We share our thoughts on the AWS Sustainability Pillar that forms part of the Well Architected Framework. This is the sixth and final part of a series of talks.
The AWS Sustainability Pillar is a new pillar that was announced at AWS re:Invent. What’s nice about sustainability is that it rolls up a lot of good practices and it’s a very simple measure.
AWS Sustainability Pillar on our Well Architected Series
Dave Anderson
We’re continuing our conversation on the well architected pillars.
Today we’re talking about the AWS Sustainability Pillar. We’ve talked quite a lot about sustainability. It’s a new pillar that was announced at AWS re:Invent. So we’ve talked about it a fair amount, because (1) it’s new and exciting and (2) because we’ve been following this for the past two years and it’s a brilliant addition. What’s nice about AWS sustainability is that it rolls up a lot of good practices and it’s a very simple measure.
It’s very hard to measure carbon, but very simple to understand when someone, like AWS measures for you. This pillar is slightly different, it doesn’t have all the same kind of questions. Maybe they might change it eventually. But it’s more like a list of best practices that are broken down into sections. So let’s list them out:
- Region Selection,
- User Behaviour Patterns,
- Software and Architecture Patterns,
- Data Patterns,
- Hardware Patterns,
- and Development and Deployment Process.
There’s a bunch of questions within those. Let’s fly through them.
Region Selection
Mark McCann
Some regions are supplied with greener energy than others. Some regions are using non sustainable resources, depending on where you are in the world. If you don’t have massive latency requirements, or a real need for super fast, low latency, then you’re probably best putting it into a more sustainable region. US 1 and Ireland are sustainable regions. There’s others across the AWS ecosystem. Google and Azure have the same. So where you put your workloads can have a sustainability impact. All other things being equal, go for the greener one. Go for the most sustainable region for your workload.
Dave Anderson
Some regions are not sustainable, and some are very sustainable. There’s a nice phrase that Werner Vogels used during the launch that Adrian Cockcroft also uses: ‘the cloud providers look after the sustainability of the cloud, they’ll make the data centres as sustainable as possible. It’s our responsibility to look after sustainability in the cloud.’. So they’ll have a sustainable datacenter, we have to design sustainable workloads. So I thought that was a nice kind of split. With region selection, if they say a certain region is green, you should try and put workload there.
The difficulty of sustainability and your own data centres
Mark McCann
With a move to the cloud, if you’re on your own data centres, you’re going to have a hard time trying to be as sustainable or green as the cloud providers, because they’re investing hundreds of millions or billions probably across the globe to sustainably power, cool and provide all the resources they need for the for the data centre.
Mike O’Reilly
What’s going through my head is the shared responsibility model. A lot of what the white paper gets into for that pillar is understanding the separation of responsibility. It’s good to touch upon those as we run through it. It’s major that cloud providers are going to keep their side of it as green as possible. And also coaching you, as an architect or business owner, on how to make good decisions in relation to what you’re doing in your space and what you’re building on the cloud and the various sustainable approaches.
Dave Anderson
This is the whole premise of the cloud. An organisation like Amazon, Microsoft or Google will do data centres better than you. If you look at User Behaviour Patterns and you use their assets in an elastic way using the latest technology, it will be more sustainable, efficient and cheaper on modern cloud. If you go down the legacy cloud route, and treat the public cloud like a data centre, then you’re not going to be very sustainable, you’re not going to be very cheap, and you’re not going to be very efficient. So it’s that elasticity thing that I know was popular 10 years ago when the cloud started, but people seem to have forgotten about it.
User Behaviour Patterns
Mark McCann
Under the User Behaviour Patterns section it’s about how do you align your SLAs with your sustainability goals. That screams out at me, especially with the COP events. With the requirement for boardroom level mandates these concerns are going to be more prevalent and visible to all teams, up and down the hierarchy. You’re going to be asked about how green your solution is. How green is your product? How green is your business? Making sure workloads and solutions on the cloud are aligned to SLAs is what we’re going to be concerned with over the next number of years.
Dave Anderson
It’s no different to when you buy a soft drink. You’ve got to know if it’s in sustainable packaging. It’s a big issue. In the future, when you use a digital service, you’ll want to use something sustainable which is effectively cloud.
Software Architecture Patterns
The next section is Software Architecture Patterns. I think this one is interesting. It’s about keeping your code base and your architecture really efficient such as refactoring optimization and more effective data access.
Mike O’Reilly
It’s good practice as it ties back into efficient design. When you work in enterprise spaces, you do question the value of older business products that are running in the background. You’ve got to constantly assess if this is worth the compute? Is this worth the cycle times? Are we really getting value for money? Now you’ve got to factor in sustainability. Is it racking up your carbon footprint? It’s another factor to consider. Should we invest time in reducing the amount of maintenance? How efficiently or inefficiently is something running? This is definitely an interesting area for people working in large organisations where they’re supporting lots of different apps and different workloads.
Impact on devices
Mark McCann
There’s an interesting question on optimising the impact on customer devices and equipment. If you have a really inefficient client side app with a lot of data processing on the device that it doesn’t need, you could sustain the lifetime of that device by having a more effective and efficient client side app or web app or mobile app. There are lots of things we can concern ourselves with.
Dave Anderson
If you’re on a mobile app, do you need to send the biggest image possible over the network? And make the device or optimise it accordingly? Or can you optimise it ahead of time? There’s a bunch of stuff around not being as fast as you need to be. Another thing about architecture is that in the olden days, there were constraints. So you had to do proper architecture. I would imagine that there’s some modern apps with massive amounts of scale. As it scales, the architecture is not ready for it and they are relying on the network. You’ve got modern technology, but it’s not optimised.
Mark McCann
The question is ‘how do you use software patterns and architectures that best support data access and story patterns?’. If you’re following a serverless first approach, you’re well on the road to being sustainable.
Think in terms of limited resources
Mike O’Reilly
Bad design has an impact on users and their devices. They could be using unsustainable resources to charge their machines or phones. I always think about IDEs and the bigger IDEs in terms of auto completion and indexing because they get very warm very quickly! I’ve seen a lot of IDEs moving onto the cloud: Cloud9 and VS code. And you’re thinking that all of that should be done in the cloud too. Right now in the UK, energy prices are going through the roof. My energy bill has doubled in the last month. So if I can do anything to reduce that, it helps me as well.
Dave Anderson
I wonder if it has come full circle. I’m old enough to remember the olden days when CPU, memory, disk and network were limited. You had to design and take into consideration your poor resources. I’s going back that way again. We should be thinking of limited resources, so you’re not using what you don’t need.
Mark McCann
A very thin client for all your needs. And everything is done in the cloud, server side, from your IDE to everything else.
Photo by Casey Horner on Unsplash.com
Data Patterns
Dave Anderson
So the next section is Data Patterns. It’s a huge one. There’s an awful lot of waste with data flying around the internet. Again, we quote, Adrian Cockcroft: ‘If you collect a piece of data, and don’t put it directly into a model, you should just delete it.’. Because you don’t need it! It is a bit extreme, but it’s a great idea.
Mark McCann
There’s a lot of good practice here. Data can be quite toxic for various reasons from privacy breaches and security points of view. You should have a good handle on this. Your data classification is critical. If you don’t extract value from it, get rid of it or it’s going to be unsustainable.
Mike O’Reilly
This is what I’m interested in. I’d love to learn more, dig into it and just see how it unfolds. Everything’s becoming more data centric, and the amount of compute that goes into chomping data is 90% of what IT. I’d love to see how much electricity or energy is used on processing data. I am keen to see how organisations approach this one.
Mark McCann
How do you minimise data movement across networks? That’s huge, right? Only move what you need for when you need it and not just move everything for the craic!
Dave Anderson
I wonder when they will start egress charges when you’re going out to the cloud? I wonder if they will start doing more of that. That’s a big one.
Hardware Patterns
The next one is Hardware Patterns, which is right sizing our stuff correctly. We’ve all been in teams where the question asked is: ‘what size box do you need?’ And the answer back is: ‘the biggest one humanly possible!’. It’s a natural reaction but you don’t need that.
Mark McCann
This is where a serverless first mindset and approach really kicks up a gear. You don’t even have to concern yourself with a lot of these questions. It automatically scales up and down appropriately. We don’t have to worry about picking hardware or incident sizes ahead of time.
Dave Anderson
A good one is the Lambda Power Tuner. It will help you pick the optimum hardware size. And sometimes it’s not the biggest that is the most efficient.
Mark McCann
With Graviton chips coming in, immediately, you are on a more sustainable compute platform without having to do too much.
Development and Deployment Process
Dave Anderson
The last section of the AWS Sustainability Pillar is Development and Deployment Process. How do you increase utilisation of build environments? Build everything all the time!
Mark McCann
We see this quite a lot where environments sprawl and asset’s sprawl for no real benefit. So again, it’s all about being smart about how you set up your clients, how you set up your pipelines and how you set up your environments to make sure they’re actually delivering value. And they’re not just there because that’s the way we have always done it.
Dave Anderson
And people running huge test suites that go on for days. And publishing results that no one ever looks at.
Avoid the sprawl
Mark McCann
The question here is: ‘how do you adopt methods that can rapidly reduce or introduce sustainability improvements?’. If you’re on a serverless spectrum, the cloud providers are working for you and they’re introducing new capabilities. Graviton is a great example. It’s lower cost, more powerful, better performing, but it’s also more sustainable. If you’re on a serverless way of delivering your solution, you can just take advantage of that instantaneously.
Mike O’Reilly
You need to make sure your design and architecture can move with those innovations. That’s a large part of the spirit of that, like Graviton and different chipsets and making the shift towards getting advantage of that.
Dave Anderson
So that’s the AWS Sustainability pillar.
That’s the craic. Next time we’ll be looking at all six pillars to see which one we like the best. Thanks for listening. There is more on TheServerlessEdge.com, our YouTube channel, Twitter @ServerlessEdge and Podcast. Thanks very much!
Transcribed by https://otter.ai
Top comments (0)