Article Preview
Top1. Introduction
The remarkable power of swarm-intelligence lies in a coordination of all individuals and communication of “knowledge” without any supervisor. Every individual in the population makes local decisions, and acts in a decentralized manner. Using different natural mechanisms of social interaction, the individual agents reach intelligent solutions. From another point of view, in highly dynamic distributed systems, autonomous agents interact without a central control. Therefore, swarm-inspired intelligence could help highly dynamic systems to cope with environmental changes.
The problem of efficient routing nowadays faces many challenges and copes with highly dynamic nature of the Internet. The main task of routing algorithms is to solve the problem of path selection when sending information from one node (source) to another (destination) over multiple hops within a network (Buford & Yu, 2010). An efficient routing1 is especially an important topic in case of unstructured P2P networks, since no global view on the network exists and no address mapping is maintained. A delivery of data packets is neither guaranteed nor bound to a specific upper limit of hops (Buford & Yu, 2010). Therefore, it requires an advanced approach that is able to manage and solve the above-mentioned problems in an autonomous, intelligent manner and that is sufficiently adaptable. Unstructured P2P overlay networks support dynamics very well, but their “week point” is scalability. Therefore, we start with the main research question that concerns the efficacy of an intelligent routing in unstructured P2P overlay networks.
In this paper, we propose the usage of an intelligent routing based on two swarm-mechanisms: the lifecycle of slime molds Dictyostelium discoideum (Dd) and the bee foraging in fully unstructured P2P overlay networks. Our decision to choose these swarm inspired mechanisms in particular was based on the fact that they already showed potential and additionally, the slime molds do not compute the shortest path per se, but the optimal path for the amount of resources involved (Adamatzky, 2015; Šešum-Čavić et al., 2016), which is important for the path-optimization process in routing. The proposed solution combines the unstructured P2P and space-based computing (Kühn, Craß, Joskowicz, Marek, & Scheller, 2013) for intelligent routing, i.e., the routing in unstructured P2P overlay network is realized by using swarm-inspired intelligence, whereas space-based computing is used for the implementation of (sub)spaces in the overlay network serving as an asynchronous communication medium. As the focus is put on the routing algorithms, the underlying software architecture is only very briefly mentioned. Namely, the diversity of routing algorithms and the resulting substantial amount of different options for their implementation make routing algorithms hard to compare fairly. Existing popular network simulators, even when focused on P2P networks offer only a general environment to simulate distributed systems and do not provide a generic abstracted pattern for benchmarking routing algorithms. Our underlying software architecture for unstructured P2P networks enables the fair and systematic benchmarking and comparison of routing algorithms by providing a meaningful component-based abstraction in a form of coordination patterns (Kühn, 2016). It benchmarks routing algorithms in a generic manner, stripped from specific areas of applications, and supports the easy exchangeability of routing algorithms (simply through “plugging”) and different network topology settings through configuration. Note that the framework itself does not solve the routing problem, but serves as a necessary basement for the used algorithms and abstracts the general requirements. Software agents play the roles of artificial species. A pallet of three swarm intelligence-based algorithms (AntNet, BeeHive, slime molds Physarum polycephalum) and two conventional approaches (Gnutella and k-Random Walker) are adapted for unstructured P2P networks, plugged-in and compared with each other.
The novelty and contribution of this paper include: