A QoC‐based model for performance and QoE trade‐off in distributed systems

Abstract

Summary This paper proposes a quality of context (QoC)–based model to enhance trade‐off between system performance and users quality of experience (QoE) in distributed systems. The QoE considers the users subjective perceptions of quality on technology services and depends on context attributes related to the users expectations. In addition, context information can be used to improve performance in context‐aware systems. Therefore, the QoC‐based model proposed in this paper aims to establish performance and QoE trade‐off for resource allocation in distributed systems. With the massive data growth over the past years, the demand for computing capacity has been notably growing in a proportional way. In view of that, the demand for distributed systems has also increased significantly due to the advantages of distributed computing such as cost‐benefit gains, large processing capacity, and an extensive capability for resources sharing. In this sense, the large‐scale utilization of distributed systems has generated problems concerning resource allocation because of the heterogeneous characteristics and continuous evolution of distributed computational devices, resulting in several challenges in terms of environment performance and users' satisfaction trade‐off. The model conceived in this paper provides for an experimental evaluation considering performance and QoE trade‐off by means of simulations. Simulations were performed in order to study complex distributed systems scenarios. Experimental results showed that the proposed QoC model policy has an enhanced performance in comparison with classical approaches.

Adaptive video streaming testbed design for performance study and assessment of QoE

“…40 Moreover, we modified the open-source application to record the activity of the objective metrics, and the recorded information is sent to the main server over HTTP protocol.…”

“…CPU usage and battery power consumption for the VMs and the entire virtualized system are measured by using the Linux monitoring tools. In the VMs, 39,40 the number of cycles executed per second is analyzed when a user watches videos of different segment lengths. The average CPU usage is influenced by some factors such as high encoded video quality, quality video switching, CPU environment, and the behavior of the video playback application and availability of clients' throughput.…”

“…We recorded the track of the user such as IP address, and after finishing the playback of the video, all information of the metrics is transported to the main server to assess the QoE in real time. 40…”