資料量龐大是三維 (3D) 資料如今還未在網路上被廣為使用的主因之一。雖然近來各種壓縮法、漸進式傳送法，或格式標準，已漸漸使「三維資料串流」(3D streaming) 能有效的即時傳送3D資料，但在實際應用中，系統上線人數是否具「擴展性」(scalability)，是至今還未受正視的重要問題。我們認為，由於 3D 內容本質上就資料龐大及具互動性，傳統的主從式架構 (client-server)，基於伺服器之資源上限及成本考量，恐怕無法真正提供具網際網路般規模的服務。 我們因此提議用同儕式網路 (peer-to-peer networks) 來支援 3D 資料串流，以分散伺服器的資源消耗。因 3D 內容為非線性式存取，同儕式3D資料串流 (P2P 3D streaming) 之主要議題和現今的多媒體串流不同，而需要新的探討。我們亦提出第一個以同儕式網路為基礎的 3D 資料串流架構，稱為 FLoD。它允許 3D 虛擬地球或 3D 虛擬環境的用戶端，直接透過其它使用者來取得相關資料，使伺服器的資源使用可最小化。為了展示 FLoD 如何用於現實環境，我們也建了個雛型系統，將原本以 JPEG 2000 為基礎的 3D 物件串流改由同儕式網路傳送。實驗顯示，伺服器端的資源用量因此可大幅降低；而模擬也顯示，同儕式 3D 資料串流本質上較主從式架構更有擴展性。此外，原始 FLoD 模式的一些改良設計，也將在本論文中介紹。 本論文透過歸納 3D 資料串流的需求及挑戰、及替現有作法分類，點出 3D 資料串流的主要問題及設計空間為何。我們亦在最後說明待解的議題，以期促進未來相關系統的設計與探討。 One of the most serious issues holding back the widespread of 3D content on Internet has been their inaccessibility due to large data volume. Many compression and progressive transmission techniques, as well as format standards, have been proposed in recent years to make 3D streaming increasingly viable for the efficient and accessible delivery of 3D content. However, existing proposals have yet to seriously address one of the most important issues in practical adoption -- a system's scalability in terms of the number of concurrent users. We argue that due to 3D content's large data volume and interactive nature, client-server architecture, with its inherently fixed resource availability and high cost, will not be suitable to support popular Internet-scale 3D streaming. We propose the use of peer-to-peer (P2P) networks for 3D streaming, and argue that due to the non-linear access patterns of 3D content, P2P 3D streaming is a new class of applications apart from existing media streaming and requires new investigations. We also present FLoD, the first P2P 3D streaming framework that allows clients of 3D virtual globe or virtual environment (VE) applications to obtain relevant data from other clients while minimizing server resource usage. To demonstrate how FLoD applies to real-world scenarios, we build a prototype system that adapts JPEG 2000-based 3D mesh streaming for P2P delivery. Experiments show that server-side bandwidth usage can thus be reduced, while simulations indicate that P2P 3D streaming is fundamentally more scalable than client-server approaches. Certain improvements and extensions of the original FLoD framework are also presented. In this dissertation, we define both the problem and solution spaces for P2P 3D streaming, by outlining its requirements, challenges, and categorizing existing proposals. Open questions are also identified to facilitate the design of future systems.
