In search of eternal life for IT hardware
Wed 24 Jun 2020 | Astrid Wynne
The dream of a circular economy for IT hardware is not a fantasy, says Astrid Wynne
Circular economy is all about creating a roadmap from ‘cradle to grave’ to ‘cradle to cradle’. The ideal is to create a system that is regenerative by design, which minimises harmful emissions and turns waste into manufacturing resource. ‘Eternal life’ for our equipment is the great modern-day challenge, but it is by no means an impossible dream.
Energy Efficiency vs Waste
Conversations about sustainability in the data centre sector have often revolved around energy efficiency and electricity supply. However, with so much great work having been done on PUE and the provision of renewable energy, the focus is shifting to IT hardware.
Research released by organisations like the Uptime Institute suggests the next big area of gain is server optimisation, and with the slow-down in Moore’s law, this may look different from previous expectations.
Servers account for the lion’s share of energy usage in the data centre. The traditional argument has been that stripping out older machines and replacing with newer ones is the smart move for reducing energy usage overall.
However, studies carried out recently show that this is more complex than it used to be. Refreshing machines that are older than five years has undeniable benefits for reducing energy use, increasing compute power and saving space. However, for later machines the picture is less straightforward. With more incremental improvements between CPUs, component level upgrades could be a better option.
In November 2019, two of my colleagues presented the results of benchmarking experiments on refurbished and new equipment at DCW Frankfurt. The first section of the presentation demonstrated there is no difference between refurbished and new servers for power draw or compute power.
The second part was even more interesting. It showed that the immediate past generation of server can outperform the latest just by doubling the RAM. A previous generation CPU upgrade
made a further a big difference.
This approach saves on materials usage and cost without losing on energy efficiency. It makes the case for rebuilding rather than refreshing for newer machines and creates more emphasis on materials efficiency. Now we need to place a real
value on what this represents.
Inside Our Servers
Before we can come up with a decent plan on how to reuse what we have already, we need to have some quality information on what is out there. This is something that the Circular Economy in the Data Centre Industry, or CEDaCI, a European research project, is working on with Techbuyer as one of the associate partners.
As an industry, we are faced with a plethora of different server manufacturers with distinct component parts for each generation. These are made remotely by a number of different manufacturers, which means accurate assessment of all the IT equipment out there is an incredibly tall order.
We know from studies like the JRC Science and Policy Report, which the EU used to inform the Ecodesign Directive that assesses exactly what is in servers and what happens to the material flow, these are sticky issues without exact answers.
Beginning with what we do know, the hundreds of millions of servers out there (121 million deployed 2019-2023 alone according to Gartner) contain massive amounts of steel, aluminium and plastic, which are all costly in terms of Greenhouse Gas emissions to produce.
They also contain precious metals and 12 of the 27 critical raw materials (CRMs) identified by the EU as in short or politically unstable supply. CRMs are predicted to run out in decades unless we find alternative sources. Some have suggested the seabed; using what we already have seems like a much better option.
The Cost of Production
The energy it takes to produce this hardware, also known as embodied energy or Scope 3 emissions, needs to be taken into account. This is often expressed in its carbon equivalent.
An absolutely agreed-upon figure on this is hard to come by. The JRC study puts the embodied energy of the average server at around 380kg CO2 equivalent based on data from 2012. More recently, Dell has released comprehensive figures of the carbon footprint of its equipment.
These calculate the proportion of embodied energy at 9.7% – 29.5% depending on model using software from MiT. The range of embodied energy according to this method would be around 1200 kg – 1800 kg CO2e, dwarfing the JRC figures.
There are some great moves towards improving recycling technologies for precious metals and CRM recovery. Traditional methods have revolved around crushing and melting to extract copper, gold or silver, which yielded imperfect and very limited results.
New technologies such as bioleaching are improving on this but are some way off widespread commercial application. For example, extracting Cobalt from the coating of a hard drive may be technically possible but it has to be economically viable to do so.
Multiply that by the number of trace elements, components, makes models and generations out there and we begin to understand the challenge. Reuse, reallocation and refurbishment are generally accepted to be much higher up the value system for now.
A Systemic Approach
Google has published details of its inhouse refurbishment system for servers and how much it saves as a result. Smaller end users are more reluctant to do this, publicly at least. Myths about refurbished including a perceived risk of second-hand equipment often hamper reuse, despite the evidence. A misunderstanding that data destruction demands hardware destruction and disposal limits sale of equipment back onto the secondary market, which in turn creates needless waste.
Education of the sector through research, publications and honest conversations will help and there are good signs that this is already happening. Major manufacturers take back servers from customers at the end of the lease term, refurbish them and make these available with a one-year warranty on the secondary market. Refurbishment specialists like Techbuyer go further, offering a three-year warranty on parts and a vendor neutral attitude towards supply.
In an ideal world, we would look at further collaboration amongst manufacturers and with the secondary market. Components are often incompatible between manufacturers and also between generations. In some cases, there is good reason for this because the system is designed for the performance increase of particular parts like CPUs.
In other cases, such as power cables, fans, Power Distribution Units, it is not clear why these need to be redesigned for each machine. Looking into this as an industry and finding consensus would go a long way to making best use of resources.
Changing the eco-system around delivery and take back would further improve reuse and remanufacturing. Pulling equipment back through a two or three-tier distribution model is a major challenge for manufacturers, as it is for end-users.
Involvement from accredited ITAD services helps here, as do standardised processes for restoring, repairing and rebuilding equipment. All it needs is a more cooperative approach and some dedicated – possibly disruptive – thinking.
Since the tech sector is famous for innovation and tearing up rule books, it is in a great position to find solutions. Eternal life for our IT hardware? We can do that.