Smarter computing brings data processing closer to home
When a drone needs to identify an object in real time, or sensors detect a sudden change in a building’s air quality, waiting for data to travel to a distant data centre can slow the response. Processing information closer to where it is generated can make computing faster and reduce unnecessary data transfers. But deciding where each task should run is a complex challenge. The EU-funded AC3(opens in new window) project has developed a platform that uses AI and machine learning to manage computing resources across cloud data centres and nearby edge servers, adapting as workloads and conditions change. “By processing data close to where it is generated, the AC3 platform reduces communication delays and network congestion, enabling the system to react more quickly to changing conditions,” says project coordinator Christos Verikoukis of Athena Research Centre(opens in new window) in Greece.
Finding the right place to compute
Modern applications often consist of multiple smaller software components, known as microservices, which can run in different locations. The AC3 platform automates decisions about where these components should be deployed, accounting for factors such as computing capacity, network conditions and energy use. “The main innovation of AC3 is the use of AI and machine learning to automatically decide where and how these components should be deployed and managed,” explains Verikoukis. Three scenarios were used to test the platform: smart building management, drone-based video surveillance and astronomical data analysis. In a smart building, for example, sensors can detect a sudden drop in air quality. AC3 could help the system respond faster by moving data processing to a nearby server, allowing it to trigger an alert or adjust ventilation while reducing the amount of data sent to a distant data centre.
Adapting when conditions change
The drone demonstration showed how the system can respond when computing resources become scarce. If a drone cannot run an AI model for real-time object detection efficiently, AC3 can move the workload to a nearby edge server. When the drone has enough capacity again, the task can move back. “This intelligent and seamless migration ensures that the object detection algorithm continues to operate optimally, regardless of changes in the drone’s available computing resources,” explains Adlen Ksentini, AC3’s technical manager at EURECOM(opens in new window) in France. The researchers also demonstrated the platform’s potential for handling large astronomical datasets. By distributing processing tasks across multiple computing resources, AC3 reduced the time needed to process 5 GB of data by at least 50 % compared with stand-alone systems. The application also maintained 100 % reliability in the project’s tests.
Saving energy and opening up the technology
AC3 can use predictions of computing demand and energy consumption to consolidate workloads onto fewer servers. This allows unused nodes to be switched off while applications continue to meet their performance requirements. Several software components have been released as open source, giving developers and researchers a starting point for further development. Some tools have also been integrated into established open-source ecosystems, including Kubernetes(opens in new window) and OpenShift(opens in new window). For Verikoukis and Ksentini, the next challenge is turning technical progress into wider adoption. Cloud services and edge-based systems often remain separate, and stronger integration and viable business models are needed to bring them together. The AC3 framework has demonstrated how intelligent coordination can make computing more responsive and efficient. The next step is to help organisations put these capabilities to work in real-world services.
