以實驗為依據對隨機線性網路編碼非均等抹除保護能力之探討

Abstract

在這篇論文中，我們將探討如何使用線性的隨機網路編碼技術 (Random Linear Network Codes, RLNC) 如何達成非均勻的資料抹除保護機制 (Unequal Erasure Protection, UEP)。在眾多的資料流中，有些子分流與其他的相比之下有較高的優先權。為了對這些相對較重要的分流有比較安全的保護機制，必須要選擇性停止某些網路中的傳輸中繼端點的資料傳輸。這個編碼方式可量化為分離矩陣(Separation Vector)；並在實驗中更進一步採用一夠大的整數有限域來提升解碼的成功率，而產生並推導這些編碼方式的可行性可透過圖型理論來解決。 除此之外，此研究的最大貢獻是以 Java 語言實做出UEP-RLNC 的資料傳輸系統。並設計實驗模擬網路中有可能遇到的傳輸問題如封包遺失，觀察此編碼方式如何能面對常遇到的傳輸問題。實驗結果主要為顯示在網路環境不穩定的情景下，網路終端的接收點收錄並成功解碼封包所占的比例。

We investigate in this thesis how Random Linear Network Codes can achieve Unequal Erasure Protection. In many data streams, some layers of data have priority over the others. In order to enhance protection of particular layers with RLNC, we need to nullify local encoding coefficients at some specific nodes. The resulting coding schemes can be quantified by using Separation Vector. Furthermore, granted a finite Field large enough to have high theoretical decoding probability, evaluating and generating those coding schemes can be achieved using Graph Theory. Besides this model, our main contribution is an actual implementation of UEP-RLNC in JAVA. Some simulations were performed to verify the validity of our UEP mechanism as well as its quantification. The results obtained show the profiles of retrieval rate with respect to packet error loss for both hotspot losses and background losses scenarios.