Climate activist Mary Annaïse Heglar wrote in The Rolling Stone: “Dealing with one crisis at a time was a luxury that is now over, and climate change has become the ‘threat multiplier’ we knew it would be.”
Yes, we are still full throttle in a pandemic. Yes, we are in the midst of multiple economic and social crises. But the climate crisis proliferates everything. It is simply too urgent to be ignored by any of us.
The tech industry has its part to play. An essential role even. It’s not just that we shouldn’t be building the future while also destroying it. Tech touches just about everyone and everything. We have a responsibility to lead the charge. And anyway we have to live and operate in this world, too.
With this in mind, we kick off 2021 by offering you simple ways to make a huge impact, even as a single developer in a massive company. Heck, even as a consumer that uses videoconferencing or streaming.
Bonus, it’ll probably save you money.
Why Should You Care? It’s So Out of Our Lowly Control
For this piece, we held two separate Zoom interviews. First, with Dr. Holly Cummins, innovation leader at IBM. Then with both Anne Currie, tech ethicist and science fiction author, and Paul Johnston, interim CTO and tech environmental consultant. Each conversation started with asking: Why? Why should we care? Don’t we have enough to worry about in 2021?
There’s no doubt wrapping our heads around the immensity and urgency of the climate crisis has become a challenge unto itself. Johnston said these weather extremes and breakdowns in biodiversity are such massive problems that it’s nearly impossible to see how we as individuals can make any change.
He said, “It’s the kind of thing you can’t physically understand and mentally understand. So complex and big and difficult. You can ignore it, but it still affects you.”
“This matters because nobody has done enough yet and we will not do enough, so everything that we do and we can do matters.” — Paul Johnston, Roundabout Labs
We tend to distance ourselves. But that thing that seemed a far-off idea when we were growing up is already here and it’s getting worse year after year.
Currie quoted fellow science fiction author William Gibson: “The future is already here, it’s just not very evenly distributed.” The climate crisis is already happening in Africa — and the United States — with droughts, plagues of locust, hurricanes and floods. We are on the cusp of mass migration with environmental refugees. And long-term exposure to air pollution is linked to a 15% increased risk of dying from COVID-19. Of course, these are just a few of a plethora of long and short-term consequences.
But that’s all caused by big industries like oil, transport, manufacturing, and agriculture, right?
There is admittedly still a lack of agreement around precise measurement of carbon footprints, but data centers are pegged as contributing between 1% and 3% of all carbon usage. If you include also the indirect secondary and tertiary emissions, from raw materials, devices and delivery, from data storage and processing to machine learning models, the IT and communications industry is at about ten to 12% of all emissions. And that was last counted by Greenpeace in 2017, so it could be higher.
The first step is getting the tech industry to understand its own responsibility. Currie says the basis of morality is to do no harm. So even if you measure the tech industry down at 1% mark, that’s still a lot to sort out.
She says it starts with “Just getting everybody used to the fact that you have responsibility for the world. Then, fundamentally, we are an industry that exists in the world and the world has a problem to be sorted and we are in a very good position to sort it and we should. It’s a group effort and tech is very powerful group.”
Cummins says we have two reasons to care. First to save the world. But, she says, even if you don’t care about the planet, you probably care about money — “all that energy is our company’s money that we should also be trying to economize.”
Add to these significant reasons, while they are mostly slow to start, governments are on a track to require all organizations of a certain size to report their carbon footprint, including the U.K. in less than four years.
Bonus, companies — with the notable exception of Amazon — that are perceived as being greener are more successful. There’s a lot to be said for consumer power — especially developer consumer power.
As Currie said, the tech industry doesn’t need to fix the world, but “they at least need to fix their own stuff and then go a bit further because they need a stable world to operate in.”
How Do You Solve a World-Sized Problem?
For a problem that’s nearly inconceivable, solving it is magnitudes harder.
Currie says the power is with the consumer. She says if you do one thing to help curb global heating this year, let it be to ask questions. Make a list of any paid-for IT services. Start by saying to your customer support and sales reps that you care about the climate crisis.
Then specifically ask any cloud service or tool you pay for — or your company pays for — where they host and with whom. They may not know, but flagging that question is an important step.
“I think because it [climate crisis] is such a huge issue, it’s so easy to say ’It’s so big so nothing I can do will make a difference so let’s all do nothing’. Our industry is one that consumes a lot of energy so small actions that we take can actually have a big impact.” — Holly Cummins, IBM
Currie says we have to remember: “What my supplier emits is what I emit.”
Then, the bigger the contracts, the more important the customer you are, the more questions you can ask:
- What are your plans?
- What is your corporate statement on this?
- What are your milestones?
- How do you want to address this?
Try to get past the cacophony of green-washing press releases and demand your suppliers to stop causing problems and then to start fixing them.
For example, Amazon Web Services — by far the biggest provider of cloud services — has been oblique at best. At last December’s Re:Invent, AWS made some big promises of using more than 6.5 gigawatts of renewable energy. First, you need to understand what a gigawatt is. Then you’d need to know how many gigawatts AWS uses in general — is that a small or large step toward their stated goal of 100% renewable energy? Especially with AWS, we just don’t know enough.
The consensus seems to be to give AWS a high five for a good start and then push them to do more. Much more.
Then, once you’ve begun addressing vendors, you can start — or join — the conversation within your own organization. Johnston adds more questions to ask including: Who is the point person for making us more sustainable? Who is it that I talk to to be part of the solution?
When I talk about sharing Earth’s resources people call me a socialist. When I talk about Kubernetes and sharing compute resources I’m called a thought leader.
— Kelsey Hightower (@kelseyhightower) December 26, 2020
From there, the organization needs to make a strategy. You start by measuring and trying to understand the data. How many emissions do you have? What is your impact on the world? And, finally, how can that be changed?
This year will hopefully see the fulfillment of the demand for tooling that measures climate impact, and hopefully, ones that automate lessening carbon impact at scale. Argos is a new open source tool released last November that seems so far unique in its ability to create and measure the technical metrics of an organization or product’s carbon footprint.
Once you know your baseline, you can look at ways to reduce your impact. A simple but dramatic way is turning off auto-play of any videos or media on your website.
Cummins used to work on the Java Virtual Machine (JVM). She says if she made just a half a percent performance improvement with JVM, it was at the base of so many stacks that it would improve both the user experience and use significantly less energy.
She says for any product being used widely, a small change can make a big environmental impact.
Energy Use as a Proxy
When looking for a way to measure your carbon output, energy makes for a useful proxy. And the easiest direct measure of that energy is the number of servers you are using. Most organizations will find a lot of waste and even a zombie servers (those with no current useful workload that are nonetheless still running) with this examination.
For instance, many companies pursue single tenancy even when it’s unnecessary, which means the cost of excess servers and excess energy.
Cummins said, “All of us provision servers to one team because talking is hard.”
She points to Conway’s Law as often causing us to have more instances than we need. Conway’s Law states that:
Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization’s communication structure.
This leads to a lot of waste, with excess servers or virtual machines at each department level.
Money-saving multitenancy isn’t the easiest thing to do, but a lot can be done with a culture of breaking down silos and increasing communication.
Cummins said that single tenancy is often more a cost accounting than compliance measure, with organizations wanting to make sure costs flow to the right place. By redirecting these services to be multitenant, it reduces costs but it makes it harder to track. FinOps is an emerging trend that looks to ease this workflow.
Cummins says that you could start by sharing administrative rights in a cluster.
Of course, you don’t want everything to be multitenant. Communication is key because, as Cummins said, if two services are sharing and one uses 100% CPU, it will hurt the other. And there are security concerns. You probably don’t want a multitenant production environment, but the multitenant staging, testing and development usually work well.
Similarly, there’s a ton of unnecessary waste — which costs both money and the environment.
“We have so many servers, all of us, that we have set up and forgotten to take them down. Sometimes we’re not really sure they are being used — you do need tooling to monitor inbound and outbound traffic. Most of the time we’ve just forgotten them,” Cummins said.
Some organizations know that no one is going to use their services overnight so they turn off internal services from 5 p.m. to 9 a.m. This of course doesn’t work for distributed worldwide organizations but is a good solution for some that also enforces work-life balance.
Another change may be going from lots of servers to serverless computing.
mostly I’m deep in rest mode but I still love thinking about how some of the stuff we’re building explicitly intersects with environmental factors, the opportunity ahead to have impact is, in my engineering professional opinion, super cool ?
— Miss Amy? (@MissAmyTobey) January 1, 2021
Johnston said, “Serverless is efficient, it tends to use fewer resources in terms of maintenance in the long term. It’s just better in terms of efficiency of using the cloud resources instead of your own resources. You’re pushing your carbon emissions to someone who is actually managing them better than you can. You are using others’ expertise.”
Since everything with measuring emissions is really challenging, Currie recommends to “Just get a supplier to do it for you. Get a supplier that is both reducing the emissions and measuring it. Each cloud company measures it for you.”
Then decide on a change for the better.
“If people go onto Amazon because it’s greener than their dirty data center,” Currie says,” that’s “better than making no decisions.”
Simple Ways to Dramatically Cut Your Carbon Footprint — And Cloud Costs
A simple proxy for your energy usage in the cloud is your cloud computing cost. If you reduce your dollars spent, you’re probably reducing your electric usage.
Jeff Smith, director of product operations at Centro, spoke at DevOpsDays 2020 about how the company cut its cloud cost over the last year. Centro cut it by 37% by just shutting down instances from 7 a.m. to 7 p.m. and over weekends. Smith did warn that if you do this, you need to carefully coordinate how your dependencies go up and down.
Centro also uses Amazon EC2 Spot Instances — Amazon selling its excess compute at a discount rate of marketed as up to 90%. Again Centro saw, on minimum, a 37% cost reduction. Smith did say there’s a reasonable fear that if that spot becomes full, your instance will drop off, but he said you can get instance flexibility, which allows automatic rollover to other instances. You can also mix a few on-demand instances alongside the spot instances to make sure you’re covered. Spot also comes with notifications when your instance is about to be terminated.
Smith says to start with spot instances, look at interruptible workloads, background infrastructure tools, and stateless applications. His team started with ephemeral workloads within their continuous integration pipeline. From May through November 2020, Centro saw less than 2% of interruptions.
Smith also strongly recommends using the AWS Config tool, which ensures that all of your resources are meeting tagging responsibilities, which in return helps clarify that cost accounting.
Just Let the Cloud Do It for You
Yes, in the above example of Centro, there was no mention of environmental cost. But by making wise economic decisions, they are optimizing their server energy usage, which in turn decreases their carbon footprint.
Cummins says this is all “Really enabled by the cloud — now we have all of this technology for spinning things up and down and we have optimized our infrastructure to do the same things at the same time.”
This can be running things as a batch at night rather than in real-time. Awareness of cloud infrastructure and cloud workloads allows you to know which you can shut off and which you cannot.
Cummins says another place to look at is performance. This is also clearer in the cloud rather than running things locally.
She calls this “a double win — either you’re saving the planet and saving money or if you start looking at ways to improve performance, your users are happier.”
Data-driven automation of cloud services is the fastest way to make a difference in carbon output, while cutting costs. How much money is the best start we have right now for calculating carbon footprint?
Just don’t try to manage it yourself.
Currie explains that cloud companies are just so much better at machine efficiency than other companies. Specifically, she said Google is really good and Amazon is quite good.
She explained, “You really want them to manage your workloads as much as possible. They will ring every CPU out of their servers. Don’t even attempt to do efficient. And that’s where serverless comes in because that is right now the most managed thing.”
Currie continued, “Then you are really getting your cloud provider to do as much work as possible and you should be writing as little code as possible because you’re so much worse at making it more efficient.”
So, yes, the climate crisis is a multidimensional, almost insurmountable challenge, but if every tech team — or every tech consumer — takes a few small steps, it could lead to a big impact.
Amazon Web Services is a sponsor of The New Stack.