• 沒有找到結果。

Dynamic Circumstance

在文檔中 中 華 大 學 (頁 37-41)

7 Performance Evaluation

7.2 Dynamic Circumstance

圖十六為比對

RLM 與 ACE [12]在 S 為 40,平均節點度數 4~10,ACE 的 h 值為 1~2,最

佳化次數為1 的情況下的效果比較圖,圖中菱形曲線為 RLM,正方形曲線為 RLM 加上 cache,

三角形曲線為

ACE h=1,叉叉曲線為 ACE h=2,圖中,RLM 表現的比 ACE h=1 好,但當 ACE

h 值調高時,則會表現的比 RLM 好,但是所花費的成本(overhead)將會大幅的增加,這邊 簡單的分析,RLM 複雜度為O(N*平均節點度數),ACE 的 h 為 1 時做 1-hop 的區域最小生成 樹,每個節點傳送鄰居表給1-hop 鄰居,複雜度為 O(N*平均節點度數),h 為 2 時做 2-hop 的 區域最小生成樹,資料交換改為每個節點將鄰居表傳送到2-hop 鄰居,複雜度為 O(N*平均節 點度數2 ),以及較多的資料計算,當 S 大於 30,ACE h=1 時,RLM 表現的比 ACE 好。

10 20 30 40 50 60 70 80 90 100

degree

query%

RLM 87.99 76.86 68.53 60.25

RLM+cache 84.99 75.14 67.34 59.35

ACE h=1 96.28 89.17 78.89 67.2

ACE h=2 55.05 26.63 16.60 11.96

4 6 8 10

Figure16: Performance evaluation of RLM and ACE

表平均節點度數為 6 時使用淹沒法,三角形點曲線代表平均節點度數為 10 時使用淹沒法,

菱形點曲線代表平均節點度數為 6 時使用 RLM,叉叉曲線代表平均節點度數為 10 時使用

RLM。圖中淹沒法的訊問訊息數量會慢慢的成長到一個門檻值,是因為節點度數小於平均節

點度數的節點離開後,新加入節點度數為平均節點度數的節點,導致總連線數會慢慢的增 加。不論總連線數如何增加,RLM 的詢問訊息量維持不變,表示了 RLM 的在動態的環境下 一樣有不錯的效果。圖十七(b)中顯示在不同的平均節點度數下淹沒法與 RLM 的總回應時 間,淹沒法所產生的總回應時間為100%,菱形點曲線代表平均節點度數為 6,正方形點曲線 代表平均節點度數為10。不論平均節點度數為 10 或 6,在 10 分鐘以前,總回應時間比例都

會慢慢的變長,這並不是

RLM 總回應時間變長,而是因為總連線數的增加,使得用淹沒法

得到的總回應時間變短。相比較後,RLM 的總回應時間與淹沒法比較起來就變的比較長,而 到了10 分鐘後,總連線數達到門檻值,並且不在劇烈的變化,總回應時間呈現穩定的狀態。

1000500 15002000 25003000 35004000 45005000 55006000 65007000 75008000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

time(minute)

number of query

flood of degree6 flood of degree 10 RLM of degree6 RLM of degree10

(a)

100105 110115 120125 130135 140145 150155 160165

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

time(minute)

response time%

RLM of degree10 RLM of degree6

(b)

Chapter 8

Conclusions and Future work

P2P 網路的目標就是讓所有的客戶端都能提供資源,包括頻寬,存儲空間和計算能力。

P2P 網路透過在多節點上複製數據,增加了防故障的健壯性,網路中的節點不需要依靠中心 索引伺服器來收尋檔案。並且P2P 網路不會出現單點崩潰的現象。因此 P2P 網路近年來已成 為一種新的分散式計算模型。

但分散式非結構化P2P 網路中的資源搜索往往導致大量冗餘的通訊負擔,造成網路頻寬 的佔用,使網路的效能無法被充分的利用。對覆蓋網路進行優化,可以大幅的改善上述問題。

使得網路頻寬得到解放,以及更有效的利用,甚至加快檔案收尋速度。

在本篇論文中,我們提出了一個完全分散式的改善技術,Redundant Link Minimization (RLM),減少 P2P 網路中多餘的詢問訊息。RLM 的主要作法是利用鄰居集合將網路節點進行 分群,根據群組後的網路拓樸,判斷出多餘的邏輯網路連線,進而建構出最佳化的網路拓樸。

除此之外,RLM 亦可以保證網路中檔案搜尋的涵蓋率。根據實驗結果,RLM 所產生的詢 問訊息數量,在最好的情況下,只有淹沒法的 11%。在網路連線稠密度越高的網路中,RLM 改善的效果越顯著。結合檔案索引快取的機制,RLM 可以再降低網路的訊息量與縮短回傳時 間。在動態環境中,RLM 的效能表現的與靜態環境一樣好,並且還可以避免因為關節點的離 開,導致網路搜尋涵蓋率減少的問題。在未來的研究方面,我們將致力於改善訊息回應時間,

使

RLM 能更適應真實的動態網路。

References

[1] Gnutella Network Size, http://www.limewire.com/index.jsp/size

[2] The Gnutella protocol specification 0.6, http://rfc-gnutella.sourceforge.net

[3] Stefan Saroiu, P. Krishna Gummadi and Steven D. Gribble, ”A Measurement Study of Peer-to-Peer File Sharing Systems”, Proceedings of Multimedia Computing and Networking, Vol. 4673, pp. 156-170, 2002.

[4] KaZaA, http://www.kazaa.com

[5] I. Stoica, R. Morris, D. Karger, M. F. Kaashoek and H. Balakrishnan, “Chord: A Scalable Peer-to-Peer Lookup Service for Internet Applications”, Proceedings of IEEE/ACM

Transactions on Networking, Vol. 11, No. 1, pp. 17-32, 2001.

[6] Ai Wu, Xinsong Liu and Kejian Liu, ”Efficient flooding in peer-to-peer networks”, Proceedings of International Conference of Computer-Aided Industrial Design and

Conceptual Design, pp. 1-6, 2006.

[7] Wen Tsuen Chen, Chi Hong Chao and Jeng Long Chiang, ”An Interested-based Architecture for Peer-to-Peer Network Systems”, Proceedings of International Conference of Advanced

Information Networking and Applications, Vol. 1, pp. 707-712, 2006.

[8] Lakshmish Ramaswamy, Gedik B and L. Liu, ”A Distributed Approach to Node Clustering in Decentralized Peer-to-Peer Networks”, Proceedings of IEEE Transactions on Parallel and

Distributed Systems,Vol. 16, No. 9, pp. 814-829, 2005.

[9] Alberto Montresor, ”A Robust Protocol For Building Superpeer Overlay Topologies”, Proceedings of the Fourth International Conference on Peer-to-Peer Computing, pp. 202-209, 2004.

[10] Kunwadee Sripanidkulchai, Bruce Maggs and Hui Zhang, ”Efficient Content Location Using Interest-Based Locality in Peer-to-Peer Systems”, Proceedings of INFOCOM 2003, Vol. 3, pp. 2166-2176, 2003.

[11] Juan Li and Son Vuong, ”ECSP: An Efficient Clustered Super-Peer Architecture for P2P Networks”, Proceedings of International Conference of Advanced Information Networking

and Applications, Vol. 1, pp. 278-283, 2004.

[12] Yunhao Liu, Zhenyun Zhuang, Xiao Li and Lionel M. Ni, ”A distributed approach to solving overlay mismatching problem”, Proceedings of the 24th International Conference on

Distributed Computing Systems, pp. 132-139, 2004.

在文檔中 中 華 大 學 (頁 37-41)

相關文件