Transcript
Lawson: We need to radically reduce the CO2 we emit as a planet. Currently, the earth is 1.1 degrees warmer than it was in the late 1800s. Science tells us that in order to avert the worst impacts of climate change, we need to restrict that 1.1 degrees to 1.5 degrees. In order to do that, we need to reduce our carbon emissions by a massive 90% by 2050, and 45% by 2030. If we don't manage to do that, and we let the temperatures rise to a whole 2 degrees, then this is going to result in billions of people being exposed to extreme heatwaves on a regular basis, millions being exposed to drought conditions, and also sea levels rising a whole 10 centimeters, which is going to cause global flooding and hurricanes. It's imperative that we try and keep the global warming down to 1.5 degrees.
How are we doing with respect to those goals? In order to keep to 1.5 degrees, we need to reduce our carbon emissions by 45% by 2030. This is an infographic taken from the United Nations IPCC website. It shows that we're not doing very well towards that goal. In fact, we're currently on a course for increasing our carbon emissions by 10% by 2030. It's clear that the task ahead of us is massive. We're going to need governments doing everything they can. We're going to need individuals to take personal responsibility. Also, businesses are going to have a huge role to play in this too.
Outline, and Background
I'm going to talk about the tech industry and how that's impacting the climate crisis. Next, we're going to have a look at the anatomy of a data center footprint. We're going to see, why does a data center have a footprint in the first place? What are the various components of that? Next, we're going to take a deep dive into the cloud, which is one of the dominant forces when it comes to tech carbon emissions. Then, finally, the main takeaway from this talk is, we'll look at things that everyone can start doing at their company to get them on the road to reducing their own carbon footprint.
I'm Paul. I'm a Software Architect working at a company called Supercritical. I'm on a team that's building a platform to help businesses measure and reduce their carbon footprint. This is a topic of particular interest to me.
Tech and the Climate Crisis
Next, we're going to have to take a look at the tech industry and see how that's impacting on the climate crisis. If you follow the news at all, then over the last few years, you've seen an ever-increasing number of articles that are concerned about the impact the tech industry is having on the planet, in particular, in terms of global warming. Some good articles are grounded in science, and a few of the others are perhaps a little bit more sensationalistic, like this one. There's a common theme that people are concerned about the environment, and they're concerned that the tech industry is having an outsized impact on it. What can we say about that? A few facts. In 2020, Amazon, Google, Microsoft, Facebook, and Apple, when you combine all of the electricity usage, it comes in at 45 terawatt-hours of electricity, which is approximately the same as New Zealand. The tech sector accounts for 2% to 3% of global greenhouse gas emissions, which, for a bit of a comparison, that's actually more than the airline industry. The airline industry is responsible for just under 2%. The tech industry is expected to grow rapidly in the near future. Their share of emissions will also grow. Data centers, specifically, their electricity consumption is currently around 1% of the world's electricity consumption. That is forecast to grow to more than 3% by 2030. This is actually excluding cryptocurrency as well, which I think is roughly a quarter to a half percent.
If you work at a tech company, then you will have a carbon footprint, and not all of that's going to be related to the tech part of your company. This is actually what a typical tech company at the moment looks like. Roughly 15% of their carbon emissions will be down to employee remote working. This is, for example, employees turning their heating on to heat their homes, charging their laptops. Carbon and software comes in in second place. This is their own software that they're running, and also software of third-party services that they're using. Then after that, there's things like business travel, marketing and advertising, consultants, food and drink spend, and there's a long tail of other things. Cloud and software comes pretty high up, as you might expect.
The Anatomy of a Data Center Footprint
To help us understand why tech has such a large carbon footprint, we're going to take a look at data centers and how they actually generate carbon emissions. To give a sense of scale, a server inside of a data center is responsible for around 2 tons of CO2 emitted into the atmosphere every year. To give a perspective, the average UK person's personal footprint is 10 tons of carbon dioxide. A server is roughly like one-fifth of a person. Servers are not the only things responsible for emissions, also transmitting data has a carbon footprint too. Transmitting 1 terabyte of data over the internet emits around 5 kilos of carbon dioxide. Storage also has a footprint too, so like storing a terabyte of data for a year will be responsible for emitting around 1 kilo of CO2. For most companies, running servers, running infrastructure, running instances in the cloud is going to be the biggest part of their carbon footprint, followed by data transmission, followed by storage. That's going to vary from company to company, though. If you're a streaming service, then data transmission is probably going to be the largest part of your footprint. After the actual footprint itself, you can roughly break it down into two different components. One of those is called operational emissions. Operational emissions is the emissions due to the day-to-day electricity usage of your hardware. Then the other component is called embodied emissions. What embodied emissions is, is the one-time manufacturing footprint of your hardware.
The relative size of these, say for a production server, it's likely that your operational emissions will account for around 85% of the carbon footprint of the server. Embodied emissions, that's the manufacturing emissions, will be responsible for around 15% of the footprint. There's a lot of information available on the internet about the footprint of specific servers. Dell actually is really good at providing quite detailed information on their servers and their carbon footprints, and even breaks it down into operational emissions versus embodied emissions. If you're running Dell servers, then that information is going to be easy to find. If you're running some other servers, then maybe they publish their own information, or if not, then you could just find an equivalent Dell server, one with similar specs and get an estimate that way.
I'm going to look a little bit more closely at operational emissions because this is going to be useful later on when we come to look at the cloud. Operational emissions have three main components. The first one is the hardware power draw. This is just the emissions related to the direct electricity usage of your hardware. After that, there's data center overheads. It's not only your hardware that's drawing electricity, there's also the data center itself. Data centers are very power hungry, and most of their electricity usage is going to be powering the hardware that's inside the data center. Also, the servers that it's running, generate heat and require cooling. Actually, a surprising amount of energy gets used by the cooling systems and other overheads inside the data center. There's actually a term for this, it's called power usage effectiveness. Power usage effectiveness is a measure of energy efficiency of the data center.
The equation itself is that it's the total power that's in use by the data center, divided by the power that you use just for the IT equipment, and so the remainder is mainly things like cooling. An optimal data center, the best possible number for power usage effectiveness will be 1, which would mean that you'd have zero overhead. It's like all of the energy that you'd be using will be going to power the IT equipment. If you had a power usage effectiveness of 2, for example, that would mean that 50% of the energy was being used to power the equipment, and 50% was being used to power the cooling system and other overheads.
A typical data center has a power usage effectiveness of 1.6. This, we're talking here commercial data centers rather than cloud data centers. A power usage effectiveness of 1.6 means that 60% of the electricity draw is going to power the servers, and a whole 40% of the electricity draw will be going to power the cooling system and other overheads, which is quite a large number. If we compare that to cloud data centers, they actually do a lot better. Cloud data centers have a power usage effectiveness of around 1.2, which means that 85% of the energy is being used to power the IT equipment and only 15% is going into overheads. This is an area that cloud infrastructure providers excel at. They've innovated an awful lot over the last 10 years to make this number a lot lower. It's pretty clear why they would do that. Obviously, if they're much more efficient, then they're going to be paying a massive amount less on electricity. That also means that they can make their services cheaper and be more competitive. There's been a huge amount of innovation in this area.
To give a bit of a flavor of what that could look like. Back in 2016, Google published this paper, their DeepMind team built a large neural net model, which managed to reduce their data center cooling bill by 40% in one data center. What this did was it took data from a whole bunch of their different data centers. This will be data from temperature sensors, data from other electrical equipment in those data centers. They built a model to optimize for the power usage effectiveness of the data center. The data from that, and fed back into the cooling systems, turn various parts up, turn various parts down to optimize the cost effectiveness of it. In a single data center, they achieved a 40% reduction in the cooling bill, which is pretty amazing. Unfortunately, I can't find any much real follow-up to this. I don't really know whether this got rolled out to lots of other of their data centers or not. It gives a picture of the type of work that's going on to improve data center efficiency.
The top part of the equation and the hardware power draw plus the data center overheads combined is really a picture of the electricity usage of your infrastructure. In order to translate that into the amount of carbon that's being released, you also need information on the local grid where you're drawing the electricity from. You've got this term called grid carbon intensity. Grid carbon intensity is a measure of the amount of CO2 that's being released into the atmosphere for when you use a unit of energy, a unit of electricity. This differs from country to country and grid to grid. The main reason that this differs is based on the energy mix in that area. If you're in a country that relies purely on coal and gas, then you're going to have a very high grid carbon intensity, and you're going to be releasing a lot of carbon dioxide when you're using the grid there. Once you get in a country that has lots of renewable energy sources, or is using lots of nuclear, that has a low carbon intensity, then your grid is going to have low carbon intensity too. If you're actually interested in this stuff, then I highly recommend going to this website, electricitymaps.com. They've got a really cool product. It's just really fun to click around and see a bunch of data, but you can get real-time information on the grid carbon intensity for various different countries. Not only that, they also give you a breakdown of the fuel mix. You can see here, for example, Germany's had like medium grid carbon intensity. They're using a lot of coal, but also, they have quite a lot of wind and solar. You get lots of really interesting information just going here and looking around a bit.
An extreme example of the difference that the grid carbon intensity of an area can make is Iceland. Here you can see, on the top left is Iceland's largest hydroelectric plant. On the top right is Iceland's largest geothermal plant. In Iceland, as a matter of fact, their grid runs 100% from geothermal and hydroelectric. On the bottom right, you can see a picture of a data center in Iceland. That means that if you have a data center in Iceland, then you don't need to worry about the operational emissions part of your carbon footprint at all, because the electricity that you're using will be 100% renewable. You only really need to worry about embodied emissions.
Summary (Data Center Carbon Footprint)
A data center carbon footprint has four components. There's the embodied emissions of your infrastructure. That's like the emissions that are related to the manufacturing process. There's the electricity usage of your infrastructure. There's the electricity usage of the data center that's housing the infrastructure. Then there's the carbon intensity of the local grid.
How Green Is the Cloud?
We're going to take a look at the cloud, which is one of the dominant forces when it comes to tech carbon emissions. If you go to one of the major cloud providers' websites and look at their sustainability page, you'll see some statements like this on their percentage of renewable energy that they're using. In fact, AWS and Microsoft both say that they're on target for using 100% renewable energy by 2025. Google, in fact, say they've already been using 100% renewable energy since 2017. When I first read this, it really confused me, because I'm thinking things like, on the cloud providers, just taking energy off the grid, and the grid isn't 100% renewable, so how can the cloud providers be 100% renewable? I was also thinking, in this case, why do we actually need to be really worried about our cloud carbon emissions? It seems to contradict a whole bunch of stuff I've been reading about how worried we should be. I started trying to figure out how to really make sense of all of this. Unfortunately, it's very difficult. Most of these pages don't particularly elaborate on what they actually mean by 100% renewable energy. Where they do, it's very jargon filled and difficult to understand. This section really will hopefully help demystify some of these claims.
What do they actually mean by 100% renewable energy? The most obvious thing they could mean would be that they all have on-site renewable energy. You could imagine that every data center has a massive bank of solar panels, or a wind farm attached to it, or something like that. Actually, this is impossible, because, firstly, just the amount of space it would take to do that is much bigger than these data centers have available. Also, the sites that these data centers are on are not very often an optimal site for renewable energy. They're sort of on the wrong places, really. Also, renewable energy tends to be quite patchy. The sun shines during the day, and not at night. The wind isn't always blowing. Data centers are power hungry throughout the day, throughout the night at all points in time. For a number of reasons, on-site renewable energy just isn't feasible.
What does happen? What actually happens is that data centers take their energy from the grid the same way that all of us do at home. What they actually mean by 100% renewable energy is something called power purchase agreements. The problem that power purchase agreements are trying to solve are that there are renewable energy plants that are providing renewable energy onto the grid. There are cloud providers that would like to just purchase that renewable energy and not use dirty energy from the grid. Unfortunately, the grid is a mixture of renewable energy and fossil fuel energy. It's not like cloud providers can just somehow filter out the electrons that are from the renewable energy plants and just pay for those. What power purchase agreements try and do is address those problems. What a power purchase agreement actually is, it's a contract. It's a contract between the cloud provider and a renewable energy plant. What this contract says is that all of the energy that the renewable energy plant can produce in the next 20 years, the cloud provider agrees to pay for all of that energy and to pay a set price for that. That price is often a little bit more expensive, potentially, than the local price that you would get for taking energy off the grid locally. The way that actually works in terms of the money exchanging hands and how it all gets paid for is different from country to country, and can get a little bit complicated as this diagram shows. Ultimately, that's what they're doing. These power purchase agreements are a contract between the two parties, where it's really a good deal for the renewable energy provider because they get a stable source of income and gets a guarantee that they're going to sell all of their energy at a fair price. Based on this then, the cloud providers would then like to claim that they're running on renewable energy because of that agreement.
Google, since 2017, basically each year, if you add up all the energy contracted for in the power purchase agreements, and then you also add up the amount of energy that they use and take off the grid, the power purchase agreements have exceeded the amount of energy that they've been using from the grid each year since 2017. That's what they mean by we've been using 100% renewable energy since 2017. Amazon and Microsoft also have goals to be doing the same thing by 2025. Power purchase agreements, these are definitely really good things. They're really positive. From the renewable energies' providers point of view, they provide security, they provide less exposure to risky market conditions. Because of that, they attract outside investment. Because of both of those things, they result in the grid becoming greener. This is all really good stuff. From the cloud provider point of view, there's some upsides as well, because, as we're especially aware of recently, fuel prices can be very volatile. It's really useful to be able to have a constant steady source of energy where you've got a price agreed in advance, especially when it's renewable energy, too. That's what these power purchase agreements provide for both parties.
There is actually a limitation to what you can really claim about power purchase agreements. There's a couple of slight wrinkles in this story. One of those is around this concept of additionality. Additionality is a word you'll see a lot if you spend much time reading about climate issues. In the context of power purchase agreements, what additionality means is that to be additional, your power purchase agreement needs to actually result in new green energy being added to the grid. You shouldn't just be laying claim to existing green energy that would have existed anyway, even outside of the context of the power purchase agreement. To make that a little bit clearer, imagine that there's a wind farm that's existed for 15 years. It's been profitable for that whole time, but it has no plans for expansion. It's just happily churning away putting energy into the grid. Imagine that Microsoft comes along and says to this wind farm, we'll pay you for the energy you produce for the next 20 years. We'll give you this nice, fair price for it. Then the wind farm is definitely going to say, yes, please, to that. It's only good things for them.
At the end of the day, after this agreement's been made, nothing's really changed. There's no additional green energy going on to the grid. It would have been there even if Microsoft had not come along and made that agreement. This is the idea of additionality, that in order to make a claim that you're using renewable energy, your power purchase agreement needs to be additional, and be adding new green energy and not just shuffling the numbers around. If you do have a non-additional power purchase agreement like that, then you use a 1-gigawatt hour of energy, it puts 400 tons of CO2 into the atmosphere. Just the fact that you have this power purchase agreement doesn't magically make that 400 tons go away. The 400 tons would still be in the atmosphere, whether or not you made that power purchase agreement or not. This is why additionality is important.
Cloud providers are actually quite good at making their power purchase agreements satisfy additionality, at least to some extent. I think all the major cloud providers that are making power purchase agreements tend to focus on new projects rather than existing projects. They tend to focus on regions that are struggling for green energy investment. They tend to be actually adding new green energy to the grid, which is great. I think this shows it's a bit more of a complex picture that even when you are making a power purchase agreement for a new project, it's not obvious whether or not that new project would have happened anyway, even if you haven't made that power purchase agreement. It's not very easy to assign, should we give you 100% credit for all that renewable energy? Should we give you 50% credit for all that renewable energy, something else? It's not quite as simple as just saying we're 100% renewable.
There's another limitation, a second one, which I think is more evident that it's a problem. This is the hour-by-hour availability of renewable energy. The claim of 100% renewable is that if you just take all of the power purchase agreements for a whole year, then that energy adds up to more than the amount of energy that you've drawn from the grid over a whole year. The problem is that renewable energy is not available at all times of the day. The most obvious example would be, if you're relying on solar power, then there's going to be solar power during the day, but there's going to be no solar power at night. The same is true for wind. The wind isn't blowing at all times of the day. It's true for all types of renewable energy. Data centers have a constant requirement for a lot of energy. What cloud providers tend to do to make this equation add up, they overbuy, in this example, during the day, and they are purchasing nothing at night. Overall, the whole year, that then adds up to their renewable energy claim. The problem with this is that there's many examples of grids and periods of the day on those grids, where there's not even enough renewable energy in the whole grid to even theoretically power one of these data centers. It's clear in this case that the claim of you're running on 100% renewable energy is misleading. There's literally not even enough energy on the grid at certain times of day to satisfy that. In those cases, the data center would have to be drawing energy from fossil fuels, which is resulting in CO2 emissions.
This is actually a graph published by Google, of one of their data centers over a year. The green spiky part of the chart you can see is the amount of carbon-free energy supply that's actually available on the grid at the time. The black flat bar at the bottom is the energy demands of their data center. You can actually see that at most points during the year, there's not enough green energy on the grid to supply that data center. It's quite variable by region. These are figures published by Google. In Iowa, actually, 58% of their energy is from wind power. In their Iowa data center, if you look at their energy consumption hour-by-hour, 97% of the time, they're actually running on renewable energy. Their claim of 100% renewable energy here is pretty close to being true. In Singapore, where the vast majority of their energy is from gas, if you look at an hour-by-hour basis, Google's data center in Singapore, only 3% of the time is it actually running on 100% renewables. That's a long way away from 100% renewable energy claim.
Google and Microsoft actually have targets around this already. They've both said they're aiming for 24/7 carbon-free energy by 2030. 24/7 carbon-free energy here means matching power purchase agreements hour-by-hour, as I was just talking about, rather than matching them annually. This is actually a really big ambitious goal for them, definitely not guaranteed to happen. It'd be really amazing if they actually succeed. It's actually impossible in a lot of areas at the moment. It's going to require some pretty major investments in renewable energy in areas that currently doesn't have any renewable energy, for example. It's also going to require a pretty major overhaul of the way that energy accounting works in many places. Right now, in most places, you can't actually track hour-by-hour what energy is being produced and what the energy mix is on that basis. Yes, hopefully they make it. I think if they do manage to do that, for me, that meets the definition of what could be claimed to be operating on renewable energy. AWS, who's the biggest cloud provider, unfortunately, have not got any targets around this right now. It's unclear as to whether they're going to aim for this at any point or not.
Key Takeaways (How Green Is the Cloud?)
The key takeaway from this section is that even if you're running on the cloud, you still need to care about your software emissions. Power purchase agreements are great. They're helping to transform the grid and make it greener. At the end of the day, your cloud infrastructure use is still causing CO2 emissions.
What Can We All Do Right Now?
With that, what can we actually do right now? When we go back to our companies, what can we do to start to reduce our carbon footprint? There's two main ways you can do this, the first one, fairly obvious, using less energy. Then the second one is around using green energy. We'll start there. What you can do when you get back to the office is you can start trying to figure out which of the workloads that you've got is amenable to moving location. This is more likely to be things like batch jobs, and potentially latency-insensitive applications or services. In particular, you want to be looking out for batch jobs that are using a lot of resources. Maybe you've got a machine learning model you're running. Or maybe you've got a data warehousing job that's transforming a bunch of data and putting it into a data warehouse. These are going to be the most effective things you can move, and also potentially the easiest things for you to move. Then, when you're identifying regions that you want to move to, you could use the Electricity Maps app that I talked about earlier. Also, the cloud providers publish tables of their regions and the grid carbon intensity of their regions, so you can use those. If you're on Google Cloud, they also publish that on how carbon free they are in that 24/7 carbon-free definition I talked about.
The other category of recommendations that you can look at is around using less energy. There's a happy coincidence here, which is that, if it saves cost, it tends to save CO2, so it's a win-win scenario. As a result of that, a lot of the things I'm about to mention, you'll already be doing, or you'll have already considered doing purely on a cost saving basis. Maybe some of them you haven't prioritized yet. The fact that it can make your software greener might be the thing it needs to push them over the edge and actually implement them. First, you want to look into reducing idling resources. Thinking things like, monitor your resource usage. Find instances that are only running at 30% CPU all the time, or only running with 20% memory usage all the time, and redistribute those workloads. Either reduce the size of those instances, or put more things on those instances so they're actually utilizing more of their potential. You should also be looking at autoscaling, which is a pretty major thing you can do to reduce idling resources. If you're on AWS, you can look into using burst instances. You might decide that now's a good time to start looking into serverless because serverless infrastructure on average uses a much higher percentage utilization of the infrastructure that it's running on than your typical self-run infrastructure. Another one is, if you're running Cron jobs every day that take a lot of resources, don't run them all at the same time, so that you need a massive instance to run them on. Try spreading them out a bit, and so you can use a smaller instance. Also, when you're not running those tasks, turn those instances off. These are all things you can consider.
There's a few recommendations around network traffic. Reducing network traffic is good. Utilize browser and edge caching where you can. You're probably already doing this, but maybe it's not very optimized, and so you can consider how to optimize it to get a better hit rate. Using edge services is an interesting idea. You can use CloudFront Functions, or Fastly have an edge service called Compute@Edge where you might be able to do more of your processing at the edge, so that it involves less network traffic and doing that at your data center instead. Compression is another one. Again, research has shown that compressing a file, sending it over the network, and then decompressing it is less energy intensive than just sending the whole file over the network. Look into compressing things where you can.
There's couple of energy efficiency recommendations here. You can look at using more energy efficient processors. AWS in particular have Graviton 2 processors and Graviton 3 processors you can look into. Graviton 2 processors are two to three times more energy efficient than the equivalent other processes that they have on offer. Where you've got the option, consider using a more appropriate language for CPU intensive workloads. Some interesting research was done to look into the energy efficiency of various programming languages.
Unsurprisingly, perhaps the compiled languages are much more efficient than interpreted languages. Maybe it's a bit more surprising just how less efficiently interpreted languages are with Ruby, Python, and Perl coming in at around 70 to 80 times less efficient than C. You're clearly not going to go away and just rewrite your entire stack in C, that wouldn't be a very sensible thing to do. Maybe you've got two teams, you've got a Python team and a Java team, and you're looking to write a new data warehouse data processing job or something like that. Then maybe you give that to the Java team rather than the Python team based on this. That's another option.
Finally, with storage, consider using long-term storage where you can. For example, use AWS Glacier if you've got files that you don't access very often. You can potentially also automate moving files into the most appropriate storage type, energy efficiency-wise. In terms of data processing, and machine learning, don't always train your models for maximum accuracy. Sometimes you can spend twice as much time and energy getting a 10% increase in accuracy in your model. What you should do is only train your model as much as you need to, for it to operate within predefined accuracy that you require. Don't just always go for optimal accuracy. Along similar lines, only retrain your model when necessary, don't just automatically run it every single week or every single month if you don't need to. Only rerun it when you actually need to.
Summary (Things to Do)
The summary here really is, the easiest thing you can do is moving to a greener location where possible. The golden rule, if it's saving cost, it's saving carbon dioxide. Start making some cost savings. That is the simplest place to start.
Tooling to Measure Your Impact
There's a bunch of tooling out there that can help you with all of this as well. All the major cloud providers have their own calculators. This is the AWS one. It does all the things you'd expect it to do. It tells you what your carbon emissions are by month and by service. It helps you track that over time. This is super easy to get set up with. You automatically have access to this if you're an AWS customer. There's no setup required. Google have their own equivalent of this. Azure also has its own one as well. No matter which provider you're with, you should have access to some way of measuring your emissions. If you want to get a bit more granular and track something that's a bit more concrete, a good proxy metric, then if you're on AWS, you can use Cost Explorer to track, for example, your vCPU hours every month, and try and optimize that. Or you can track memory usage, or storage delivery and try and optimize those.
Some limitations of the calculators, you have to be a bit careful about exactly what they're telling you. For example, the AWS one, it doesn't give you your embodied emissions, which is a bit strange. Hopefully they'll add that at some point. Then, I think another quite big limitation of both the AWS one and the Azure one is that they subtract all of the power purchase agreements that they've got. They don't give you your gross emissions at all. For AWS in particular, if you're in a region that is covered 100% by power purchase agreements, because they're not giving you your embodied emissions either, you can end up with a footprint that's very little or even zero. That's of limited use. I hope that they'll add gross emissions at some point as well. That's just something to be aware of.
If you find those calculators too limiting for you, and you want to know what your gross emissions are, there's a tool by Thoughtworks that they've published and open sourced called Cloud Carbon Footprint. This plugs into all of the three major cloud providers. It tells you your gross emissions. It estimates them. This is a useful tool, again, if the cloud providers calculators are too limited. If they are too limited, then just talk to your cloud provider about it. The only way that they know that their customers care about this and they need to make improvements is by them being told. Definitely talk to your cloud provider, tell them what you'd like to see in the calculator.
Getting Started Plan
With all that said, if you take away one thing from this talk, it's a simple, three steps you can do to get started when you get back to your company afterwards. The first one is, just get started, set up some monitoring. If you're in the cloud, plug in to whatever the calculator is from your cloud provider. If you're not in the cloud, maybe start measuring your energy usage, or just set up something that helps you monitor your carbon footprint. The second one is moving workloads to low emission regions where possible. This tends to be the very simplest thing that you can start doing. Number three is, start making energy and cost reductions.
See more presentations with transcripts
