Replicating Data Centre Conditions | Techbuyer’s Digital Twin

After having spent decades within the data centre and IT hardware industry, what has become clear to our team is that each data centre is unique. The server environment can change depending on location, size, cooling equipment and hardware, to name but a few factors that come into play.  

This is why we have a dedicated research team examining the effects of server environment and configuration on device efficiency and data centre modelling. Kat Burdett, Stephen Clement and Nour Rteil provide invaluable research and information that enhances our understanding of server efficiencies; this information is then passed onto our customers through the services we provide.  

This team is made up of experts in the field of server efficiency who all approach the topic from their individual specialisms. However, many within our team prefer to refer to them as the people who steal servers! 

How Do We Test Server Efficiency?  
 

The question of how we can replicate and adapt server environments comes down to research and design conducted by Dr Kat Burdett, Research and Development Engineer at Techbuyer. During her PhD, Kat designed and built a bespoke wind tunnel that allowed her to isolate a device and control the environment in which it runs. Kat is now on her 3rd iteration of this tool and the team uses the wind tunnel daily in their research.  

Within this wind tunnel, we can replicate the server conditions for different temperature setpoints and pressures to test exactly how servers can run as efficiently as possible.  
With this tool in place, our research team have provided invaluable findings to our wider company that have informed how we advise and recommend server configurations and technology to our customers. Their research has also been published in leading academic journals, including IEEE.  

The team have now been working on their latest project of developing a digital twin – a cloud-hosted, co-simulation tool based off data centre modelling. 
 

Why is Techbuyer Developing the Digital Twin? 
 

The digital twin is designed to increase our understanding on how data centres function and how the hardware within them works. 

The nature of data centres is changing and the highest growth rate in the industry is occurring outside of the developed world meaning that new research is needed into creating efficient data centres within varying climates.  

Similarly, with the growth of new paradigms like Edge Computing, we are seeing a shift in data centre design towards smaller data centres in cities becoming more commonplace. This means that, for our sales team as well as the wider industry, understanding how server devices work, and work best, is essential for data centre efficiency.  

Now that we have conducted lots of modelling on servers and deepened our understanding of server efficiency, we are now combining those models with physical models of a data centre to create a digital twin. By composing multiple models together as a system of simulations we can more fully understand the dynamic behaviours within a data centre, driving further efficiency changes and understanding of sustainability metrics. 

With this understanding, data centres can make better decisions for their hardware and software to increase efficiencies and performance. With the digital twin we can mimic each aspect of a data centre within a controlled environment and then predict changes and effects.  

Currently, every aspect of a data centre is siloed, including hardware, software, cooling, power consumption. Because of this, there is a lot of redundancy built into data centres, in both units of equipment, and the capacity of that equipment, which impacts inefficiencies. We want to paint a fully detailed picture to understand all inefficiencies and reduce them altogether.  
 

How Does the Digital Twin Work? 

 
The digital twin is a cloud hosted tool that can be accessed from anywhere and will function through co-simulation. The system will utilise any models or data that a data centre already has. In a way similar to a Lego set of building blocks the data and models can be composed in different ways to answer different questions. With this combination of systems, we can model all aspects of the data centre, avoiding a large monolithic build which would be very difficult to create and manage. 

We model the different aspects - hardware, workload, cooling etc – that have previously been siloed within the industry, so that we have a complete image of what the data centre is doing and we can predict how it will respond to changes in environment, operations, hardware etc. 

Unless you model all aspects together, you can’t understand the full effects of decisions and inefficiencies within the entire system, nor can you understand the impact of environment, climate or location.  

A good example can be found when examining Power Usage Effectiveness PUE and cooling techniques. The biggest metric for data efficiency is PUE – if you manipulate setpoint to reduce cooling your PUE improves despite the fact that your data centre can be consuming more energy. This is due to the fact that it is marked as a ratio rather than a number of consumption. These are limitations that we aim to change with the digital twin by looking at all the environmental aspects holistically.  

For a full rundown of our team and their research backgrounds, head to our article where they have each introduced themselves.