The Multi-access Edge Computing (MEC) system interconnected by metro optical network is gaining much attention recently. The proximity of computing resources in MEC provides the possibility of low-latency access for user equipment (UE). However, since the computing capacity of edge server is limited, the task failure may occur due to the unacceptable computing latency if the corresponding server is overloaded. Note that there is more than one MEC server at the network edge and the computationally intensive tasks on the overloaded server can be sent to nearby light-loaded servers for processing via peer offloading. In this article, we design a latency-aware task peer offloading (LA-TPO) scheme, which is an economic solution for the overload problem of MEC. We formulate the LA-TPO into a mixed integer non-linear programming (MINLP) model and develop a GA-based heuristic algorithm to jointly optimize the offloading decisions and the routing and computing resource allocation. We compare the MINLP and proposed algorithm under various scenarios, and the results show that the GA-based LA-TPO can achieve low latency, high success ratio of tasks as well as the optimization of resource utilization.

The rapid growth in Internet traffic has contributed to the need to expand network transmission capacity. Multi-band (MB) using existing standard single-mode fibers (SMFs) in the free band is an ideal way to increase the capacity of the fiber. However, the introduction of MB has transferred space division multiplexing - elastic optical networks (SDM-EONs) to multi band-SDM-EONs (MB-SDM - EONs), and changed routing, spectrum and core assignment (RSCA) problem to the routing, spectrum, core and band assignment (RSCBA) problem. At the same time, it introduces a new issue named stimulated Raman scattering (SRS), which affects transmission quality. Although an expanded fiber can carry more services, a large number of services will be interrupted if a link fails, which will bring huge losses. Therefore, it is necessary to propose a reasonable protection strategy to ensure reliable transmission quality. In this paper, the survivable RSCBA problem in MB-SDM-EONs is studied, which copes with link failures and considers SRS in signal-to-noise ratio (SNR) analysis. In order to solve the problem, we propose a band partition protection scheme: working and protection resources are allocated in different frequency bands for increasing SNR in a fault-free network. We formulate this problem and band partition protection scheme as an integer linear programming (ILP) model, which takes into account inter-core crosstalk and SNR analysis. Furthermore, a heuristic algorithm based on genetic algorithm is proposed for large-scale networks. We evaluate the performance of the proposed approaches through experimental simulation. The results demonstrate that both proposed approaches are effective in finding the optimal solutions.

In recent years, hybrid fiber-wireless (Fi-Wi) networks that integrate optical network technology and wireless communication technology have received widespread attention. The technical integration in Fi-Wi networks provides users with the possibility to establish high-quality connections that guarantee flexibility and economy. However, there are heterogeneous structures and resource types between the subnets of the Fi-Wi network. This paper considers the segmented application of network slicing technology in the optical backbone and hybrid access parts of Fi-Wi networks. In this way, the seamless connection of wired and wireless networks can be realized, and meanwhile adapt to the multiple network service forms and differentiated service quality requirements introduced in the 5G network era. In this paper, we design a dynamic slicing (PDSM) scheme based on traffic prediction. This scheme includes all four procedures of service provisioning, which are network slice creation, traffic prediction, online reconfiguration of network slices and intra-slice routing and resource allocation based on the prediction results. It is more flexible than the traditional static slicing scheme in the scenario, where the network state changes dynamically. We develop a prediction-based dynamic slicing algorithm (PD-MSRA) with intra-slice multi-path routing and shared resource allocation for the entire workflow of PDSM. We evaluate the performance of PD-MSRA on a variety of network topologies, and also implement several comparison algorithms including static slicing strategies. Simulation results show that the PDSM scheme using PD-MSRA achieves low bandwidth blocking rate and the optimization of resource utilization.

To alleviate the performance interruption between elephant and mouse flows, a combined optical timeslot switching (OTS) and L2/L3 Top of Rack (ToR) architecture is experimentally demonstrated for the first time in a data center scenario. In this system, elephant flows are identified on the fly by a newly built sFlow-based traffic analyzer and delivered through a separate OTS subsystem. In this way, elephant flows are physically isolated from mouse flows; thus, fewer mutual interruptions are expected. However, previous OTS research has not sufficiently addressed how the timeslot allocator and timeslot allocation should be implemented to adapt to traffic fluctuations, and it is obvious that the current SDN-based timeslot allocation cannot satisfy the OTS system's requirements for time efficiency. Therefore, in this paper, data-plane development kit (DPDK) platform was used to enable an ultrafast online timeslot allocator (TSA) that can process timeslot requests at traffic levels exceeding 1 Tb/s. This demonstration employs two Intel Xeon CPU e5 logical cores for TSA, a private protocol between the TSA and servers and a novel timeslot allocation algorithm. Using the proposed architecture, mouse flows are seldom interrupted by elephant flows. The power penalty of the OTS system is less than 1.2 dB, which is acceptable for server-to-server interconnections. Furthermore, simulation results show that the system achieves a consistently lower end-to-end latency for all flows than does a traditional ToR-based intra-rack DCN.

With the continuous development of the Artificial Intelligence of Things, deep neural network (DNN) models require a larger amount of computing capacity. The emerging edge-cloud collaboration architecture in optical networks is proposed as an effective solution, which combines edge computing with cloud computing to provide faster response and reduce the cloud load for compute-intensive tasks. The multi-layered DNN model can be divided into subtasks that are offloaded to edge and cloud servers for computation in this architecture. In addition, as bearer networks for computing capacity, once a server or link in optical networks fails, a large amount of data can be lost, so the robust reliability of the edge-cloud collaborative optical networks is very important. To solve the above problems, we design a reliable adaptive edge-cloud collaborative DNN inference acceleration scheme (RACAI) combining computing and communication resources. We formulate the RACAI into a mixed integer linear programming model and develop a multi-agent deep reinforcement learning algorithm (MADRL-RACIA) to jointly optimize DNN task partitioning, offloading, and protection. The simulation results show that compared with the benchmark schemes, the proposed MADRL-RACIA can provide a guarantee of reliability for more tasks under latency constraints and reduce the blocking probability.

Multi-access edge computing is a promising technology for solving delay sensitive, computationally intensive applications. It aids users with low-latency access by deploying computing and storage resources close to the user side. However, as the computing resources of edge servers are relatively limited, directly offloading tasks to neighboring edge servers can easily overload them. Tasks that arrive at the overloaded edge server will be blocked due to latency exceeding its threshold. To make better use of edge servers in the metro optical network to support delay sensitive and computationally intensive tasks, we divide the tasks into two categories and model them. Partial offloading tasks can be distributed to multiple edge servers for cooperative computing, while binary offloading tasks can be offloaded to a lightly loaded edge server as a whole. In this paper, we put forward the problem of task optimal allocation. Furthermore, we design and implement a heuristic algorithm for cooperative offloading of dependency-aware tasks based on genetic algorithms. The algorithm jointly optimizes the allocation of routing and computing resources. Simulation results show that our proposed algorithm has superior performance.