Baris Yamansavascilar

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.

Earthquakes are one of the most destructive natural disasters harming life and the infrastructure of cities. After an earthquake, functioning communication and computational capacity are crucial for rescue teams and healthcare of victims. Therefore, an earthquake can be investigated for dynamic capacity enhancement in which additional resources are deployed since the surviving portion of the infrastructure may not meet the demand of the users. In this study, we propose a new computation paradigm, air computing, which is the air vehicle assisted next generation edge computing through different air platforms, in order to enhance the capacity of the areas affected by an earthquake. To this end, we put forward a novel paradigm that presents a dynamic, responsive, and high-resolution computation environment by explaining its corresponding components, air layers, and essential advantages. Moreover, we focus on the unmanned aerial vehicle (UAV) deployment problem and apply three different methods including the emergency method, the load balancing method, and the location selection index (LSI) method in which we take the delay requirements of applications into account. To test and compare their performance in terms of the task success rate, we developed an earthquake scenario in which three towns are affected with different severity. The experimental results showed that each method can be beneficial considering the circumstances, and goal of the rescue.

There is an ever-growing race between what novel applications demand from the infrastructure and what the continuous technological breakthroughs bring in. Especially after the proliferation of smart devices and diverse IoT requirements, we observe the dominance of cutting-edge applications with ever-increased user expectations in terms of mobility, pervasiveness, and real-time response. Over the years, to meet the requirements of those applications, cloud computing provides the necessary capacity for computation, while edge computing ensures low latency. However, these two essential solutions would be insufficient for the next-generation applications since computational and communicational bottlenecks are inevitable due to the highly dynamic load. Therefore, a 3D networking structure using different air layers including Low Altitude Platforms, High Altitude Platforms, and Low Earth Orbits in a harmonized manner for both urban and rural areas should be applied to satisfy the requirements of the dynamic environment. In this perspective, we put forward a novel, next-generation paradigm called Air Computing that presents a dynamic, responsive, and high-resolution computation and communication environment for all spectrum of applications using the 6G Wireless Networks as the fundamental communication system. In this survey, we define the components of air computing, investigate its architecture in detail, and discuss its essential use cases and the advantages it brings for next-generation application scenarios. We provide a detailed and technical overview of the benefits and challenges of air computing as a novel paradigm and spot the important future research directions.

Failures in networks result in service disruptions which may cause deteriorated Quality of Service (QoS) for the end-users. Since SDN is becoming the mainstream paradigm for networks, implementation of a robust fault tolerance scheme for SDN-based networks is crucial. Existing SDN data plane fault tolerance approaches can be classified as reactive and proactive which may or may not rely on the controller, respectively. However, none of them qualifies as a complete solution, providing only partial remedies. In this work, we propose Dynamic Protection with Quality of Alternative Paths (DPQoAP) that considers not only the existing faults within the network but also the quality of alternative paths. As a result, we can sustain the QoS throughout the network after the recovery. We also investigate how application based parameters are affected by link failures. To this end, we explore the change in Quality of Experience (QoE) caused by link failures under different cases using Dynamic Adaptive Streaming over HTTP (DASH) for video streaming. On the other hand, even though DASH is proposed as a solution to improve the QoE affected by the dynamic conditions of the networks, it remains insufficient to handle the congested links that show the symptoms of a link failure. Thus, we apply the data plane fault tolerance approach in SDN to improve the QoE of DASH clients in the case of congestion as well as the failure. The performance of the proposed solutions is evaluated through various experiments considering the QoS and QoE parameters. It is observed that DPQoAP enhances the efficiency of the networking operations and adaptability of the applications.

In this paper, we develop a novel throughput-maximizing OFDMA scheduler for the multi-user MAC framework for the IEEE 802.11ax networks. The scheduler works both in the downlink and uplink directions and assigns resource units to stations using a linear programming technique considering load of each client, possible resource unit configurations, modulation-coding scheme of each client, and ageing factor of each client’s load. The performance of the proposed scheduler has been evaluated using the NS3 simulator and compared against the legacy MAC layer mechanism of IEEE 802.11 protocol (i.e., DCF/EDCA). Simulation results show that our proposed throughput-maximizing scheduler increases the total throughput in the network as well as decrease the average end-to-end delay regardless of the number of stations connected to the access point by prioritizing the traffic of clients connected via high modulation-coding schemes.

Since tremendous amount of traffic is generated in modern networks as a result of the mobile devices and video streaming, the congestion issue is faced more frequently in the networks. Accordingly, failures and performance losses in networks due to congestion result in deteriorated Quality of Experience (QoE) from the end user perspective that may cause financial and reputation loss for the service provider. Even though the new video streaming paradigm, Dynamic Adaptive Streaming over HTTP (DASH), is proposed as a solution for the changing condition of the networks, it is not sufficient considering the heavily loaded links that show the symptoms of link failures. Therefore, the flexible implementation of the data plane fault tolerance scheme that can be applied for other problems like congestion in networks is crucial. Thus, in this study, we apply the data plane fault tolerance approach in the Software-Defined Network to improve the QoE of DASH clients in the case of congestion rather than the failure. To detect the congestion in the network level, we use the Bidirectional Forwarding Protocol (BFD) that is originally implemented for link failures. In our experiments, we investigate the effect of the BFD interval, video segment size, and traffic load on QoE parameters. Our results show that if the fault tolerance approach is applied using a small BFD interval with a large segment size, QoE parameters are noticeably enhanced considering the non-applied case.

Recent developments in Internet technology have led to an increased importance of network traffic classification. In this study, we used machine-learning methods for the identification of applications using network traffic classification. Contrary to existing studies, which classify applications into categories like FTP, Instant Messaging, etc., we tried to identify popular end-user applications such as Facebook, Twitter, Skype and many more individually. We are motivated by the fact that individual identification of applications is of high importance for network security, QoS enforcement, and trend analysis. For our tests, we used UNB ISCX Network Traffic dataset and our internal dataset, consisting of 14 and 13 well-known applications respectively. In our experiments, we evaluated four classification algorithms, namely J48, Random Forest, k-NN, and Bayes Net. With the complete set of 111 features, k-NN gave the best result for the ISCX Dataset as 93.94% of accuracy using the value of k as 1, and Random Forest gave the best result for the internal dataset as 90.87% of accuracy. During the course of this study, the initial numbers of features were successfully reduced to two sets of 12 features specific to each dataset without a compromise to the success. Moreover, we observed a 2% increase in the success rate for the internal dataset. We believe that individual application identification by applying machine-learning methods is a viable solution and currently we are investigating a two-tier approach to make it more resilient to in category confusion.

Recent developments in Internet technology have led to an increased importance of Software-Defined Networks (SDN). Due to advantages of this new network model that controls the network centrally, many service providers and vendors expect that traditional networks should be superseded by SDN. However, because of their centralized nature, they are vulnerable in terms of reliability and fault-tolerance issues both on data and control planes. Thus, developing such a fault-tolerant SDN design is quite important. In this study, fault tolerance on the data plane is targeted by considering various network and performance measurements. In the experiments, the impact of the topology size, frequency of packets, and the number of flows in the current route on the recovery time is tested. Moreover, local and global recovery approaches are compared.

Great hardware and software capabilities of mobile devices allow us to research new scientific fields. Activity recognition is one of the main research areas for smartphones. Built-in sensors of a standard smartphone, such as accelerometer, magnetometer, gyroscope, enable us to recognize the daily activities of a person. In this study, we focused on the window sizes and the sampling rates in order to observe how they affect to the accuracy and CPU utilization. For our test scenarios, we built a dataset including a tri-axial accelerometer sensor data of 7 daily activities: walking, jogging, sitting, lying, standing, walking upstairs and walking downstairs. We collected these activities with a sampling rate of 80 Hz by using 5 seconds window size. Then, we downsampled the collected data to 40 Hz, 20 Hz, 10 Hz, 5 Hz and 1 Hz by using 4, 3, 2, and 1 seconds window sizes, respectively. Thus, we could evaluate the variation of the accuracy and CPU utilization. Our test results showed that when the sampling rate increases, both the accuracy and the CPU utilization become greater. Moreover, we observed that, for a fixed sampling rate, when window size increases, CPU utilization decreases.