Bringing computation to the data: A MOEA-driven approach for optimising data processing in the context of the SKA and SRCNet
The Square Kilometre Array (SKA) will generate unprecedented data volumes, making efficient data processing a critical challenge. Within this context, the SKA Regional Centres Network (SRCNet) must operate in a near-exascale environment where traditional data-centric computing models based on moving large datasets to centralised resources are no longer viable due to network and storage bottlenecks. To address this limitation, this work proposes a shift towards distributed and in-situ computing, where computation is moved closer to the data. We explore the integration of Function-as-a-Service (FaaS) with an intelligent decision-making entity based on Evolutionary Algorithms (EAs) to optimise data-intensive workflows within SRCNet. FaaS enables lightweight and modular function execution near data sources while abstracting infrastructure management. The proposed decision-making entity employs Multi-Objective Evolutionary Algorithms (MOEAs) to explore near-optimal execution plans considering execution time and energy consumption, together with constraints related to data location and transfer costs. This work establishes a baseline framework for efficient and cost-aware computation-to-data strategies within the SRCNet architecture.