Delivering sustainable technology is becoming mandatory for organisations across the planet, from both a regulatory and moral standpoint. We have written a lot on how using Well-Architected and Serverless First principles is the best way to build a secure, cost effective, reliable and performant applications. More importantly, we firmly believe that a Serverless first approach is the sustainable approach for Long Term Value.
We have been thinking about this topic for quite some time and we believe that these 5 steps will get you moving in the right direction:
- Measure your carbon footprint.
- Move to Public Cloud and close your Data Centers.
- Run your workloads in a low carbon region where possible (keeping in mind data location regulations, latency requirements and service availability).
- Eliminate waste by understanding and meeting your users needs in the simplest way possible.
- Adopt a Serverless first mindset and approach.
Sustainable technology guidelines
In addition to our 5 steps the UK Government’s Central Digital & Data Office recently published guidance on how to make your technology sustainable which contains awesome guidance for any businesses or government agencies. They have 12 questions that are important to consider when aligning with sustainability goals.
At the start of your project, you should consider these questions:
- What are your organisation’s sustainability goals?
- If the contract is more than £5 million per year, has the supplier committed to meet the government’s net zero target, and published a Carbon Reduction Plan? I would update this to: has your supplier committed to meet a net zero target, and published a Carbon Reduction Plan?
- Can you include specific project objectives to meet your organisation’s sustainability goals?
- Have you identified potential benefits for meeting sustainability objectives, or risks that would stop you meeting those objectives?
- Does your organisation have processes for recording and reporting on sustainability goals? For example, reporting on the targets for greenhouse gases, waste and water.
- Do your project plans include user research to more clearly define requirements and reduce the chance of buying software and hardware you do not need?
- Do you have a process or plan for recording the impact of future upgrades to software and hardware?
- Are you able to recycle or repurpose any equipment you are replacing?
- Are you able to use existing datasets for your project?
- Are there any opportunities for minimising processing, transmission and storage?
- Can you put in place processes which reduce printing and paper trails in back office systems and user facing services?
- Have you assessed whether home working is a practical and more sustainable option for your project team?
The whole document is worth a look, but I would draw your attention to the steps to reduce greenhouse gas emissions.
Without an accurate Carbon Score capability from your cloud provider, we feel using your Cloud Bill as a proxy can be a useful indicator of your Carbon usage.
Lower cloud bill = less carbon used
To back this up, AWS released a report by 451 Research that EU businesses that move to AWS Cloud can improve energy efficiency and reduce carbon emissions that states that “Businesses in Europe can reduce energy use by nearly 80% when they run their applications on the AWS Cloud instead of operating their own data centers”.
As well as the advantages of improved time to value, security, cost, performance, reliability, and increased pace of innovation, moving to Public Cloud is now the right thing to do from a sustainability point of view as well.
As ESG and reducing carbon footprints are becoming a Board/C-Suite priority, Cloud providers that can provide rapid and accurate answers to questions on their workloads will be at a significant advantage.
- What is our Carbon footprint?
- How can I reduce our Carbon footprint?
Microsoft Azure have launched the Sustainability Calculator which helps their customers to understand the carbon footprint of your Azure cloud resources.
The pressure is now on AWS to provide something similar.
Update 2nd December : AWS announced a Customer Carbon Footprint Tool that is coming soon ! :
With AWS re:Invent on next week, we are hoping for some significant announcements to help teams deliver not only a Well Architected Serverless solution, but a sustainable solution. Check out our thoughts on Serverless Craic Video, Podcast and Transcript below.
There is a Sustainability Track this year and this attendee guide has some great sessions that we will be keeping a keen eye on.
Serverless for Sustainability
In the same way that for certain workloads at a certain scale, a serverless approach may not be the cheapest, the same will apply for carbon score. But for the vast majority (and growing number) of use-cases, the serverless option will prove to be the cheapest and most sustainable.
We believe a Serverless first approach is the sustainable approach. With these emerging Carbon Score calculators we will now be able to prove it as well.
Serverless Craic Episode 4 Transcript
We are going to talk today about sustainability. We’ve been doing a lot of reading and writing about sustainability and cloud compute. And what certainly sparked this one is Simon Wardley’s tweet this past couple of days. Plus ‘COP 26’ is in the news and it is just finishing. In his tweet, Simon says he hopes AWS bring out carbon cost per lambda execution. So today when a lambda executes there’s a cost per execution. But it would be nice if they also gave the carbon cost. Mark, do you think they’ll do this? Obviously AWS re:Invent 2021 is on shortly.
I’m hoping more than expecting. I have no insider knowledge or anything like that. I think it would be very good timing as being able to complete the carbon footprint for a particular workload is especially important with the renewed interest in sustainability, saving the planet and environmental concerns on the back of COP 26.
It’s one of the big wins that all Cloud Providers can easily put out there. We’ve seen the Microsoft Azure calculator and GCP have their carbon footprint capabilities as well. So two of the big cloud providers are up on their game in this regard. So I am hoping AWS do something similar.
Cost as a proxy for carbon footprint
We’ve used cost as a proxy for carbon footprint up to now, but that’s a big assumption. There’s lots more to it than just that. With our serverless first approach, sustainability is not for free, but you’re getting it as a side effect. So I’m hoping we get carbon emitted per function in the same way that we get our our cost per function on your bills. In my mind, I have beautiful dashboards and reports that we can talk regularly to our teams about. But let’s see what happens.
I have a good question for you, Mark. Because I know you’re the sustainability guy, right? If you look at the Amazon set up in the EU/Dublin, that’s all green energy. So why would would lamda be more efficient if they were running containers or a k hits cluster running on EC2?
There’s an architectural answer to that: it depends on your workload.
How green is your data centre?
There’s an illustration there that’s important. How green is the data centre?, For example Finland ‘s sustainability is very high and that is different for each region. There’s four different basic models for compute:
- buying a computer;
- using an instance in the cloud;
- using a container in the cloud
- and using serverless.
To explain you go to PC World because you have a terabyte of video that you need to convert from mp3/mp4 to a different format. You drive to PC World, buy a box, come home and plug it in and run that workload. That’s the worst way to do it.
Secondly, you configure an EC2 instance on AWS, and you give it that workload. So there’s time, memory, CPU, and storage with your EC2, which is just a file system connected to it.
Then there’s a cloud native way which is a container using S3 for storage. And then there’s the Serverless way just using the native services. And it stands to reason that the fourth way has less carbon burn than the first way.
The first way is easily the worst as you’re driving to the shop and physically putting something into your house. It’s too hard for the top cloud providers to give you that exact comparison.
Be wary of Jevons Paradox
As it’s all ephemeral in the serverless world, there are shared resources and servers in the cloud with multiple customers/tenants on that hardware, so it’s very hard to do that. But, sustainability is all about reuse and only using what you need. You’re reusing that hardware and it’s being reused across many 1000s of customers. So there’s effectiveness and efficiency underlying that compute resource that you’re not going to get if you buy dedicated hardware to run your workload because it sits idle if you’re not using it.
You have got to be wary of Jevons paradox, the more efficient you make something, the more you use it, and then that takes over anyway. And you end up burning the same amount of carbon.
We want to maximise the amount of work not done, we want to truly understand what it is we’re trying to accomplish, and then use that serverless first mindset to only use/leverage what we absolutely need to meet that need. That’s a very carbon effective/efficient way of solving a particular problem. We’re not going to build a big platform. We’re not going to build a big ‘deathstar’ to solve a small issue.
That’s the interesting bit. There are value streams and carbon scores at each point and you have got to work out the equation for the most effective approach to dealing with your carbon spend. The first step is to get to the cloud and the second step is to think about your energy utilisation.
And I think you’re right. The first step is moving to the cloud, and in general, close down your data centres.
AWS is 80% more efficient than your data centre
There was a report from AWS this week, that AWS is around 80% more efficient than running your own data centre
This is really interesting and people never understand this when you explain it to them. When you move from a box sitting in your house to a box you rent from Amazon, you know they’re going to run a box much more efficiently. But then you are renting that 24/7. Are you going to take 100% of utilisation? The answer is no.
I describe that as legacy cloud. You’ve got a box in the cloud, but it’s legacy cloud because you don’t have elasticity in that box. It might running on 10%, so you’ve 90% of wasted resource sitting there, but you still need to power it.
AWS call this modern applications, and that includes containers and serverless. By the way serverless actually has servers and containers, so they are both the same thing. So if you do modern applications, you are renting a box, but you’re letting AWS make sure that it’s maximised. There’s still a physical computer sitting in the cloud data centre, but you’re letting Amazon, make sure it’s effective.
There’s a jump up in efficiency moving to the cloud, and there’s an extra massive jump up from moving to modern architecture. And cost is the way you can measure that. But then, cost savings plans can also mask that sometimes in reserved instances. So there’s a nuance and that’s one of the reasons why it’d be great to see Amazon come out with a carbon score for your workload. And it’s easy for legacy cloud, because it’s a box. It’s harder for modern applications.
Well architected and serverless leads to ease of adoption
We’ve been talking about well architected, serverless first for quite a while now. And we feel it is the best way to deliver that secure, cost effective performance and a reliable solution. If teams have embraced that, have a really good handle on it and have delivered solutions in a serverless first way, then whatever comes out from the cloud providers for carbon, they’ll be able to adopt that within minutes or hours.
We’re seeing this emerge from COP 26 and ESG boardroom/C Suite level mandates for all organisations large and small. They’re going to need to be able to answer the question: what’s your carbon footprint? How sustainable is your business? What’s the environmental impact for whatever it is you’re trying to do? And I think teams who have embraced serverless and are in the cloud will be at an advantage because they’ll be able to articulate very clearly on what the solution is delivering eg. here’s the carbon score, whenever that becomes available. Google and Azure have this already in place.
That’s a very powerful argument to have for their business partners. To be able to say, ‘yes, we’re doing okay’, or ‘no, we’re not doing okay, we need to do better’, gives the visibility or transparency, that will help drive the right sort of conversations. We all driving towards the same sustainable goals. We all feel this is the right thing to do, for various reasons.
If they struggle with being able to measure it, do you think it would be a good extension to the well architected framework to provide that guidance on running your workloads as effectively/efficiently as possible? And they probably do do that but they don’t frame it in relation to through your carbon footprint.
It’s a combination between performance and cost. It’s like a performance test from years ago, when you could see the headroom of application and how efficient it was. There’s a couple of factors you have to bake in. It’s an attribute of well architected and it is a sign of good architecture to have a sustainable system, But (at the moment) it’s too hard to measure.
SCORP becomes SCORPS to include Sustainability
Mike, you might need to update your blog posts on SCORP to be SCORPS!
Security, cost of access, performance, reliability, and Sustainability? We did a map a few weeks ago. We talked about your company and your carbon burn. There’s stuff around employees and keeping an office open. So obviously, remote working, and being responsible about employees and staff is good. One trip to the states and back means you’re burning a whole lot of carbon. So there’s employee burn, then there’s physical supply chain.
If you’re selling a physical product, there’s ‘the how’ and things you physically bring in. If you’re a software company, you don’t have that but there’s a software supply chain. Are you getting sustainable compute or are you building in an unsustainable way? That is still very cutting edge.
If you take an holistic view of the value stream for an organisation/business you can talk about things like your developer IDE: do you need to have high powered m1 macbook pro’s shipped to all your developers when there are compelling cloud IDEs available. So bare earth minerals that go into making the laptops, the resources and the carbon footprint of those high powered machines, is that an expense worth paying, when there are very compelling options with GitHub code spaces, Cloud9 and GitLab out there. These things are in our purview.
When you look at a business as a whole, their tech footprint might be quite small, but we can show what you should be able to do and what the future should look like. We can tell you what your tech stack costs are and the environmental or carbon footprint. But what about the rest of the business? Tech may not be a big part of the overall footprint for a large enterprise, but at least you can start driving things the right direction. I can showcase what is possible.
It’s a bit like electric cars being more expensive early on. It was more expensive to be eco friendly. But with this stuff, it’s cheaper to be eco friendly. I don’t understand why people haven’t clicked onto the fact that it’s cheaper to build in a more sustainable way.
Providing sustainable solutions is a competitive advantage
From a competitive perspective, if you’re a service provider, consultancy company, or building solutions for companies and you can offer solutions that are low carbon footprint and sustainable, that’s a competitive advantage going forward. Because that’s going to be a big objective or mandated for all businesses. Standing up a load of containers and burning a load of energy/carbon isn’t going to cut it. Maybe I’m being optimistic here, but I think the push for better sustainability will drive the need for more serverless first approaches.
The other thing is government standards coming in, which puts pressure on companies.
I was delighted to see the UK government announce their sustainability guidelines. So the Central Digital and Data Office just published it yesterday. And if you haven’t read it, there’s loads of great stuff like 12 points on your organization’s sustainability goals, and some really interesting ones about reducing your greenhouse gas emissions.
So use public cloud, optimise your cloud use by managing your cloud spending, but also use less energy, audit your data, which is a really cracking one to call out. We used to have the whole line about ‘data is the new oil’. But oil can be quite toxic and harmful to the environment. So you need to treat the data the same way. Only store it if it’s absolutely necessary and you are getting value from it.
As Adrian Cockcroft says: if you collect a piece of data and you don’t put it in a model, just delete it because it’s out of date.
Audit your live environment and switch off unnecessary environments where possible. We’ve seen that a lot in our in our time, right? There’s environments that sit and burn all day and all night long, and nobody’s using them. And decommission unnecessary systems your project might replace. So how many projects have we been on where we built the new thing, but the old thing still remains?
Lots of goodness there.
We’ll put the link in the show notes. It’s a cracking read and it’s very well written. And it’s a great start. If the UK Government are doing this, then what’s stopping your organisation?
And fingers crossed we move forward on this. All right, very good. So that’s the craic! We’ll keep writing about this on the blog: theserverlessedge.com. Follow us on Twitter @ServerlessEdge and we’ll have some more in the book in November 2022.