Data centers are having an increasing effect on the environment and climate due to their rapidly rising energy consumption. In 2018, data centers already accounted for 1% of global energy consumption. Depending on efficiency gains and growth of the computing industry, some studies project that data centers may make up 3-13% of the global energy consumption by 2030.
Sino-German Cooperation on Green Data Centres
In this episode of Environment China Podcast, experts discuss their joint research of data center flexibility in China and Europe. The research project was performed under the Sino-German Energy Transition Project, which is implemented by GIZ in China on behalf of the German Federal Ministry of Economic Affairs and Climate Action (BMWK), in partnership with the China National Energy Administration (NEA), and German and Chinese implementation partners including the report's lead author, dena.
The research includes interviews with industry experts and companies on topics such as time-shifting of data and cooling loads, real-time geographic shifting of loads, the pros and cons of relocating data centers to colder climates for greater cooling efficiency, and using on-site energy storage to participate in power markets. They also discuss whether the many obstacles data center operators currently see to becoming more flexible are likely to be overcome - and what policies would help.
Transcript of the podcast
Welcome to environment China, a podcast from Beijing Energy Network, I’m your host Anders Hove. And in today’s episode I posting a special interview related to my ongoing work with Ye Ruiqi (Angel) of Greenpeace East Asia, Prof Zhang Sufang of North China Electric Power University, and Katerina Simou of the German Energy Agency (dena) about the topic of data centres, which are having an increasing effect on the environment and climate due to their rapidly rising energy consumption. In this episode, we begin first by talking about how data centre operators are trying to go green (a topic we discussed with Angel in November 2019), and in fact that the Environment China episode was actually the genesis of our work together on this topic, then we going to move together to the topic of flexibility and joint research which was performed under the Sino-German Energy Transition Project, which is implemented by GIZ on behalf of the German Federal Ministry of Economic Affairs and Climate, in partnership with the China National Energy Administration, and German and Chinese implementation partners including the report's lead author, dena.
So first of all, Angel, welcome back to Environment China.
Hi everyone, so happy to be back!
I'd like to also welcome Prof Zhang Sufang of North China Electric Power University.
Prof. Zhang Sufang:
Hi everyone, Hi Anders, hi Ruiqi.
And third I'd like to welcome Katerina Simou who is an expert for International electriciry grids at dena who supports the implementation of Germany's Energy Partnerships in the field of renewable energy integration. Katerina, welcome to Environment China.
Thank you Anders, hello.
Ruiqi, last time we talked about strategies for renewable energy use in data centres, but a lot has changed since then. Could you start by just updating us on a few findings from your recent work about data centre efforts to boost efficiency or pick up more renewable energy?
Sure, most of our work recently focuses on cloud and data centre sectors, 100% renewable energy transaction. Actually, we have so much news since when president xi announced the carbon peaking and neutrality goals.
First of all, we see more commitments towards carbon neutrality, and a 100% renewable energy just in the last 6 months. Just a couple days ago, china's top data centre operator GDS commit to 100% renewable energy and carbon neutrality by 2030. GDS is the largest independent data centre operator in china is taken up roughly a fifth of the domestic market. Last year, it consumed roughly 2,800 GWh of electricity, nearly double the size of Tencent. This marks the third data centre company in a row committed to 100% renewable energy since we launched our clean power projects in back in 2019; the other two companies are both chinese data centre companies, Chindata Group and Shanghai Athub .
So you may see that this is so much new commitments coming. On the other hand, we're seeing progress in the market mechanism side. In September, there is a new green power trading market mechanism and the pilots rolling out involved 17 provinces. As far as we know, tech companies are the leading buyers in this pilot, such as Alibaba, Chindata and Tencent, GDS are participating in it. And we are for the first time seeing 10-year PPAs being signed between GDS and Zhongguanghe. So not just the commitments from the company side, but also on the ground we're seeing more action. These are the company updates. I'd like to also share that we have two reports this year. First one is still the flagship ranking in cloud ranking 2021 that's tracking the corporate renewable energy performance and commitment from the cloud and data centre sector. The other one is a study about China’s 5G and data centre carbon emissions outlook towards 2035.
We will of cource have links to those in the show notes as always. Ruiqi, if I could just follow up what are the main drivers that are pushing data centres to go green? You've seen a lot of progress, but what would be their interest in purchasing renewable energy?
I would say the biggest driver is the investors' expectation of companies to be more climate friendly. We've heard from them that having these carbon neutrality and 100% renewable energy commitments will help them improve ESG ratings which will help them get more finance.
And on the other hand, they also understand that going 100% renewable energy and just having a renewable energy strategy is the long-term future. They understand that China's net-zero transition will require the energy sector to go from both mostly coal to renewablea. So they just want to get there in advance and explore ways to procure renewable. So far for the pilots, in September, the green power PPA price is more expensive than the coal benchmark price and they are paying a premium to buy that green power. So, that is a statement that it's not just about short-term economics. They are seeing this as a long term strategy, and they're willing to invest up from just to get in some of the advantage later on in this transition to green energy and to a more open market in power.
On the one hand, you have the possibility to purchase renewable energy from third parties using PPAs; on the other hand, data centres have a lot of other ways to go green, one of which is just energy efficiency, and there are a lot of national standards for energy efficiency.
But another one is to actually become more flexible so that data centre usage could help the grid stabilize the consumption of renewable energy. Professor Zhang Sufang, why is it important for data centres to be flexible for renewable integration?
Prof. Zhang Sufang:
In recent years, the dramatic growth of data centres has led to the increase of power consumption and carbon emissions in data centres. Reducing energy consumption and carbon emissions requires efficient and flexible energy use in data centres. The Chinese government called for building a new type of power system with new energy in a dominant position.
So this suggests that in the future, in china's power system, the proportion of variable solar PV and wind power will increase significantly, and the power system would require more flexible resources. So data centre can provide flexible load resources for the power system. So in short, it is important for data centres to be flexible, both for energy saving and emission reductions and for enhancing the flexibility of the power system to meet the challenge of a high proportion of renewable energy.
How can a data centre be flexible given that data centre operators can’t really control the demand for their data operations?
Prof. Zhang Sufang:
I think there are at least four strategies for data centres to be flexible. One is to put off workloads that do not require immediate treatment. For example, we can put off some workloads to the peak period of renewable energy generation: this is a time shift. Secondly, shift some workloads to regions with rich renewable energy: this is geographic shifting. The third is to change the temperature of cooling systems to flexibly adjust the power load. It is evident that it will not have much impact on the operation of the server. In a short period of time, the cooling system accounts for a large proportion of the total power consumption in data centres.
The fourth strategy is to use the UPS energy storage system to optimize the charge and discharge time of the energy storage system. For example, charge during the peak period of the new energy generation and discharge during the peak period of power load. This can provide reserve and operational flexibility for the power system.
Some of these are actually being practiced in China, and I think in particular we heard that some operators are already experiencing geography load shifting, correct?
Prof. Zhang Sufang:
Katerina, for the study, you've talked to data centre operators and researchers in Germany and Europe, what is the situation in Germany? Are data centres already implementing measures to operate flexibly?
Yes, last year we conducted several lectures from the data centre landscape, especially here in Germany. Let’s just say that some companies are doing baby steps and implementing measures to operate data centres flexibly.
There is, of course, a long way to go, even though German data centres are among most energy efficient data centres in the world, any efficiency gains cannot offset this increasing demand for energy in data centres. We discussed a lot of energy efficiency measures with the interview data centre operators and researchers as well as flexibility measures. Some [measures] were evaluated more positively than others, but all in all, we encountered hesitation towards these measures.
In any case, there are, some positive examples of flexibility measures being implemented in Germany. We have, for example, a data centre which uses battery storage for backup power, but also to supply power to the grid and in case grid imbalance occurs. This power is also marketed with support of an energy trading company; this can also lead to additional revenue for the data centre from the energy market.
Katerina, we just talked about in China the potential for geographic load shifting. Do you think that this is feasible for data centres in Europe? Do you think that it's feasible for anybody but the largest kind of mega scale operators that have multiple large data centres and large data loads of their own?
This largely depends on the type of the data centre, especially the hyper-scale data centres can perform a geographical shifting and shift the workloads and maybe process them also offsite. The reason why this might not be an option for other types of data centres—maybe smaller data centres—is that not all workloads can be transferred offside, even the small workloads require big data files to be retrieved, even though they might be small in terms of computation. That’s probably something only hyper-scale data centres can do for now.
What other options are feasible for data centres to become more flexible?
Other feasible options, which we also identified in our workshops with Chinese and German experts, are time shifting of cooling loads, but also time shifting of IT loads. Time shifting of cooling loads is considered more effective. Time shifting of IT loads is considered to be faster.
When you say faster, you mean that the response time is faster for the grid?
Exactly. To implement this shifting of cooling loads is still required some ice or water storage to be in place, but also a well-developed grid that could absorb this surplus cooling load from the data centre.
Also a more dynamic configuration of servers so that they don't only perform on a maximum level, but also in a more dynamic and energy efficient way can also be feasible for some data centres. This is a second flexibility option.
The utilization of data centre battery storage in periods where there is no data utilization to provide electricity frequency regulation. The solution doesn't really provide data centres great revenues, but it can be very valuable for the grid operator, because it provides balancing capacity.
But the type and location of the data centre are the aspects that determine which solution the data centre could eventually apply.
What are the main obstacles to these flexibility options in China?
Prof. Zhang Sufang:
As I just said, technically data centres have the potential to use power flexibility. However, our research results show that most data centres in China prefer not to use power flexibly. I think there are four reasons for this. One is that data centre operators lack awareness of flexible power use. Their concern is that the time and geographical shifting of workloads may undermine the reliability of data centre operation, and this will in turn affect the quality of services for their customers.
Secondly, the data centre industry is a high value-added industry. They are not very much interested in the economic benefits of participating in the demand response of the power market. The third reason is that china's spot power market is underdeveloped. The price signal has not been fully utilized to guide the power load. The last reason is that the type and number of workloads in data centre are increasing, and it is hard to analyze the flexibility of all kinds of workloads.
Right. So just to repeat a couple these items or clarify a little bit:
First of all, China doesn't really have the spot market that would enable the data centres to receive price signals in the first place. And then secondly, even if those price signals were available, the data centres are concerned that the value would be too small compared to their main business, which is providing data service.
Ruiqi, how could we imagine that any of these problems might be solved if they could be solved? There’s potentially the argument that data centres just will never be flexible because the need for reliable data services, but are there policies that could work in China aside from just having spot markets?
I think this will have to be considered in relation to carbon emissions control targets and renewable energy consumption targets for specific sectors. Now the data centre sector is under very strict regulations on what's called the dual targets, that are distributed from central government to local provinces and then to each sector. The dual targets refers to the energy consumption and energy intensity of each project level or company level targets.
Now, the current system looks at the total amount of the energy consumption instead of looking at both the amount of energy and the amount of carbon emission. I understand that the policy regulators are shifting this dual energy targets towards a carbon dual target system in the future, meaning the total amount of carbon emissions and the carbon intensity targets, so the system is being shifted. If we have that administrative system sorted out, we will have more incentives for companies to procure renewable energy at different times to meet the administrative target. That's one case.
And on the other hand, spot market is still along the way, so before that, I think China is rolling out policies to allow power prices to fluctuate more than before. So you see recent policies loosening up the coal power benchmark price and allowing the fluctuation up to 20% compared to previously 10% only. They specifically mentioned that for energy-intensive sectors—and we may think that [energy-intensive sectors] only refers to steel and so on, but now it also includes data centres—for these sectors, there will be no cap for the power price rise, it will depend on local provincial of power markets.
So I'm seeing a lot of articles written by experts from the sector discussing how rising power prices will create a lot of pressure for their business operations. That will serve as another incentives for them to better manage their power usage behavior to reduce costs. In that case, if in the future, we can have more renewable-penetrated system that reward more flexible energy usage and to encourage flexible usage of renewable energy, then I think that will provide a lot more incentive for the data centres.
Just to clarify one thing on the dual control administrative targets. This is not a system that just affects the power industry like the current case with the carbon emissions trading. This is something where industry, when they're thinking about planning their expansion, they have to keep in mind that the provinces and their own industry have caps on emissions and energy consumption, correct?
This is not only referring to one specific sector or rather it is an administrative target on the officials of each provinces, they must meet the KPIs, otherwise they will have different levels of punishment. So, that's why it's very strict control at the provincial and local and on-the-ground level.
So, in the future, if we can only shift and link these dual energy targets to a dual carbon emission targets, we can see more incentives to procure renewable energy to meet this administrative target.
I also want to mention energy storage. Other than the power market or spot market mechanism. VNET is also a data centre company in China have shared their recent progress of building extra energy storage capacity along with the data sectors to make use of the peak trough price difference (峰谷电价差). They actually are able to make money, and have economic incentives from this pilot in the Guangdong power market.
What about policies that were introduced in some provinces after the power outages this fall—the orderly power consumption rules? Have data centres been impacted by those and with that encouraged them to be more renewable friendly?
Yeah, data centres are definitely impacted. We've heard of data centres in Hebei being shut down multiple times because of these orderly power consumption rules and they are also struggling to deal with this. Like I mentioned before when they choose who to shut down, what loads to left on is still not link to carbon emissions.
So, I think when these are more linked, we can have a better regulations and incentives, to incentive the companies to use renewable energy.
Right. Katerina, I wanted to ask about the potential tension between efficiency and flexibility. We know that a lot of data centres already have located in cooler climates as a way to reduce their cooling load and also use the external environment as a way to absorb their cooling load.
I wanted to know if the strategy of maintaining a very high utilization at data centres in order to take advantage of that constant cooling load and improve overall efficiency or power usage effectiveness (PUE), dos that contradict the idea that data centres can and should be flexible in order to provide value to the grid and respond to the changing power price over the day?
Yes, especially in the example you provided, there could be attention between efficiency and flexibility. There is no silver bullet with regard to the energy efficiency and flexibility measures that data centres should apply. There is also no one-size-fits-all solution, especially because we have to deal with different types of data centres. These different types of data centres will also eventually follow different pathways to improve the flexibility and energy efficiency.
For example, co-location data centres or a smaller university data centre may, for example, have it easier to implement flexibility solutions which would be more focused on their IT load. This is not possible for hyper-scale data centres, for example. What remains of great importance is ensuring stable data centre operations. And it could be that at this point, energy efficiency measures are evaluated more positively than flexibility matters because there is not as much intervention in the data centre operation when energy efficiency measures are implemented.
The location of the data is another very important point. If data centres, for example, are located in densely populated areas, which they are in Germany, they might cause grid bottlenecks. On the other hand, if they are located more remote areas, they will have a grid connection issues or bandwidth problems.
What remains important is that data centre operators will sooner or later have to apply some form of energy efficiency or flexibility measures. And some flexibility measures could work well in consolation with efficiency measures. If we take the example of optimizing server utilization, like adjusting service to not have maximum power but the more dynamic configuration, this can also help adjust the data centres' power consumption without having this contradiction between flexibility and energy efficiency.
As part of the study, Zhang Sufang and Ruiqi, you interviewed a number of data centre operators and held workshops and heard from the industry. What would you think is the most interesting or new thing that you've learned through this process? And I know that you've been working on this issue for years, but maybe just in the last few months of working on this topic, have you learned something new that seems promising or seems perhaps like a bigger challenge than you realized?
Prof. Zhang Sufang:
One very interesting view from an expert with regard to the incentive to be flexible for the purpose of enabling the integration of renewable energy in the power system: One expert thinks if we can link the power usage right with their carbon emissions. They care more about the right of energy use. As Ruiqi just mentioned, the data centre was shut down, due to the orderly use of power rules. So maybe this is a very interesting suggestion to be a very good incentive for them.
Right, I definitely agree with Sufang. That's exactly was the same policy that I mentioned before where we need to transition from a energy dual target system to an emission target system. Because in the current system, a data centre permits are still being about evaluated as the total amount of energy that they consumed and the GDP value that they bring, even if it could be a zero-emission data centre and they could not get the permits.
And for me, personally, the biggest findings I learned is that the bandwidth is also quite important to enable the geographical load shifting in China. A lot of companies have mentioned that bandwidth speeds are quite expensive in China. I haven't done much comparative analysis to other countries. First of all is the price, and the second point is that it's not very connected from the city level to the western part of China. So if we want to do load shifting, we must need cheap and very well-connected and fast internet to enable that.
So I think that fortunately we see a new policy coming out: China has picked out data centre hubs and wants to highlight their development in the future. They will try to increase the bandwidth capacity around these data centres hubs.
Katerina, what would be the most interesting surprise that you came out of the study for you?
When I first started working on this project a year ago, before I did an interviews, I thought the most interviews would be will-less and hesitant to implement flexibility measures in data centres. I had this idea in my head that the expected outcome would be that the interviews would contradict with interviews of the Chinese experts. We would maybe be able to compare the two opposing views in our study. In the end, that wasn't the case. We were faced with more or less the same hesitation as in China towards flexibility improvements.
But in the end, this also proved to be a very good base for a study and the workshop discussions because any challenges and solutions were commonly identified by both sides, and we were able to have constructive discussions of what we could apply in the next years.
But it is clear to me now that data centres will in the future assume of central role in demand side processes and will also become an integral part of the energy transition. In our study, we identified such a wide palette of flexibility and energy efficiency options that data centres can choose to apply. So in the end, there is always the possibility to apply at least one of these options.
The energy transition of both countries should be a common effort and involve as many actors as possible in order to be successful, and data centres should be included in this process.
Yeah, I agree. Much more similarity between Europe and China in terms of the resistance and in terms of the difficulty of being flexible. Yet, on the other hand, the data centre industry is already undertaking serious efforts to integrate renewable energy and also become more flexible in terms of their cooling loads that you mentioned, even though it may seem difficult to many both in the industry and academic studying, the field nevertheless progress is being made in this field and it is not impossible to imagine the data centres becoming both flexible and efficient at the same time.
Okay, that's all the time we have for today and we've certainly covered a lot of ground. I want to thank all the three of you for joining.
Further reads on the topic:
“China 5G and Data Center Carbon Emissions Outlook 2035” Greenpeace East Asia, 2021, at https://www.greenpeace.org/eastasia/press/6608/electricity-consumption-from-chinas-digital-sector-on-track-to-increase/.
“Clean Cloud 2021” Greenpeace East Asia, 2021,at https://www.greenpeace.org/static/planet4-eastasia-stateless/2021/04/03a3ce1a-clean-cloud-english-briefing.pdf.