of change in the way these things were getting done. The people who owned these fabrication units started taking work from outside. So there were a lot of innovative designing companies that started coming up into being. Like a very good example is NVIDIA. NVIDIA doesn't do their own fab. They use existing fab units because it's very expensive. So NVIDIA is very good in designing the chips and they design the chips and they take it into the foundry, get it done, and then take it to market. That's exactly what happened to cloud computing. And utility computing is really coming into being for that. There's a lot of money that goes into building these data centers. And why should you, when you have companies who are already having them? And believe me, at least in our case, we had these humongous data centers that we built for our own internet IT and for our life services and so on. This is actually a manifestation saying, you know what, we have these data centers. Why don't we do this? And this is what has happened with many companies. So it's very easy for you to then take your technology up there and use somebody else's platform. Well, this is just a brief on how really a SaaS and a utility computing collaborates together. So literally a SaaS user, for example, if I'm using a technology online, I'm going to be based on using somebody else's platform or application, which is ultimately being host on a utility computing platform. And that's really the relationship between these things. So utility computing basically gives you that flexibility. This, I thought it was a very good example of why it breaks a lot of sense. So this is an article that had come way back in 2007. And I thought I'll get some slide out on that. So if you see, it's very clear that, so the server cost is a line in the blue. You have a line in the deep red, which is the annual infrastructure and energy cost. And if you see in this graph what is happening that in 2001 onwards, the annual infrastructure and energy cost started becoming more than server cost. The reason I'm giving you this slide is in giving you this information is because if you understand that the cost of infrastructure itself started exceeding the cost of servers way in 2004, I think so. Yeah, that's right. So after that, it really didn't make sense because even the cost of energy started overshooting cost of servers. Now these are some things that really make you think that how feasible it is for me to have an on-premise data center for the kind of stuff I do. Well, it's not really true for all companies or for all organizations, but there are situations in which you will make, you will decide to kind of go to the cloud. So these are some things I was talking about. We have a data center in Idaho. So for example, Quincy. Now Quincy, we have a data center there. Literally, it's all middle desert. There's nothing there. The most important thing is that it's next to the river. So we get free, very cheap hydroelectric power. So that's the reason we base the data center there. And these are the driving forces for which we decide which location and everything we're going to base our data centers on. In fact, if you go and search on YouTube, you'll find a video on the Microsoft data centers. It's a very interesting one. Our architects are talking to you about how and why they chose the location and what are the driving forces behind that. So I encourage you to go look at it. So this is just an example. In Idaho, you have some 3.6 cents for power. And then if you see in California, it's 10 cents. You know why? Because the grid really goes all the way. So it costs money to take it there. And Hawaii, literally power is generated by using the energy that has been shipped on board a ship literally, because they don't have any generation there. So it takes a lot of money to really do that shipping of raw material to burn power. So that's the reason why you will not find anything in Hawaii. I mean, you only vacation in Hawaii. So the point is that you need a data center. We already have one. And then you use the data center of existing large companies to get things done. I want to quickly jump into the economics part of it. This literally, why I put this is basically some compelling reasons why companies started going and becoming cloud providers. So if you ever want to start your own venture, and if you have a lot of venture capital funding, you can probably think about these things. There's a lot of money involved. It's a good business to be in. This is the application model I was talking about. So I'll just give you a quick overview of this. So like I said, the less constraint and the more constraint, the more constraint will have more automation, because it's very easy for us to know. It's very predictable. I don't let you fool around with the platform and so on and so forth. Now, the question which one model will dominate, it's very difficult to say. A very good parallel that we can take from earlier days is if you're a C++ programmer, and if you're right now using Ruby or you're using ASP.NET or whatever, C++ you can do a lot of good things. And personally, that's all our favorite, because in school we used to all write code in C++. But the point is, if you want to write a website in C++, how difficult is that? It's really a lot of work. You will typically use a Ruby on Rails and give you that infrastructure to really develop a very fast, nice-looking, attractive website. So the point is everything has its place today. And what I believe is that all these app models are going to thrive. There is a place for everything. You have Amazon which is doing amazingly well. And I believe personally they are the leaders in cloud computing. They came up with it first way back and they're doing a great job with that. But it's a different app model completely. They're on the left-hand side of the spectrum. So there's a lot of space for innovation and let's wait and watch what's gonna come. This is the favorite section literally because everything is about money when you talk about the cloud. The technology is fantabulous. I mean it's very good for us to work on the technology, but there is a lot of compelling reasons to drive those things. So we'll just look at those reasons out here and then try to figure it out. So in the cloud it's very easy for you to get a resource whenever you want it and that's what. So in here, if you have a peak, you would generally provision for peak. When you have a data center, you will provision for peak for that particular load. And this graph will show you that these are the things in gray. You know this color is not showing, but it's gray. So these things between the troughs, those are wasted capacity. That is the time when the data center was idle for so much time. It's a lot of money and a lot of energy, a lot of infrastructure, a lot of people getting wasted. Now that's really the problem that you can solve by going to cloud computing. So it's a simple example. You're assuming that your service is, you need a peak of some 500 servers, you're averaging about 300 servers of compute power. Now what would you do when your actual utilization will really be 300 into 24 hours, which is going to give you 7,200 server hours per day. Now normally you would end up provisioning some 12,000 hours. And a lot of those hours will go wasted when you're not really utilizing the data center. So in a pay as you go model, the break even point with reach, if the cost of your server computing hours is not more than 1.667 that cost, right? So it's a very simple equation. You know that you can provision 7,000 and grow as and when you need the capacity rather than having that huge data center and provisioning for 12,000 hours when you don't need it. So it's, even though the pay as you go hourly rate, if you would find on Amazon or Microsoft or Google the leading technology, cloud technology providers is a little higher than what you would generally incur inside. But still it makes a lot of sense when you look at the economics over a long period of time. So it's a economies of large scale game and it's not saying that why should I pay that X amount of dollars per hour or whatever to get my cloud computing done. And you will figure it out. So there are a lot of these examples, very important examples, seasonal demands. You know recently in the Super Bowl, there's a Super Bowl event that happens in the US, right? Of American football, what they did, what Domino's did, they put the entire website on Azure. Because a lot of pizzas that get ordered during that time. So the site didn't go down and they could provision, they could scale. Now imagine if they have their own data center, they have to put so much of money just to provision during the Super Bowl and the rest of the year, the demand is absolutely flat. That's really one of the scenarios where things will happen. Now second one, the risks of under provisioning. Now if you see what happens if you under provision, the area on top of the red line, that will be the time when people will not get access to your website. And what will happen as a result of that, you will basically end up losing business. You see this? The graph will really go down because people will start going away from the website. Because they could not get access to the website. So you don't really want to get into the situation. This is a very good scenario again for me moving into the cloud. There's a couple of examples I thought will make sense. Now both of these examples is on AWS. It's none on Microsoft, none on Google. I think AWS has some amazing examples. So Animoto and target.com, everybody knew what happened to target.com and Amazon during the time when it was Black Friday. These were the only two sites which were up. Because of the thought leadership and saying that I'm gonna host it on AWS and whatever and I'm gonna get the economies of large scale and the infinite amount of resources that I need. So very simple, this equation is literally what you will end up getting to when you decide when you want to move to the cloud. And we look at this equation in a minute. So it's very simple. It's basically the amount of revenue that you would make for the cost of the cloud that you're going to take for hosting it. And how much are the user hours that you're going to use on that cloud? This value should always be greater than the right hand side which is literally hosting it in your own data center. Now the reason it will always be greater is because the cost of the data center is one but the bottom part, the utilization, divided by the utilization, the percentage utilization that you're having over a period of time. And believe me, the amount of revenue that you would end up losing because of doing this is going to be much more. So the compelling reason will definitely be the left hand side which is the cost of running it into the cloud is definitely greater, is definitely lesser and the revenue is definitely greater than the right hand side which is hosting it in your own data center. It may not be true for everybody but it may be it's very relevant to some businesses and to some organizations. So this is another thing. Now this is based on Jim Gray. I assume everybody who knows who Jim Gray is. He was a great researcher and very unfortunate that he got lost and lost his life and when he was sailing out in the sea he was part of the Microsoft research group and very respected person. And he did this research way back in 2003, very simple analysis, saying that what was the cost of things way back in 2003 and after that what I've done in this slide deck is literally saying that what was it costing for network, for storage and for CPU in 2003, what's costing in 2008 and what is the cost performance improvement? So if you see in short, the cost performance improvement for a van is 2.7X for CPU hours is 16X and for disk storage is 10X. Now the point is that how much did it take to rent for a one dollar in way back in 2008? That's the rate that you see down. And the slide deck is gonna be up so you can analyze later, we don't have time now. But the point is that the cost, if you compare it by running it into your own data center the cost of renting it in the cloud seems to be, it seems to be a bad deal. It seems to be expensive at the first look but you go down, dig deep and you'll find that we have not accounted for these costs. There's a lot of additional costs that you incur in your on-premise data centers especially when you're running huge ones. Now when you add up all these costs, believe me, pay as you go model makes a lot of sense and you can do the math later on. So that's really the premise of cloud computing the compelling reason why enterprises are thinking about it and that's really why I spent so much time in the economic section because for you to articulate applications now what I would look at you folks as students people who are doing research saying what new things can we come out with what new scenarios can we come up with to help India for example. There's a lot of scope that we can do and help governments using the cloud technologies. If you ever get a chance please view go to the professional developers conference website that Microsoft hosts every year and Vivek Kundra who's the CIO of the US federal government. He has a brief speech on what the cloud can do for the US and they've already started doing it. You know hosting NASA data on the cloud and stuff like that, right? So it's very important that we can do the same stuff for India and make our country much more with technology and helping it. So these are some workload patterns. I just thought I'll put this for you. Some analysis reasons why you would want to go to the cloud and get things done. There's some obstacles and some opportunities very common sensical things but these are some basic things that we're still trying to answer. A lot of people are doing a lot of work. It is both technical as well as policy and there's a lot of bureaucracy involved as well. For example data location, how can I have data located in some other country if I'm a bank, a lot of these issues. I mean you have some data lock in then you have consideration like we have a very nasty thing of being a platform lock in company. So people say oh man if I go to Azure and you give me your APIs and you lock me in. Well that's true, you are locked in. You don't have a choice, right? The same with Google and the same with Amazon. That's the way it's going to be because they cannot give you flexibility to an extent and then make sure that the app is always up. It's not going to work. So there has to be some level of proprietary that's going to be built into that. And people are working on a common ground to say that we expose the APIs as REST services and so on and so forth. That's the Microsoft part. I'm not going to delve deeply into that. We don't have too much time but these are some case studies that you can see how we have sort of Domino's one that I was talking about is here and we just don't do Azure. We have a host of SaaS services that are online as well and Coca-Cola for example is running completely on Microsoft online on the cloud. All of Coca-Cola including mail, communication, everything, right? These are some fast facts on what we have done. There's what investments, some numbers into how many years we've been in the cloud and what services we have online. You can refer to it later, okay? And these are all the services that we do on the cloud, okay? And this is literally really how we view the cloud. So from Microsoft perspective, we say that we believe in giving you power of choice. So if you want to have a combination of on premise and on the cloud, you are free to do that because we have an identity management and syncing framework that we can make sure that the application can actually stretch across your private cloud onto the public cloud using and without losing your identity of the corporate security that you generally do. So this really is the view of our view of the cloud and you can read more on that later on. And you can see that there is a power of choice on every hand, so you don't really have to stuck to the web model or go only on the cloud completely. I would like to take you to this and this is why I'm jumping the slides. Now these are some things that are very interesting and I'm giving you these examples because it's about the environment. Now we have come up with some energy-based projects. Microsoft Home is one such project which is basically a web service-based system and it literally, it gives you this whole online portal wherein you can log in and you can create the profile of the house you live in, okay? And it works only in the US right now because the data that has been generated is for the US and they are going to extend it to the rest of the world. So you can fill in the data for your house and it uses some statistical models and the model that it uses basically is the DOE model and it uses the one, we have licensed the analytics model from the Lawrence Berkeley National Laboratory which is the best model existing in the world today for this, okay? And it's a lot of number crunching. So if you say that I have two doors, four windows, you know, my carpet is of this quality, my air conditioning of this voltage, whatever, this algorithm needs to compute a census data over a period of many, many years and tell you how you can save electricity. And it's front table is go to microsofthome.com you will realize that it can really help a lot of environment related issues, okay? It will actually tell you how to save power. So why are we running this on Azure? The reason is because the computation that is required to generate that kind of reports without even visiting your house requires a lot of compute power, okay? So we spawn up so many instances of the application and this compute and make sure that you get a data done. Similarly, Amazon has a lovely case study when Hillary Clinton was the freedom of speech, the freedom of information act was exercised. Her history of the first lady of the US needed to be publicized onto the website. So they need to convert a lot of PDF documents into searchable text documents. You know what they did? They spawned 200 instances of EC2, okay? And they process this PDF documents into searchable text. It's a brilliant and very simple example of why cloud makes sense. You can find it on the Amazon website. So this is the whole market of electronic vehicles. We have fuel-based vehicles today. A lot of companies are thinking about electronic vehicles and there's a lot of complexity involved in this. There's a billions of amount of transactions per day and a lot of records getting stored. This is why we're using Azure to really take it up saying we're working with Nissan and we're working with the BMWs of the world saying that how we can use this model saying if you fill in the battery, how far more do you have to go before you find the next battery station, okay? These are the things which really make sense for the cloud and not just taking some application for the heck of it and putting on the cloud. You need to have scenarios that makes a lot of sense and that's where we look up to you folks to come up with new ideas, okay? I'm going fast because we're kind of done with this and what I would like you to see is the takeaways basically that cloud computing and the apps delivered, the economics are changing. There's a lot of economical belief that is driving this particular initiative and there are many obstacles to cloud but it's going to become ubiquitous as we go. There are a lot of people doing a lot of work, okay? What I would like to close my talk with is there is tremendous potential. This is not a new phenomenon, okay? This is an old phenomenon but with a new flavor. There are a lot of people doing a lot of work now which making a lot of sense and we're taking a lot of strides in the world today and I would just like to show you a brief video which will give you an idea of what really the kind of investment that companies are making in this whole initiative. These are some references, all the stuff in my presentation is from here, okay? The thoughts are very suboptimally original, okay? I've taken a lot of stuff from Berkeley because I believe they are a great paper and a lot of talks from Jim Gray who's written a lot of papers also and you should read, I just put this last night, you know, Microsoft on cloud computing, Steve Burmer gave a speech yesterday at the University of Washington. Please watch that. It's a great speech and it talks about how we are doing a lot of work on the cloud. It's a lot to learn from him, yeah? So I'll just play the video so that we can give you a perspective of what I was talking about. Thank you so much.