eResourcing: Handling data sustainably

Humans today store vast amounts of data. Travelers upload photos of mountains and beaches to the cloud, sensors send thousands of machine parameters to control centers, and online stores ship packages daily based on millions of stored customer details. The immense quantities of data in the global computer network seem to be completely natural, as they can be stored and retrieved at any time. But dealing with data isn’t all that natural, says Prof. Alexander Martin, director of the Fraunhofer Institute for Integrated Circuits IIS in Erlangen: “Data is a resource, and just as with the consumption of any other resource, generating and storing data produces substantial amounts of carbon dioxide.” After all, a great deal of electricity is required to manufacture and operate transmission lines and large server farms. Even a simple query in a search engine triggers an entire cascade of actions that result in the emission of a total of about 10 grams of carbon dioxide. “It’s time we learn to handle data in a sustainable way – and to consider very carefully why we collect data and how we can reduce the amount.” Martin calls this economical and sustainable strategy eResourcing. And just as with fossil resources, he sees recycling as an integral part of this strategy. “We put a lot of effort into generating data today, so we should reuse this data as often and in as many different ways as possible.” 

Getting the most out of data

One example of using data sustainably is the cooperation between Fraunhofer IIS and household appliance manufacturers. The latter have to stock thousands of spare parts for a number of years in case a customer reports a component failure. The problem with this is that spare parts take up a lot of warehouse space and tie up capital. The trick is to estimate the future need for parts as accurately as possible in order to keep inventory small. Currently, however, there usually isn’t enough empirical data to calculate the service life of individual appliances and components. A few dozen data points on porous coffee machine tubing aren’t exactly a sound basis for a comprehensive service life analysis. “But if we merge and analyze the tubing data for many similar products, we can gauge the durability and demand for spare parts quite well,” Martin says. “This is one way to amalgamate unused data sets to create a new use.” It’s like aluminum yogurt lids that are melted down and reshaped, then begin a new life in a car body. “With metals, the entire life cycle is considered – from the cradle to the grave. Achieving this with data would enable us to leverage great potential.” 

eResourcing - from “big data” to “relevant data”

Not all data is equally valuable or equally meaningful, and yet it requires the same investment of effort to produce and store it. Systematically deciding which data is really necessary and relevant can prevent or at least significantly reduce the volume of long-term, persistent data. For example, when sorting foreign matter out of food, the only relevant data sets are the few in which impurities were identified. Images of the faultless food make up the bulk of stored data, but they are virtually inconsequential for further processing: for example, they are less relevant – or can easily be reproduced – for any subsequent use in machine learning.

eResourcing - from “big data” to “relevant data”

Not all data is equally valuable or equally meaningful, and yet it requires the same investment of effort to produce and store it. Systematically deciding which data is really necessary and relevant can prevent or at least significantly reduce the volume of long-term, persistent data. For example, when sorting foreign matter out of food, the only relevant data sets are the few in which impurities were identified. Images of the faultless food make up the bulk of stored data, but they are virtually inconsequential for further processing: for example, they are less relevant – or can easily be reproduced – for any subsequent use in machine learning.

Smart sensors – less data

The teams at Fraunhofer IIS also deal with communication between machines. Sensors record data and send it to control centers, which then respond with instructions. Billions of these kinds of data points are constantly on the move. It is more expedient to use embedded algorithms to evaluate the majority of these data points directly at the sensor. This avoids having to send every individual temperature reading to a remote location. Instead, the cognitive sensor generates something like a situation report, which is then sent to the control center as a compact data packet. One nice side effect of this is that when sensors have fewer individual data points to transmit wirelessly, their batteries last longer.

Greater resource independence, also with data

This perspective of data as a resource may surprise some people. But there is yet another parallel with traditional resources such as oil and rare earth metals: independence and resource security. Today’s major international corporations store their data in the cloud. We don’t really know what happens with it there, yet we are essentially dependent on it. That’s why Martin says, “For the future, I recommend a European federal data management system in which we in Europe retain data sovereignty.”

Article written by Tim Schröder