Blog

Introduction

With the advent of the digital age, every aspect of our lives generates and requires data. As a result, data is considered a core business asset, and the ability to effectively collect, manage, analyze and utilize data, as well as building the core infrastructure for it, is becoming increasingly important.

Advances in data analytics, particularly using big data, and advances in artificial intelligence (AI) and its machine learning (ML) algorithms are driving this trend.

However, this infrastructure, particularly data centers, raises a number of ESG and environmental concerns.

In this article, we look at what environmental issues are being raised and what solutions are being discussed.

Managing data requires a variety of equipment, including servers, storage and networking equipment.

Most companies that deal directly or indirectly with data have a room called a server room where this equipment is stored, and IT companies that specialize in data have large facilities called data centers. You can think of a data center as a hotel for computers, with all the different equipment we’ve been talking about — servers, storage, networking equipment, and so on.

As data becomes more important and necessary, this infrastructure, the data centre, grows in size and its environmental impact increases.

Here’s how it breaks down

Electricity Energy consumption

Inside a customer data suite at Union Station. Image: Global Access Point. Public domain. 2008

One of the first issues that comes to mind when talking about data centers is power consumption. While electricity is unavoidable when using electronics, the environmental impact of data centres is more significant than most people realize.

Basically, just keeping hundreds of thousands of pieces of electronics running requires a lot of electricity, but there are additional reasons why it’s a burden on the environment.

1. 24-hour operations

Data centers operate around the clock. This is because many applications require real-time processing, and in a global business environment, services must be continuously available regardless of location or time. Other reasons include the need to optimize the use of resources based on the time of day.

2. Maintaining temperature and humidity

The core equipment in a data centre, such as computer servers, storage systems and networking equipment, generate heat as they operate, but they work best at a constant temperature range, and excessive temperature rises can cause damage to components, so it’s important to keep them cool to maintain a constant temperature.

In addition to managing temperature, it is also important to manage humidity, as high humidity increases the likelihood of condensation (dew point), which allows moisture to penetrate electronic components, causing short circuits and corrosion, while low humidity makes it easier to generate static electricity. Static electricity can affect electronic equipment and cause data loss or hardware damage.

3. Redundancy

Redundancy is a method of ensuring the reliability and availability of a system by installing and operating redundant versions of key components within a data center.

Data centres are built with redundancy in mind, which is why they have more equipment and use twice as much power. The reason for redundancy is that redundancy is almost the only solution in the event of failure of critical equipment, accidents caused by various disasters and even the security of the entire system.

On the other hand, let’s look at how much power these data centers use.

It is estimated that the power consumption per square meter of a data centre is about 1,000 kWh, which is 10 to 50 times more per square meter than a typical commercial office, and the power consumption of a large data center with tens of thousands of servers is equivalent to the power consumption of 50,000 households.

The current electricity consumption of 147 data centers (3,337 GWh) alone consumes almost half the annual electricity production of one nuclear power plant (7,000 GWh), and the expansion of two or three nuclear power plants would need to be considered to meet future electricity needs.

Water use

Steam rises above cooling towers at the Google Dalles data center in Oregon. Google photo

Data center water use is also an issue. The amount of water used in data centers is staggering. In 2011, Google reported that 16.2 billion liters of water were used in Google data centers worldwide, which is about the same amount of water used by 29 golf courses in the southwestern United States.

DWater is used in data centers for cooling, to prevent servers from overheating, and for generators to supplement power in emergencies.

In addition to conventional cooling, various cooling methods have been developed, such as running coolant through the grids of server racks, but regardless of the method, water is used in the cooling tower cooling system.

On the other hand, data centers install back-up power generators, which are rarely used because they are used to back up power in the event of a blackout, but when they are used they use steam to turn turbines to generate electricity, which requires large amounts of water.

Electronic and toxic waste

This amazing pile of English e-waste awaits recycling. Photograph by Stefan Czapski, courtesy Geograph. CC-BY-SA-2.0 https://blog.education.nationalgeographic.org/

A hyperscale data center is generally defined as one with at least 100,000 servers in 22,500 square meters of space.

There will be 628 hyperscale data centers in the world by 2021 (source: Cisco), and there are also hyperscale data centers with more than 100,000 servers in Korea, such as LG U+’s Pyeongchon data center and Naver’s Sejong data center.

The problem is that hundreds of thousands of servers in data centers have a lifespan of only four to five years, and according to the US Environmental Protection Agency (EPA), e-waste, or discarded electronic equipment, is classified as toxic waste. (70% of toxic waste is e-waste).

This is because electronics use toxic chemicals such as brominated flame retardants, lead, mercury, cadmium and beryllium.

According to a recent United Nations report, 53.6 million tonnes of e-waste will be discarded globally in 2019, with only 17.4% recycled and 82.6% landfilled or incinerated. In addition, due to strict environmental regulations and high disposal costs, most e-waste is exported to developing countries such as China, India and Africa, meaning that e-waste pollutes the entire planet.

Heat, noise and electromagnetic waves

Although less serious than the first three, data centers themselves generate a lot of heat. The enormous amount of energy used to keep the temperature inside the data center down through cooling means that a lot of heat is generated inside the data center and released to the outside. This is problematic from a global warming perspective.

In addition, the huge amount of electromagnetic waves generated by hundreds of thousands of servers and the noise generated by the many electronic devices and air conditioners can affect not only people’s lives but also the ecosystem.

The main solutions to the above problems are to

Minimizing the power required for cooling

Project Natick — https://natick.research.microsoft.com/

Various methods are being used to minimize the amount of electricity used for cooling.

Temperature control, particularly cooling, is the most power-intensive part of the data centre, accounting for 40–55% of total power consumption, so it is arguably the most important part of reducing overall power consumption.

Google has reported a 10% reduction in power consumption using water cooling compared to conventional cooling, and many other companies are working to reduce power consumption using a variety of cooling methods.

Meta (Facebook) built its pan-European data center in Luleå, Sweden, which is cooler year-round to reduce cooling costs, and Microsoft is experimenting with putting data centers under the sea with Project Natick.

In South Korea, LG Uplus’ Anyang data center in Pyeongchon and Naver’s Chuncheon data center are also actively using outdoor air cooling to reduce the amount of electricity used for cooling.

Data Centre Optimisation

Image creator from Microsoft Bing

Another issue that should be examined is how to reduce energy consumption by optimizing various equipment such as servers and storage in the data center.

This is because when operating a data centre, applications installed on servers may become obsolete over time, or servers may become unused because service subscriptions are discontinued, leaving them unattended and consuming power.

According to infrastructure management company Cormant, these servers, also known as zombie servers, make up 10 to 30 percent of the servers in a data center.

By better management, removing zombie servers, consolidating and virtualizing servers, and reducing storage equipment through data erasure, you can save a significant amount of energy simply by reducing unnecessary equipment.

Improving power efficiency

Comparison of AC and HVDC power feed systems in a datacenter or communications building.(NTT Technical Review)

In order to reliably operate the various facilities in a data center over a 24-hour period, many basic facilities are required, such as backup batteries, power generation units and highly reliable air conditioning systems for cooling equipment. It is also being investigated whether power can be saved by improving the structure and method of powering these facilities.

A typical example is DC power. Since servers use DC internally, supplying power directly as DC can save power by eliminating the need for AC to DC conversion in the server’s internal power supply. Typically, there is a loss of around 10% per power conversion, so DC power requires fewer power conversions than AC power, so in principle efficiency can be improved.

In addition, the replacement of conventional generators with alternative power solutions such as fuel cells is being investigated to minimize carbon emissions and other costs.

In recent years, ESG has emerged as an important concept.

ESG is a concept that refers to the three aspects of environment, society and governance, and is used to assess whether a company fulfills its social responsibility and pursues sustainable management.

The reason why ESG has become important is that, as regulations on corporate ESG have been strengthened, ESG performance has a significant impact on the valuation of companies.

From an ESG perspective, time series database engines can provide solutions for the environment in a different way than the previous three.

Time series data is data whose values correspond to the passage of time, such as sensor data, and a time series database engine is a software system optimized for storing and serving this type of data.

This means that if you need to process time series data, such as sensor data, you can use a time series database engine to do so with much less equipment than a traditional database engine.

This is a very significant solution that can reduce the amount of power and water required for operation and cooling, as well as reducing waste.

Machbase’s time series database engine has been rated the world’s best in a price/performance evaluation test by the Transaction Processing Performance Council (TPC), an internationally recognised performance evaluation organization, and has demonstrated a performance difference of nearly 6x over Hadoop in the US.

https://www.tpc.org/tpcx-iot/results/tpcxiot_price_perf_results5.asp?version=2

In this test, the cost of meeting the same performance criteria was $329.75 for Hadoop and $54.85 for Machbase.

This almost sixfold difference is due to the superior performance of the database engine itself and the simplicity of the architecture.

For example, in a project with Electronics and Telecommunications Research Institute(ETRI) of Korea, Machbase has demonstrated that a system using Hadoop can be simplified to Machbase.

Before: Hadoop

After: Machbase TSDB

In this case, a project to build an energy big data platform to collect electricity big data and develop deep learning algorithms, Machbase has demonstrated that applying a time series database engine instead of Hadoop can simplify the platform architecture, which not only reduces the cost and time to build hardware, but also reduces the physical size of the overall facility.

The importance of the environment and energy cannot be overstated. The fact that sustainable development indicators, known as ESG, are actually being used to assess the investment value of companies shows that environmental issues are directly related to the survival of companies.

The growth of data and the facilities to store it is inevitable. But minimizing its environmental impact will be a question of how we do it.

Although not a panacea, it is a proven fact that in areas where time series data is used, the impact on the environment can be minimized through the use of time series database engines.

Among the DBMSs currently in development, the Machbase time series data engine has been formally proven to minimize the facilities associated with time series data.

If you are concerned about the global environment and ESG issues, I encourage you to take a look at how Machbase’s time series database engine can help you minimize your environmental impact.