What actually makes a cloud sustainable?
Since Covid, working from home has taken off, and cloud usage has exploded. What goes unnoticed is that these services run on real computers, take up actual space, and use real energy, lots of energy. This energy is turned into heat and typically disposed of using even more power to cool all the servers back down. Cloud-related energy use now accounts for a massive amount of green energy and is set to rise sharply. What's needed are smarter ways of running the servers to keep reaping the benefits of the cloud.
So, what is green cloud?
Green cloud is shorthand for an environmentally friendly cloud. The term has been around for a while, often as a marketing tool used by Big Tech to refer to greenwashed cloud computing or cloud powered by (lots of) green energy. As a result, companies are going to the ends of the earth to find environmental conditions conducive to building efficient locations for their traditional data centers, co-locations, and Hyperscales. All while continuing to burn through unnecessarily large amounts of (green) energy.
Green cloud is cloud computing that reduces carbon emissions. That is actually environmentally friendly. Thankfully, some alternatives don't require polar exploration or deep-sea diving to be green. Instead, they create a climate-positive, green cloud by putting the accompanying residual heat to good use. Let's take a look at data centers' current and projected energy usage, the impact of Big Tech on renewable energy, how data center efficiency is measured, and how solutions like Leafcloud's truly green cloud can make a significant dent in energy consumption.
What we'll cover now: We'll start with an overview of energy usage by data centers, explore how and why this is expected to grow, and how solutions like the Leafcloud Leaf site solution are the way forward in combatting energy waste.
What updates to expect: I'll be adding more data on the Leafcloud energy usage and co2 savings and exploring the creative ways others are trying to either offset or minimize their waste.
How big is the problem? Amount of data centers
There are roughly 18 million active servers housed in over 4700 data centers worldwide. Already, data centers use an estimated 200-terawatt hour (TWh) each year. That's more than the energy consumption of some countries or about 1% of global electricity usage. As a result, data centers contribute around 0.3% to overall global co2 emissions; ICT accounts for more than 2% of total global emissions. That puts the carbon footprint of the ICT sector on par with the aviation industry's fuel-related emissions. The Dutch Datacenter Association says data center energy usage is closer to 3% in the Netherlands.
Energy use forecast
Forecasting projected energy use is tricky due to the rapid developments in technology, availability, and regulation. The more worrying models predict electricity use by ICT could exceed 20% of the global total energy use, with data centers responsible for about a third of that. Bitcoin and the mining of other cryptocurrencies are also having a significant impact on energy demands and in China have already provoked a response from the government.
Data centers will improve their efficiency. Unfortunately, not enough to cope with the expected growth rate of global data center traffic. It's likely data centers will account for 3–13% of global electricity in 2030 compared to 1% in 2010.
The worst-case projection, while extreme, is not wholly unrealistic. Frankly, changes in regulation, technology, and economic factors should prevent it from occurring. For example, Hyperscale data centers, which are more efficient than traditional ones, account for a growing percentage of the energy consumed, lowering the probable growth rate. In addition, the rise of distributed data centers and the re-use of residual heat can cause a marked shift in the energy usage trend.
IEA, Global data centre energy demand by data centre type, 2010-2022, IEA, Paris https://www.iea.org/data-and-statistics/charts/global-data-centre-energy-demand-by-data-centre-type-2010-2022
Stalling efficiency gains
Another factor concerning growing energy consumption comes from stalling gains in computing efficiency. Compute usage, the amount of data processed, is snowballing. Over the past few years, two factors have delayed this translating into a spike in energy usage. First, moving computing in data centers to multi-tenant cloud providers allows for more efficient use of the available computers and cooling. Second, chips were getting exponentially more efficient. Unfortunately, these methods have run their course. Most services already run in the cloud, and both AMD and Intel chips now use more power. As a result, data center energy usage is set to rise sharply; what's needed are innovations in power-saving and energy use.
In 2019, for the first time, there was no recorded improvement in the industry average PUE numbers. In fact, energy efficiency deteriorated slightly, from an average PUE of 1.58 in 2018 to 1.67 in 2019 (lower is better).
PUE, how efficient is a data center?
The savings made by Hyperscale centers are visible in their power usage efficiency (PUE), which is the amount of energy needed for the data center to run (everything: including lights, cooling, and the coffee machine), divided by the energy used for computing (a PUE of 1.0 would be a perfect score). Conventional data centers typically have a PUE of about 2.0. Hyperscales manage to cut their PUE down to about 1.2. Google boasts a PUE of 1.12 on average for all its centers.
By this year, 53% of servers are expected to be in Hyperscale facilities. The cloud computing market grew from $6bn in 2008 to $288bn in 2019, and Covid has caused a further spike in growth. Now more than ever, it's essential that cloud providers be transparent about their environmental footprint.
Hoarding green energy
The tech sector is the largest global purchaser of renewable energy. Hyperscale operators lead the way, mainly through power purchase agreements (PPAs). In 2019, the top four corporate buyers of renewables were all tech companies. Procuring all of this green energy is better than using fossil fuels. However, due to the lack of energy-saving or re-use, this hoarding of green energy for wasteful data centers results in a green energy drought. This drought forces others to seek nonrenewable sources to serve their energy needs.
IEA, Top corporate off-takers, 2019, IEA, Paris https://www.iea.org/data-and-statistics/charts/top-corporate-off-takers-2019
Since 2019, demand for green energy has only gone up. With a total of 127 renewable energy projects around the world, Amazon now claims it's the world's largest corporate purchaser of renewable energy. Their investment in wind and solar projects grew to 6.5 GW, surpassing Google's 5.5 GW, or enough to power 1.7 million homes in the US for a year.
Despite these massive investments in green energy, Amazon's carbon footprint keeps growing. Its greenhouse gas emissions grew by 15 percent in 2019 as sales increased. With more people shopping online during the pandemic, Amazon's sales have skyrocketed. This is likely to cause a further rise in emissions.
Leafcloud: a truly green cloud
On average, servers and cooling systems account for the lion's share of direct electricity use in data centers. Cloud providers worldwide are in a battle to show increases in the efficiency of dumping their waste. None of them mention recycling residual energy, waste streams of their cloud hardware, or the impact of building data centers. It's about incremental gains in efficiency, hoarding green energy, and planting trees. Think about it like this: progressively improving cooling efficiency in a giant warehouse is still like making a more efficient diesel car, no longer good enough. By rethinking the fundamental idea of a datacenter, Leafcloud is causing a paradigm shift, using servers as highly effective heaters.
The sustainability of this design rests on two key pillars:
Re-use heat to minimize energy waste
Re-use locations to minimize infrastructure waste
By placing the servers in locations where the residual heat is a welcome resource, we get housing. In exchange, we provide the building with additional hot water. We currently re-use at least 50% of the energy we use in our leaf sites and use around 2% of the total IT power required to transport the heat into the location. Re-using all that energy results in a PUE of 1.02. Compared to an efficient Hyperscale data center, this is 66% more efficient. Compared to the worldwide average data center, we are a whopping 108% more efficient. Our measurements show that we capture and deliver around 85% of our heat to the buildings' heating system at our Leaf Site. Of the remaining 15%, some of the losses come from our cooling system, and the rest is residual heat that we cannot capture.
The following graph compares Leafcloud's energy spent with high-efficiency data centers, the average data center, and an on-premise data room.
In this graph we use the unit of ERE (Energy Re-use effectiveness). This measure is similar to PUE (Power Usage Effectiveness) but takes energy re-use into account. The formula for calculating ERE is as follows:
Leafcloud's result comes out to 0.15, which is about ten times more efficient when compared to the 1.67 for the average data center.
Hyperscale data centers don't just score worse on their PUE when compared to Leafcloud. Due to massive infrastructure investments, they cost between 7 and 12 dollars per watt of installed server capacity. By contrast, our Leaf site solution comes in at 1 to 2 dollars per watt.
Data centers aren't built to re-use residual heat
Datacenters currently don't re-use significant amounts of energy. Most massive data centers are outside of built-up areas. As they're air-cooled, they can't profitably store or transport the residual heat from the building. Additionally, they use air-conditioning to cool the building down, using even more energy that's wasted. The best Hyperscale data centers use around 18% extra energy to cool the servers; on top of the 100% to run the servers, the industry average data center uses 67% extra for cooling.
Because of their data center-focused approach, cloud providers run up high extra costs and development time to re-use residual heat. Unfortunately, their core business doesn't have anything to do with being sustainable, and a lack of awareness has slowed the demand for alternative solutions. Now, governments and data center owners are in a stalemate. Governments say: "start delivering the heat." Data center owners say: "come pick it up." Meanwhile, nothing changes.
The end-user of a cloud pays to have a virtual server, security, uptime, and services. They don't care how and where the server is housed. A data center is an inefficient and costly way to run a computer. On top of that, projected data center energy usage is unsustainable when considering global emissions targets. Wider adoption of green cloud can make a massive difference in cloud-related emissions and, importantly, emissions caused by a lack of available renewable energy.
I'll be adding to this post over the coming weeks, so look out for updates as we get our hands on more data. If you have any questions or think I've made a mistake, contact me @Guy