Open Research Europe presents an interview with Thomas Batz, the author of ‘The Internet of Things Workload Emulation for Data Centers’, in this month’s Research Spotlight. Read on to find out more about the research.
Firstly, what is the ‘Internet of Things’?
The Internet of Things (IoT) refers to the concept that all devices (so long as it has an on/off switch) are connected to the Internet and to other connected devices. The Internet of Things is a giant network of connected things and people – all of which collect and share data about the way they are used and about the environment around them.
Can you share a summary of your Software Tool Article?
Our Horizon 2020-funded project, BodenTypeDC aimed to prototype an innovative and cost-efficient energy data centre in Boden, situated in the north of Sweden.
Since it’s not easy to use real applications for the design and test of data centre setups, a tool is required to emulate real applications. That is why we developed a methodology for modelling benchmark loads in data centres, using benchmarks to generate workload models, in order to have a realistic and reproducible way to emulate IT loads. A benchmark load is a unit for data centres that is used to test different distribution strategies for synthetic and application-oriented uses. We used tools like Docker and Kubernetes to generate the workload and benchmarks were used for an objective assessment of the performance metrics (including Power Usage Effectiveness) of the data centre.
The workload models were characterized for the intended application of smart cities, environmental monitoring, and smart factories. We created a functional architecture for the development of application patterns, the deployment of loads in the data centre and the measurement of performance metrics. We used a software containerization platform, Docker, for the technical realization. Containerization is the process of packaging software code so that it can run consistently on different computing equipment.
What inspired you to investigate this?
In many IoT-related projects, a lot of data must be collected, stored, and processed to fulfil the demand of the various dashboards and the end-user application, especially when it comes to maintenance or managing smarts cities. One key task was to determine the dimensions for a data centre that we could use in these projects. With our approach and the tools at our disposal, this determination could be simulated and the BodenTypeDC project allowed to test this in a real data centre.
The approach of the project was to identify synthetic and real-world workloads for a data centre, and then to test this in a real data centre which was built at Boden. The real-world workloads included predictive maintenance and smarts cities that run for 12 hours each, representing the typical characteristics of these applications, e.g., the data usage of smart cities during the morning rush hour, midday activities, and the longer lasting rush hour in the evening. As mentioned earlier, Docker and Kubernetes were used as software components with a distribution tool on top of them to specify the number of applications running in parallel on the servers in the data centre. The distribution tool enabled the identification of different distribution strategies on the servers e.g., equal distribution on all racks (the servers are clustered in racks) or whether the racks fill one after the other. And with the real data centre existing in Boden, we could perform real tests and gather real measurement data instead of simulations or estimations.
What were the outcomes of your research?
The results of our research were manifold. We were able to develop a methodology for specifying real workloads for data centres, as well as software tools to generate and distribute them via Dockers. The project also generated actual measurement data of the execution of that load in a real data centre.
Why is open research important to you and your co-authors?
As researchers working in public-funded organizations, we want to publish our results and discuss them with our colleagues quickly, so publishing with Open Research Europe was a great way of achieving this. Open Research Europe offered us a way of reaching our goals as, we post-publication peer review we can publish and share our research results quickly without waiting for a suitable conference which often has long timelines.
What are your next steps?
We are continuing our research in IoT applications, specifically in industrial production and smart cities. Therefore, our approach of specifying detailed load patterns and executing them in real data centres will be of great importance.
Read the full Software Tool Article today to dive deeper into Thomas Batz’s study and findings.
