Enhancing disk input output performance in consolidated virtualized cloud platforms using a randomized approximation scheme

Abstract In a virtualized computer system with shared resources, consolidated virtual services (VSs) fiercely compete each other to obtain the required capacity of and this causes significant system's performance degradation. The input output (I/O)‐bound applications running inside their own VS is mainly determined by total time schedule every read/write request, plus actual needed device driver complete request. To achieve right isolation resources (e.g., last level cache, memory bandwidth, disk buffer), it essential limit degradation among collocated applications, as simultaneously several I/O operations are requested VSs, perhaps different priorities. This article proposes resource allocation controller that uses fully polynomial‐time randomized approximation scheme enable concurrent requests in multiple VSs. Monte Carlo sampling approach measure estimate unknown attributes operational originating from VS. formalized an optimization problem aim minimize degree quality service (QoS) violation incidents entire platform. We associated reward function working machine represents fulfillment metric all conducted comprehensive set experiments showed proposed algorithm can reduce QoS 32%, compared result which obtained employing default policy embedded existing Linux container layer.