green technology and transformation so also you not here on site but all the other people on site let's give him a very warm welcome amol fatka so [Music] morning morning everyone for those in the room and also uh watching us live great fair so far and i can hope to see more of it uh through the rest of the day so as anna said i am amol fatke managing director and gm for our comms business at cloud um i'm going to pick up from what claudia said in the break around how we all have to strive towards building a human-centric sustainable technology and that's what we are really going to do as cloud that you will see from me in the next few minutes so i wanted to cover three things hopefully all of you can see the material i wanted to first talk a little bit about how at google we are building a cloud native uh operations model which is which is also very very important to understand as background i will then talk about how we are using ai and data to build the sustainability that claudia talked about and then i will specifically focus on what we are doing to make technology sustainable for everybody and also the work we are doing with some of our partners and clients globally in order to drive uh the green technology and transformation that's really the the key areas i'm going to focus on today let's move on so it's very important to understand how we look at the journey towards a cloud native transformation because when we talk about human centric technologies and we start talking about how we can build sustainability it's important to understand this is not something that you can transform overnight this requires a bunch of things coming together as foundational elements before you can start driving towards sustainability so for instance the first thing that you need is a common platform because anything that you engineer on top of a platform would need to be sustainable but as soon as you start creating multiple platforms automatically by definition your sustainability goals start to become challenging to achieve culture is a fundamental point which will also has an impact on sustainability and i will talk about how that is in a minute i mentioned at the beginning having a lot of data at your disposal is actually an immensely beneficial point if we use that wisely because that can really give us insights on where we are not being carbon neutral or carbon free and that's where that can really help and again i will talk about that in a little bit and of course having business models that incentivize sustainability security and scalability are also going to be key pillars so those are really some of the pillars that drive towards cloud native but what does this word cloud native actually mean right and in google we actually measure how successful we are at at driving a cloud native mode of operation uh let me talk about some of the statistics that we as a company measure and give you a bit of an insight of what we mean by what does cloud native performance look like some of the numbers that you will see here at its peak when and clouds and google started on this journey for the last two decades at its peak we are a software driven company as we all know our code that gets developed because of the nature of how we now develop code is a thousand times more roughly than what we used to be so just to put it in context if we are thinking of deploying code into a production network and let's say that was happening once a year uh two decades ago that happens thousand times a year now and that's the scale on performance benefit that you get but it's not just deployment speed it's also the lead time so we've also seen a huge advancement in terms of cloud native performance you see there are 6500 times faster in terms of lead time what does that number mean again from a conceptual perspective let's say it took us four hours from committing a piece of code to deploying it in production that four hours has now got reduced to two and a half seconds that's what i mean when we say six and a half thousand times faster in terms of real terms that's that's what it means and similarly recovery time when you had an issue with your production deployment used to be in hours and is now in seconds right so again that's something that you get through a cloud native performance all of that speed and agility that you gain also comes with a lower failure rate and the reason for this is quite obvious when you start doing these many deployments at scale your code base automatically matures and yes while you get individual failures a lot more overall your failure rate goes down statistically the failure rate has to come down given the scale at which you deploy and you are also almost twice as likely to report on better business outcomes so that's what i really mean in terms of setting the bar on what cloud native performance actually looks like um now how can you achieve that right so there are three things you have to do one build a culture from my perspective our perspective that is really a cloud native culture have sustainability at the core of everything you design and then use data to drive sustainability so you'll see for the rest of this session these are the three themes that you will see over and over again culture use of data and sustainability in order to achieve that cloud native performance so i'll first start with what we built and pioneered as sre and in a nutshell what does software or site reliability engineering mean it means when you look at operating your networks or your it systems or your production environments we started getting our software engineers people who developed the code to actually fix operational issues this happened two decades ago it created let's say a lot of issues at the beginning when you suddenly got software engineers to develop issues and solve for operational issues but it totally changed the culture of how we approached operations right and that's really what became google sre as a culture so at the heart of it there are five principles to these and why is this so important you cannot achieve sustainability by having a sustainable agenda and talking about green targets unless the culture of the company actually changes we firmly believe it is extremely tricky to look at just having a sustainability agenda and then carrying on with operations as you traditionally do that simply would not work so that's why from our perspective the five principles are very important number one we have inside google a concept called as blameless postmortems and i emphasize the word blameless because i have been in the industry for 25 years and i understand that when issues happen and i used to be with a large operator prior to joining cloud so when issues happen the first thing is you are going to call up a vendor and blame them for an issue and it is human nature to do so um getting the mental discipline for becoming a blameless post-mortem actually helps and we've seen this work in google because then there is a psychological safety to actually solve the issue so for instance if a particular issue happens in a production network inside google the engineer has the courage to report things exactly as is as quickly as possible before sugar coating it sending it through various layers of management and it is very difficult to implement in practice it's very easy for me to say this on a slide but it's very difficult to implement in practice but extremely important point to just bear in mind the second related topic around culture is almost certainly an sre mode of operation breaks down silos in any big organization google has a large organization much like telecom we have an operations department we have an engineering department we have a development organization and having an sre organization that has the mandate to look across the organizations was a very important step because think of them almost like a swat team that comes in when there's an issue and has a blank check to go solve the issue regardless of where the issue is right so they have really got that kind of empowerment top down the third important principle is when you ask a software engineer to look at an operational issue a couple of interesting thing happens one they don't like to fix it because it's not cool but secondly as you start building that they start seeing patterns that an operational engineer doesn't see because if you're a software engineer you start seeing patterns as you build solutions for an issue you see when you see those patterns you get bored and you build the automation script for it that's exactly how you start infusing automation at a very basic level when you start implementing sre across your organization that's why in google we pride ourselves on automating every possible step anytime we see an issue we want to automate that issue and that actually creates the culture that i described which is let us make it better for the whole organization let's solve for the organization as opposed to solve for a particular department or team so let me conclude by saying this culture topic is a fundamental one i do believe let's say we get a carbon neutral goal or a carbon-free goal it will be non-trivial to meet that goal if we did not really think through the culture aspect of this and only focused on let's look at the carbon offsets and let's look at emissions and so on culture is an important topic and very very key for us to solve for that so let me now turn to the second topic which is data and how you can use the power of data to drive some of the sustainability challenges that we see in the world now as everybody in this room knows everybody's watching the stream would know google is an ai first company our mission from the founding was to ultimately look at information which is data and make it accessible to everybody in the world we realized very early on in our journey that the amount of search queries because remember we were a search engine to start with before any of the other things came on the information that we started getting back overwhelmed the capability of our data centers over when the capability of our operations completely to the point where we had to infuse ai and analytics in every layer of google's organization and i will show you that in a second but it's very important to understand for us it wasn't a sustainability goal for us uh implementing ai and analytics was a survivability goal if google's business was to be surviving in the two to three decades the amount of data we get we get something like 13 petabytes of data every month going through our platform it's impossible for any ml or data structure to work without having very sophisticated technology underneath uh to process all of that data so we started off by investing as you all know we have acquired tremendous amounts of capability over the last two decades we realized that as good as we are we will not have the expertise needed to solve these problems which is why we had to invest in acquisitions deepmind api ai lambda there is a bunch of acquisitions we've done over the last two decades and then we're using it to solve real-world problems like when you type a search and you see there is a ranking on search it is an entirely through an ai analytics engine that runs in the back end uh there is we realized very quickly that we had this vast information store to connect the world to and we were getting requests coming in and the only way to solve those very quickly going back to the sre example that i gave any time we see that the search query takes longer than 0.2 seconds that's our threshold our sre teams would get into action try to solve the issue use ai analytics and so on but there are very specific targets in terms of whether ai is working or not working inside google we use exactly the same thing if you look at smart emails that gmail writes it is scary that sometimes even when we are using this for our own internal email i just have to say hi and the rest of the email is there and i just hit send so it is it is a bit scary when you see that but ai analytics is coming in in pretty much every application that we build and i want to finish that before we get into the sustainability by explaining as i said before how we have now infused ai in everything we do so we've taken a very systematic and structured approach again the reason to explain this is to say that we can't just embrace ai ai and analytics requires a complete change of mindset and culture and also infusing it in every layer of your organization so in google we we ultimately had infrastructure that had to run ai so we had to invest in building our own ai chipsets in addition to of course partnering with the market so we came up with our own ppu chipset why is that important for sustainability our tpu runs 7x times the processing power of ai with the same amount of carbon energy that of the shelf cpu gpu have this is only possible when you start looking at the problem that is let me fix this from a sustainability perspective right so that was just an example of why we have also invested in hardware once the foundation platform is there that can handle the processing of data i talked about we have then built a set of platform tools going back to the beginning slide when i said you need a single platform environment all of google uses a single platform environment whether it's for cloud whether it's for search whether it's for mail and any other application and that environment that platform environment has a set of toolkits available that we also work with partners on right so partners use the same toolkit when they want to infuse ai in their businesses when they work with google once you have the platform environment then you can see you also have a set of custom models and pre-trained models google did these so i'll give an example when we work with a partner to implement contact center ai in north america in their contact centers you know telecom operators contact centers the speech-to-text and text-to-speech pre-trained models were the same models that we use internally for hey google because as you can imagine hey google requires the voice assistant requiring speech to text text to speech so it wasn't like we just came up with a contact center ai solution for our operator clients we use the same approach of saying we have pre-trained models that can get reused and why is this important for sustainability the more you reuse the less carbon emissions you have so the more emphasis we always put in google about reducing waste and standardizing for for models to be clear we only have less than 10 custom plus pre-trained models across alphabet across the company 10 models that get reused in every solution set that you build on top right and that's very important to think of it in that standardization way and then the upper layer is as i said where we partner with clients and other isvs in our ecosystem and then they build solutions using those models those platforms whether it's contact center ai in the financial services domain we are working with a couple of partners who are looking at document ai in during the pandemic a lot of the financial decisions such as insurances mortgages were started uh were moving towards getting approvals online and so on and so forth which requires ai algorithms running in the back end to do those things so again we work with our partners and i just wanted to give you a couple of examples on that so all this to say um as we pivot towards what claudia said in the break how can we build all of this in a sustainable way we have now realized through the work that we do with our ecosystem that sustainability has now come to the forefront and which is actually a very very good thing to see uh five or six years ago sustainability used to be a csr topic and there was some funding provided by the cfo because we had to have a sustainability agenda for the organization but it wasn't really infused now it is getting infused now when we start talking to any client or partner then sustainability becomes the number one or number two topic of conversation in addition to solving the operational problems so that's why as i said at the beginning although we may want to solve operational issues sustainability becomes hand in hand with that you can't really do it in a non-sustainable way and that's why it's very important we are seeing this change we are seeing this is a decade to do it and when i looked at some of the research that google cloud looked at more than a third of our largest telecom operators globally have committed to zero targets uh by 2030 by but which is you know not not not far 2050 i should say um so i will give you a sneak peek on what google is doing with respect to making our cloud platforms and our business greener and then the work we are doing with partners and then i will leave you with uh what are the key imperatives all of us should focus as an industry in order to drive towards the green technology uh initiatives that we are seeing so firstly what did we do we put it at the core since our founding which was you know 15 to 20 years ago back in 15 years ago we were the first silicon tech company let's say that had enough high quality carbon emission offsets done that we achieve our carbon neutral target so i know we talked a lot about carbon neutrality we achieved carbon neutrality in 2007 as a business and we are always striving for doing more which means five years ago the emissions that were happening of course we were still offsetting it because we were wanting to still stay carbon neutral but even the energy we were consuming became 100 renewable so we wanted to make sure we are moving towards not even using uh let's say the the traditional energy sources and we were moving towards 100 renewable energy and then by 2030 which is going to be eight years from now our goal as google is to become completely carbon free not just carbon neutral but carbon free and that's quite a major step because you can imagine that means we want all energy consumed by the business to be entirely made up out of renewables and not of carbon it's very very very ambitious we believe we have the toolkit to do that but it is a very ambitious program how do we achieve that program some of the steps we are taking and there are more to be done over the next five to six years our data centers as you can imagine google data centers carrying 10 applications each with a billion plus users consuming huge amounts of energy we have these facilities in 25 locations across the world and 100 plus locations beyond that that are sort of edge locations if you look at the energy consumption over time this is the most recent stat from 18 months ago and i'll benchmark this against typical data centers we are about 50 percent less which means we have a 2x uh benefit in terms of how we run our data centers now if you look at this in a little bit more nuanced way most of the savings carbon savings is coming from facilities not so much servers and i'll talk about why in a second the facilities in point is important what we've actually done is in the same footprint in the same power and energy footprint we've packed 7x more computing power over the last two decades what that means is our facilities cost in terms of power consumption and energy consumption continues to go down dramatically which is why you saw the graphic in the previous chart and that actually helps us continue to drive that now that is solving for the issue of how do we pack more computing power in the same power budget of the data center but the basic unit of computing which is the cpus the gpus and the tpus are still consuming a lot more power and that's the next stage of evolution that we are looking at in terms of how we work with the ecosystem to drive innovation in the hardware in the silicon layer because that will really drive down like how do you create quiet processes there is a lot of research going on right now that we are working with our partner ecosystem on can you create processors that can actually stay quiet unless there is processing to be done which is extremely challenging to do at a chipset level but every nano watt of power that you would save from being quiet would add up enormously because of the computing power that all of us consume right so that's something that we are working on um and that's what i said at the beginning like we built our tpu chipsets to handle uh ai workloads because we believe that they are 10x more energy efficient than what's on the market doesn't mean we don't use things on the market but we are trying to infuse that innovation thoughts in the ecosystem to drive that energy efficiency to a large extent so that's why i think this is extremely important to also understand why we are investing and i said this at the beginning in terms of investments have to be done at all layers from hardware to data centers and so on so how can we now offer all of these innovations to our partners and clients right so generally we have looked at partnering with retail with telecom with financial services clients and partners of ours and they are generally in three buckets number one how can we help reduce the impact of i.t the all operations require i.t how can
we create those it operations in a more sustainable way second how can we create operational sustainability so i'll give you an example we are working with a couple of telecom operators who are looking at radio access networks brand networks and correlating that data let's say with things like transportation schedules because and you would think why that is the case so when let's say a train comes to a platform a lot of commuters will get off the train and you will see a peak surge in traffic happening because they're suddenly connecting with with people and then vice versa when the train is in motion the radio networks next to a train are quiet but the power that they consume is constant 24 7 consumption of power regardless of demand so we are trying to correlate demand performance with power consumption data to try and see if there is a way to balance this again these are small steps that have a meaningful impact and a couple of operators in north america are trying this with us to see how we can make sustainable networks that are less power hungry and then finally claudia said this in the break um telecom operators particularly deutsche telekom is a large inspiration to a number of downstream supply chain ecosystem partners so if you lead the way and drive that sustainability agenda that actually creates momentum for this to become an ecosystem play right because otherwise ultimately if only one operator does it or one cloud provider does it it's not going to be enough the whole community has to do it for the energy sustainability goals to be to be achieved i'll talk about specific tools that let you do what i just said so we have something called carbon sense sweet this is not just for cloud any cloud platform that you are using you could use google's carbon sens suite to give you an example we looked at all the data from the last year with a particular operator just on the projects they were running in cloud not just our cloud any cloud we found 600 000 kilograms carbon kilograms worth of idle projects in cloud projects that are started in cloud as soon as you start a project in cloud you're consuming cloud cpu and paying for it and there was nothing there was nothing on those projects that's equivalent of 10 000 trees per year that the operator was losing because they were just putting projects in cloud and not using it all you have to do is turn them off and that's equivalent of planting 10 000 trees so these are some of the tools that have to get implemented it's not the only one we have tons but these are some of the tools that have to get implemented across the ecosystem for people to really absorb this ethos of sustainability and having a green agenda some of the other tools standard tools are things like how can we increase the life of devices it is counter-intuitive given all of us upgrade handsets every year now i upgrade mindset every year but i have to put the mental discipline to say can i use my device longer that's a big big incentive that device manufacturers have to evolve towards and also trading refurbishments trying to reuse the heap there was an article i read in howard business review two weeks ago where if we made a heap of all the electronics that will be bigger than mount everest so you can imagine the amount of electronics we are throwing away and refurbishing that is also a key let me finish by summarizing the five things that we discussed today number one don't put sustainability as a side project as a separate department that's what our learning suggests sustainability agenda needs to be in every department and we need to have sustainability experts in every department of every organization in every industry for this to succeed um having a metric like trees per year or carbon kilograms that are getting wasted could be anything but a matrix generally is very important because it galvanizes teams that's again our experience in google that brings people together and then they all drive towards a particular agenda and then measuring it which is a third point just having a metric is no use if you can't measure it and demonstrate that you are improving over time why is that third point also important you will get a lot more cfo budget from my experience when you show them that actually you're saving money otherwise you get the standard sustainability part of money whereas if you show them the holistic transformation you will get a lot more done through that also google was very transparent on some of the things that were not going so well for us so for example our pixel handsets um did not have the most efficient processing let's say we cleaned that up third-party reports showed that we cleaned that up our most recent pixel now comes with the tpu chipset the the google pioneer a1 chip that we built which is dramatically reduce the energy efficiency uh consumptions and so therefore that's an example of being very transparent about where things are working and where things are not working and then the final point is have some magic moments that actually inspire your end customers so over the weekend you know during pandemic let's say my family switched to doing online shopping in uk i'm based in london and i started seeing when i did the online grocery shopping time slots of when my shopping would get delivered and there was a green symbol next to it from the provider so i clicked there and and it said if you take in this time slot you're going to save 20 trees or something like this so immediately i click that because that's the time that worked for me so having some kind of real world external facing moments can actually make this a lot more real and that would lead to a lot more widespread adoption of this topic so anyway i wanted to leave you with you know the request to turn this ambition into action i'm fully behind what claudia said and we really have to look at this as a human-centric uh digital innovation era thank you for listening to me thank you for having me hope you have a wonderful day and enjoy the rest of the fair [Applause] thank you so much i'm all for sharing your views and your position from the google cloud so thank you so much
2022-09-13