QOS modeling aware load balancing in multi-application environment in network on chip

Abstract

Quality of Service (QoS) in Network-on-Chip (NoC) become an important research area. Lately, increasing demands on availability, performance and reliability of Network on Chip connectivity are driving to offer Network connections with QoS guarantees efficiency. However, until now there is no standard method of the Quality of Service (QoS) measurement and less techniques have been used to provide its definition. Therefore researchers are looking for a projection of QoS on quantifiable space, since it is qualitative, subjective and not measurable. However, a few tentatives have studied QoS parameters estimation. Many applications in Networks on Chip (NoC) present a variable QoS parameters such as Packet Loss Rate (PLR), End-to-End Delay (EED) and Throughput (Thp). However, There are a few papers that have developed different method to modelize QoS in NoC. Their QoS presentation doesn't provide a multi-application parameters arbiter. The originality of our approach is based on a proposition of a QoS IP module in NoC architecture to improve network performances. We implement an extended approach of QoS metrics modeling for Network on Chip on multi-parameters and multi-applications environment. The QoS metrics model is based on QoS parameters such as PLR, EED and Thp for different applications. In this paper, we investigated into the problem of serving multi-applications with different levels of Quality of Service. To validate this work, a dynamic routing simulation for 4×4 mesh NoC behaviour under three different applications namely Transmission Control Protocol (TCP), Variable Bit Rate (VBR) and Constant Bit Rate (CBR), is considered. We propose a dynamic load balancing policy based on the principals of queuing theory to offer differentiated services to multiple concurrent applications using QoS metric measurement. The results have shown that QoS modeling approach improves load balancing on NoC and is easy to implement in hardware-software quantifiable representation. Thus, implementing a quantifiable representation of QoS can be used to provide a NoC services arbiter. QoS arbiter interacted with other routers to ensure flits flow and QoS modeling to provide a QoS value.