ML-driven classification scheme for dynamic interference-aware resource scheduling in cloud infrastructures

Computing systems continue to evolve, resulting in increased performance when processing workloads large data centers due the virtualization benefits. This technology is key factor that allows multiple applications share resources, thereby enhancing overall hardware utilization of cloud computing environments. However, cloud-services contending for shared resources are susceptible cross-application interference, which can lead significant degradation and, consequently, an increase Service Level Agreements violations. Nevertheless, state-of-the-art resource scheduling still relies mainly on capacity, adopting heuristics such as bin-packing and overlooking this source overhead. But recent years, interference-aware has gained traction, with investigation ways classify regarding their interference levels proposal static models policies co-hosted applications. The preliminary results already show a considerable improvement be considered first steps toward dynamic strategy. In scenario, paper proposes machine learning-driven classification scheme main goal present how approach, better represents workload variations, affects scheduling. place, we analyze react different workloads. Then, explore distinct formats evaluate efficiency, taking nature into account. Lastly, application classifier based learning techniques compare it related work, variety patterns. Preliminary revealed efficiency by 27%, average, applying our approach infrastructures.