دانلود رایگان مقاله یک روش طراحی شبکه تکامل پذیر با تنوع توپولوژیک

abstract

As environments surrounding the Internet become more changeable, a design approach that requires less equipment to scale up networks against the traffic growth arising from various environmental changes is needed. Here, we propose an evolvable network design approach where network equipment is deployed without a predetermined purpose. We enhance topological diversity in the network design by minimizing the mutual information. Evaluations show that, compared to networks built with ad-hoc design method, networks constructed by our design approach can efficiently use network equipment in various environments. Moreover, we show that, even considering the physical lengths of links, the approach of increasing topological diversity can lead to an evolvable network.

6. Conclusion and future work

In this paper, we have proposed a design approach (the EVN design approach), based on minimizing mutual information, to strengthen topological diversity and make an evolvable network. We have shown that a network grown using our design approach can grow with less capacity than a network grown using a method based on the FKP model. Furthermore, we have shown that capacity introduced for one environment can be used in another environment, so a network grown using our design approach shows an overlap between equipment placement in an old environment and that in a new one. We also showed that the random attachment method could not increase the topological diversity compared to the EVN design approach, so that a topology designed by the EVN design approach could have a high reuse ratio under unexpected environmental change scenarios. Although the EVN design approach did not consider the physical lengths of links, we showed that, even considering the physical lengths of links, an approach that increases topological diversity can lead to an evolvable network. Several problems are left for future research. First, further evaluation of the EVN design approach is needed. In the simulations in this paper, we only add two links for each additional node in order to keep the average degree similar to that of the initial topology. However, there are also other cases in practice. Although we believe that the topology will also be diverse and evolvable when adding three or more links for a node, this should be investigated by further simulation. Second, because the order of the EVN design approach is O(n2 · d2), where n is the number of nodes and d is the degree, there is a scalability problem. However, because the purpose is to enhance topological diversity, strict minimization may not be needed. Approximate solutions can be considered in future work. Third, analytical investigation is required to provide a clearer discussion of the evolvability of networks designed using our approach in response to several other unexpected environmental changes. Lastly, we have considered topological diversity here, but diversity at a higher-level, such as the diversity of link capacity distribution or processing capacity of nodes may help to improve evolvability.