CSCagatay Sonmez

Cagatay Sonmez

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Cagatay Sonmez has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • DeepEdge: A Deep Reinforcement Learning Based Task Orchestrator for Edge Computing

    by Baris Yamansavascilar, Ahmet Cihat Baktir, Cagatay Sonmez, Atay Ozgovde, Cem Ersoy

    The improvements in the edge computing technology pave the road for diversified applications that demand real-time interaction. However, due to the mobility of the end-users and the dynamic edge environment, it becomes challenging to handle the task offloading with high performance. Moreover, since each application in mobile devices has different characteristics, a task orchestrator must be adaptive and have the ability to learn the dynamics of the environment. For this purpose, we develop a deep reinforcement learning based task orchestrator, DeepEdge, which learns to meet different task requirements without needing human interaction even under the heavily-loaded stochastic network conditions in terms of mobile users and applications. Given the dynamic offloading requests and time-varying communication conditions, we successfully model the problem as a Markov process and then apply the Double Deep Q-Network (DDQN) algorithm to implement DeepEdge. To evaluate the robustness of DeepEdge, we experiment with four different applications including image rendering, infotainment, pervasive health, and augmented reality in the network under various loads. Furthermore, we compare the performance of our agent with the four different task offloading approaches in the literature. Our results show that DeepEdge outperforms its competitors in terms of the percentage of satisfactorily completed tasks.


  • Machine Learning-Based Workload Orchestrator for Vehicular Edge Computing

    by Cagatay Sonmez, Can Tunca, Atay Ozgovde, Cem Ersoy

    The Internet of Vehicles (IoV) vision encompasses a wide range of novel intelligent highway scenarios that rely on vehicles with an ever-increasing degree of autonomy and the prospect of sophisticated services like e-Horizon and cognitive driving assistance. The self-driving vehicle, on the other hand, entails a new passenger profile where sophisticated infotainment applications are expected to enhance the quality of travel. From the technical stand point, for this vision to become a reality a streamlined edge computing infrastructure, namely Vehicular Edge Computing (VEC), is required where computationally intensive workloads are offloaded to a nearby VEC infrastructure. However, the highly dynamic environment renders it difficult to efficiently operate a VEC system to yield the crisp performance required on an autonomous vehicle. In this setting, where to offload each task stands out as a crucial decision problem, and the conventional methods prove insufficient for its solution. In our work, we proposed a two-stage machine learning-based vehicular edge orchestrator which takes into account not only the task completion success but also the service time. To demonstrate how our approach performs in a realistic setting, we employed EdgeCloudSim to design extensive experiments where the characteristics of the vehicular applications, upload/download sizes, computational footprints of the tasks, the LAN, MAN and WAN network models, and the mobility are considered. Detailed performance evaluation of the proposed system via simulation is carried out where both overall and service type-specific performance scores in comparison with opponent schemes are reported.


  • Fuzzy Workload Orchestration for Edge Computing

    by Cagatay Sonmez, Atay Ozgovde, Cem Ersoy

    Edge computing is based on the philosophy that the data should be processed within the locality of its source. Edge computing is entering a new phase where it gains wide acceptance from both academia and the industry as the commercial deployments are starting. Edge of the network presents a very dynamic environment with many devices, intermittent traffic, high mobility of the end user, heterogeneous applications and their requirements. In this scene, scalable and efficient management & orchestration remains to be a problem. We focus on the workload orchestration problem in which execution locations for incoming tasks from mobile devices are decided within an edge computing infrastructure, including the global cloud as well. Workload orchestration is an intrinsically hard, online problem. We employ a Fuzzy Logic based approach to solve this problem by capturing the intuition of a real world administrator to get an automated management system. Our approach takes into consideration the properties of the offloaded task as well as the current state of the computational and networking resources. Detailed set of experiments are designed with EdgeCloudSim to demonstrate the competitive performance of our approach for different service classes.


  • Top Cited Papers

    EdgeCloudSim: An environment for performance evaluation of edge computing systems

    by Cagatay Sonmez, Atay Ozgovde, Cem Ersoy

    Edge computing is a fast growing field of research that covers a spectrum of technologies bringing the cloud computing services closer to the end user. Growing interest in this area yields many edge computing approaches that need to be evaluated and optimized. Experimenting on the real cloud environments is not always feasible due to the operational cost and the scalability. Despite increasing research activity, this field lacks a simulation tool that supports the modeling of both computational and networking resources to handle the edge computing scenarios. Existing network simulators can model the network behavior at different levels of granularity. The cloud computing simulators support the modeling and simulation of the computational infrastructures and services efficiently. Starting from the available simulators, a significant programming effort is required to obtain a simulation tool meeting the actual needs. On the other hand, designing a new edge computing tool has many challenges such as the scalability, extensibility, and modeling the mobility, network, and virtualized resources. To decrease the barriers, a new simulator tool called EdgeCloudSim streamlined for the edge computing scenarios is proposed in this work. EdgeCloudSim builds upon CloudSim to address the specific demands of edge computing research and support the necessary functionalities. To demonstrate the capabilities of EdgeCloudSim, an experiment setup based on different edge architectures is simulated. In addition, the effect of the edge server capacity and the mobility on the overall system performance are investigated.


  • Open Access

    Fuzzy-based congestion control for wireless multimedia sensor networks

    by Cagatay Sonmez, Ozlem Durmaz Incel, Sinan Isik, Mehmet Yunus Donmez, Cem Ersoy