完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.author | 徐英傑 | en_US |
dc.contributor.author | Shyu, Ing-Jye | en_US |
dc.contributor.author | 謝續平 | en_US |
dc.contributor.author | Shiuh-Pyng Shieh | en_US |
dc.date.accessioned | 2014-12-12T02:18:35Z | - |
dc.date.available | 2014-12-12T02:18:35Z | - |
dc.date.issued | 1997 | en_US |
dc.identifier.uri | http://140.113.39.130/cdrfb3/record/nctu/#NT860392003 | en_US |
dc.identifier.uri | http://hdl.handle.net/11536/62730 | - |
dc.description.abstract | 視訊伺服器乃是結合高網路頻寬需求,高磁碟機容量與高系統運算效能的 產物,以提供給使用者的隨選的視訊資料,其視訊資料具有連續性及不可 中斷性的串流特性,因此必須提供有效的容錯能力以滿足此一特性。在視 訊伺服器的容錯設計上,我們的研究主要著重在如何利用視訊伺服器群的 組成來達成具有透明性的容錯修復能力。視訊伺服器本身控制著儲存系統 與網路系統之間資料流向的即時性與連續性,一旦發生錯誤,就會影響整 個系統服務的品質。因此,我們在論文中提出一個分散式的視訊修復機制 ,提供給伺服器透明性的錯誤修復能力,持續的提供給使用者不間斷的視 訊串流。此外,在伺服器群的架構下,延伸而來的問題便是如何平均分配 工作,一則必須考慮在所有伺服器正常運作下,如何平均分配工作至每一 個伺服器。二則在當有伺服器發生錯誤時,也必須能夠有效的均分修復的 負擔至其他正常的伺服器。因此,在我們論文中也提出一個在視訊伺服器 群下節目工作分派的方法,使其能夠有效的達成上述兩點的要求。此外, 為了提供透明性與平滑的錯誤修復能力,每一個伺服器必須要能夠取得所 有在其他伺服器中目前節目進行的資訊。因此,我們也提出一分散式的次 序決定演算法來產生所有節目間彼此的次序與時間關係,此演算法並保證 所有伺服器所取得資訊是相同的,配合先前所提之分散式視訊修復機制, 提供最佳的分散式容錯與修復能力。 A clustered video system consists of multiple video servers which areconnected by a high speed network. While designing a video system, it isimportant to provide the non-stop video streams to clients in the presence ofserver failures. This dissertation studies the transparent recovery of aclustered video system upon a server failure. We propose a distributedplayback recovery scheme to recover the interrupted playbacks caused by aserver failure. Thus, the video system can still continue delivering videostreams to the interrupted clients. To balance the playback load both in theserver normal state and in server degradation state, we propose a playbackdispatch method which guarantees that the playback load are balanced amongthe servers in the server normal mode, and the recovery load are also evenlydistributed to the rest of the surviving servers when a server fails.To provide smooth recovery, each server needs to acquire the otherservers' playback information. For this reason, we propose a distributedorder decision algorithm which determines the timing and order informationof all distributed playbacks, and ensures every server having the same copyof this information. Once a serve fails, all surviving servers perform theproposed distributed playback recovery algorithm which uses the orderinformation to accomplish the optimal and smooth recovery. | zh_TW |
dc.language.iso | zh_TW | en_US |
dc.subject | 視訊伺服器 | zh_TW |
dc.subject | 容錯設計 | zh_TW |
dc.subject | 錯誤修復 | zh_TW |
dc.subject | 多媒體系統 | zh_TW |
dc.subject | 分散式系統 | zh_TW |
dc.subject | Video servers | en_US |
dc.subject | fault tolerance | en_US |
dc.subject | fault recovery | en_US |
dc.subject | multimedia systems | en_US |
dc.subject | distributed systems | en_US |
dc.title | 視訊伺服器群之容錯與平衡工作負載設計 | zh_TW |
dc.title | On Designing a Fault-Tolerant and Load-Balancing Clustered Video System | en_US |
dc.type | Thesis | en_US |
dc.contributor.department | 資訊科學與工程研究所 | zh_TW |
顯示於類別: | 畢業論文 |