| Summary: | 碩士 === 國立中央大學 === 資訊工程學系 === 101 === With the rapid development of Internet technology, there have been many innovative network architectures. Software Define Network (SDN), which has been adopted by a number of leading enterprises, has been greatly valued over the last few years. And OpenFlow is one of the mainstream Software Define Network architectures. However, as SDN develops, the existing Internet facility has become insufficient to meet its demand. Standford University in the US has therefore developed NetFPGA, a programmable NIC, to support OpenFlow.
Unlike traditional network architecture switches, the state of OpenFlow switches is no longer invariable, allowing suppliers to manage the networks more conveniently. One of the significant features of OpenFlow is that it has dynamic memory capacity, so it can record flow entries and modify it through an approach similar to learning. OpenFlow can bring great benefit to network experiments and management, but the advantages wouldn’t be brought to full play without a good mechanism for monitoring system states. Hence there are still issues and problems that need to be investigated and modified.
This dissertation discusses classic distributed system Global snapshot algorithms, including Chandy-Lamport algorithm, Lai-Yang algorithm, and Mattern algorithms, as well as GENE-VIOLIN snapshot algorithm, which applies the mentioned algorithms on OpenFlow. Analysis and design of a comprehensive Global snapshot mechanism that is applicable to the OpenFlow network architecture are also included here.
|
|---|
