Flooding topology is a subset of initial network's links, which can be used by flooding algorithms to send broadcast messages. Use of such topology can significantly reduce network load. From mathematical viewpoint flooding topology is a spanning subgraph of the initial network's graph with specific properties and constraints, such as minimal or close to the minimal number of edges, bounded vertex degree, bounded diameter and additional connectivity properties. In this paper one of the possible mathematical formulations of the flooding topology problem is introduced using graph-theoretic approach. Different parameters of the flooding topology are investigated and examples of possible tradeoffs between them are presented. Based on the introduced mathematical problem formulation several flooding topology algorithms were designed and their prototypes were created and tested on the typical network topologies, such as fat trees, Clos networks, spine-leaf topology and VL2 topology. Simulation results showed that proposed algorithms significantly reduce flooding for all considered topologies, while respecting other properties of the flooding topology in most cases.
The load balancing problem in a heterogeneous distributed computing system is considered. The problem is mathematically formulated as a stochastic game. As a multi-agent system, the heterogeneous distributed computing system can be viewed as a system with trainable agents at each computational node. There are some computational nodes that receive tasks according to a pre-established distribution (Uniform, Poisson). Tasks must be distributed among computational nodes in such a way that the average load on each computational node's resources (CPU, RAM, Storage) is the same. Based on the Proximal Policy Optimization (PPO) algorithm, a multi-agent reinforcement learning algorithm has been developed to solve the stochastic game. As a result of the experimental results, the proposed algorithm is capable of solving the formulated problem and is stable under various scenarios of computing system changes.
Healthcare data is generated and stored by various devices, often physically distributed, with a strong need for privacy. While applications can be custom-built to access data from these diverse sources, big data middleware can provide integrated interfaces for such data sources, avoiding repeated custom-code development. High-performance interfaces are essential for healthcare data access while adopting best practices governing the security, data privacy, and access control. These include user authentication, data authorization, role-based access control, and audit mechanisms to avoid compromising the security standards mandated by data providers. This paper presents a middleware framework for unified healthcare data access, federating data sources, and providing uniform REST interfaces. We evaluate its performance and functionality, featuring its extensibility and modular architecture with illustrative use cases.
Spanning trees are used in networking to determine routes and corresponding next hops. Topology changes due to link failure or recovery cause spanning trees to be rebuilt on each router. After the new spanning tree is calculated, the router updates its tables, where number of affected entries depends on the new tree and may be significant for large-scale network. Possible solution to minimize routing table update time is the indirect next hop approach in which we divide network hosts into clusters and split routing table into two, where the first one matches destination for its cluster and redirects to the second table that contains mapping between clusters and next hops appointed to them. In this paper we evaluate the performance of two algorithms aimed at minimizing spanning tree recalculation time and measure how many entries in routing table need to be updated. We also propose clustering algorithms that divide network hosts into clusters so that topology changes are likely to affect the next hop for the whole cluster, thereby minimizing changes in the routing table.
This paper has considered an architecture for building the Industrial Internet of Things (IIoT) using edge computing technology. The levels of this architecture and their respective roles are described, as are the benefits of implementing Mobile Edge Computing (MEC) in IIoT, the model of the considered smart factory, and the case in which the application must be migrated from the cloud to the edge device. As part of the explanation for the selection of technologies used in coordinating migration for a network with edge computing, the concepts of designing software based on a microservice architecture were considered, as well as virtualization and containerization technologies. The Docker containerization technology and the Kubernetes orchestration system are explored in depth to facilitate the migration of cloud-based applications to the peripheral.
