網路可用頻寬量測工具研究

Abstract

在現今網路科技快速演進且網路相關應用日趨多元的環境裡，網路可用頻寬資訊對於各種網路應用而言是相當重要的資訊。常見的網路可用頻寬量測技術可分為探測速率模型(Probe Rate Model, PRM)以及探測間隔模型(Probe Gap Model, PGM)這兩大類型。使用PRM模型的實作雖然較PGM模型實作有較高的量測準確性，但量測所需的時間較長，並且由於其量測是以形成網路壅塞來推估當時的網路可用頻寬，因此並不適合使用於真實的網路世界。相對於PRM，PGM模型則具有輕量、快速、準確的特性，較適合用於真實世界的網路可用頻寬量測工作。

在本論文的研究中，我們選擇了採用PGM模型的網路可用頻寬量測工具—Spruce，設計包含不同網路拓樸、頻寬、背景流量類型、使用頻寬比例、參數等情境的模擬案例，對Spruce的行為特性進行深入的探討，同時藉此了解PGM模型的特性。並於觀察Spruce在不同情境下的量測表現之後，提出改善的建議，透過模擬案例實驗證實，我們的改良確實有效。

In recent years, information technology grows rapidly. More and more network applications are adopted as parts of our everyday life, including video streaming, overlay-route selection, QoS verification and P2P applications. People concern about the network available bandwidth because many of network applications need to know the information of network available bandwidth to adjust their behaviors. There are two kinds of network available bandwidth meansuremnet tools proposed in past years: PRM(Probe Rate Model) and PGM(Probe Gap Model). PRM based tools are intrusive and cost much to get the estimation results. So PRM is generally considered not sutable for real life networks, while PGM based tools are light weight and faster that can be used in the real world.

In our thesis, we choose a PGM based network available bandwidth meansuring tool—Spruce to study its behaviors and characteristics under various kinds of topologies, background traffics, bandwidth usages and scenarios to get well understanding on PGM. We also propose our improvements on Spruce and verify our improvements by NCTUns network simulator, the results show that our improvemnets are more accurate than the original Spruce.