|
Hokkaido University Collection of Scholarly and Academic Papers >
Theses >
博士 （情報科学） >
Studies on Adaptive Routing for a Wide-Area Overlay Network System
|
Files in This Item:
|
|
|
Related Items in HUSCAP:
|
| Title: | Studies on Adaptive Routing for a Wide-Area Overlay Network System |
| Other Titles: | 広域オーバーレイネットワークのための適応的経路制御に関する研究 |
| Authors: | 柏崎, 礼生1 Browse this author |
| Authors(alt): | Kashiwazaki, Hiroki1 |
| Issue Date: | 25-Mar-2014 |
| Publisher: | Hokkaido University |
| Abstract: | This paper proposes and evaluates two types of trafficengineering algorithms.The first, REI, and its derivative, NREI, are decentralized stochastic methods that have autonomousadaptability to network traffic conditions.When a routing node has some different paths to a givendestination, these paths can be evaluated in terms oftheir latency (delay time) information,which will be given from a feedback packet sent backfrom a destination node to a source node after anoutbound packet reaches the destination node.This algorithm is based on a one-way delay algorithm.NREI was developed from the delay time based adaptiverouting algorithm, REI, and is enforced for packetlosses caused by network congestion.Based on a path evaluation using latency and penaltyscores from lost packets, every node works as adistributed autonomous agent for adaptive routing.Through network simulations comparing these algorithmsto conventional and enhanced OSPFs and evaluationson an IP network of a research test bed, it is shownthat a multi-agent based routing algorithm has betteradaptability and scalability to congested path avoidance and network load balancing.In addition, the other approach proposed is a centralized deterministic method that uses a discreteevent simulator.This method searches for a suboptimal combination of paths from the enormous number of combinations ofpaths determined from pairs of all source nodes andall destination nodes.Cloud computing resources are used to evaluate thesehuge combinations rapidly, by which we can utilizepowerful computing resources.This method was tested on an ns-2 simulator with eight-and eleven-node networks.The results show that the method has better adaptabilityand high-speed performance for cross-traffic avoidance.Finally, implementation using OpenFlow is consideredand herein a hybrid model consisting of the two proposedalgorithms is discussed. |
| Conffering University: | 北海道大学 |
| Degree Report Number: | 乙第6907号 |
| Degree Level: | 博士 |
| Degree Discipline: | 情報科学 |
| Examination Committee Members: | (主査) 教授 高井 昌彰, 教授 栗原 正仁, 教授 水田 正弘, 教授 棟朝 雅晴, 教授 山本 雅人 |
| Degree Affiliation: | 情報科学研究科 |
| Type: | theses (doctoral) |
| URI: | http://hdl.handle.net/2115/55690 |
| Appears in Collections: | 学位論文 (Theses) > 博士 （情報科学）
論文博士 (Doctorate by way of Dissertation) > 情報科学院(Graduate School of Information Science and Technology)
|
|
