1 |
基於Zookeeper服務的對稱式高可靠度電子交易系統架構 / Symmetric and High-Available Electronic Trading System Base on Zookeeper常澤民, Chang, Tse-Min Unknown Date (has links)
隨著電子商務的蓬勃發展,不論是使用者或是服務提供者,對於系統的可靠度(reliability),越來越為重視,從『能提供服務就好』,到要能提供『不間斷的服務』。這中間的轉變,推動了目前許多系統架構或產品運用在電子商務的領域。小從磁碟陣列(RAID)、資料庫鏡像(Database Mirror)、叢集(Cluster)架構,異地備援(Disaster Recovery Site),大到企業永續運作計畫(Business Continuity Planning-BCP)的一環,都在尋求一個高可靠度的電子交易系統。
所謂高可靠度的系統,必須能夠在相對長的時間週期內,系統可以持續運作而不中斷,這個概念,同時也意味著系統異常時的恢復能力,以及容錯的機制。換句話說,就是當系統面臨異常或是災害發生時,如何能在最短的時間內恢復系統的正常運作,繼續提供使用者相同的服務。因此提高系統的可靠度,對使用者的具體意義而言,即是提高其服務的可用度(availability)。
本論文的研究,即運用Apache下自由軟體Zookeeper所提供的協調服務,設計一套透過傳播監控(propagation monitor)來達到容錯機制的對稱式多主機架構,使得系統即使在面對異常狀況發生時,仍能迅速恢復系統應有的功能,提供使用者不間斷的服務。並且利用免費的資料庫管理軟體-MariaDB來儲存及操作資料,以低廉的成本,建置具高可靠度的電子交易系統。 / Due to the blooming development of electronic commerce (E-Commerce) , both users and service providers put more and more emphasis on system reliability. Thus uninterrupted service has become the basic requirement of E-Commerce systems, which now have all adopted some highly avaialable system architectures and advanced products such as Redundant Array of Independent Disks (RAID), Database Mirror, Cluster architecture, Disaster Recovery Site(DR Site). Many E-Commerce vendors also prepare Business Continuity Planning(BCP) for a highly reliable electronic trading system.
The so-called high reliability systems must continue to operate without interruption in a relatively long period of time, and this concept also means the requirements of abnormal system resilience and the mechanism of fault tolerance. In other words, when the system is confronted with abnormal situations or disaster occurs, it is critical about how to restore the normal functioning of the system to continue to provide users with the same level of services in the shortest possible time, thereby increasing the reliability of the system in terms of the user's specific meaning, that is, to improve its service availability.
This thesis applies the coordination service provided by Zookeeper to develop a propagation monitor mechanism and a symmetrical multi-host fault-tolerant system architecture, making the system work even in the face of an abnormal situation. Indeed, the system should be able to quickly restore its functions, providing users uninterrupted service. Besides, we use the free database management software-MariaDB to store and operate the system data. Therefore, we are able to build a low-cost yet highly available electronic trading system.
|
2 |
選擇性接合資料庫中表現序列跳接的容錯樣式探勘彭興龍, Peng, Sing-Long Unknown Date (has links)
真核生物在遺傳資訊核糖核酸實際轉譯成蛋白質之前,可能受環境、序列上的特定二級結構、特定部分序列樣式……等影響,而製造出目的、功能不同的蛋白質,這項生物機制稱為選擇性接合。目前對於選擇性接合機制的形成原因、根據何項資訊作選擇性調控,尚未有全面性的研究足以判斷。本研究嘗試透過發展適當的資料探勘技術,分析大量核糖核酸序列,找出可能影響選擇性接合的序列樣式。
選擇性接合可分為七種類型,我們針對其中一類稱為跳接式選擇性接合的基因資料,根據分析該資料的特性,提出兩類型的容錯資料探勘方法與流程,分別是全序列樣式探勘與轉化重複結構樣式探勘。前者對發生跳接式選擇性接合的整段intron序列,找出所有容錯頻繁樣式。再利用Kum[18]等人提出的一致性序列樣式的近似探勘方法,找出足以代表同一群聚中所有頻繁容錯樣式的一致性序列樣式。
轉化重複結構樣式探勘的作法則是先找出intron序列的前後部分區段中,可能具有容錯轉化重複樣式的序列集合。再進行容錯頻繁樣式探勘與一致性序列樣式的近似探勘方法。由於轉化重複樣式是生物序列中常見的一種序列結構,可能透過該類型結構,影響跳接式選擇性接合的發生方式。因此利用這樣的探勘方法,我們可以找到可能的具重要決定性轉化重複結構樣式。
最後,我們對兩個選擇性接合資料集合Avatar-120和ISIS-54,進行全序列樣式探勘與轉化重複結構樣式探勘實驗,討論發掘出序列樣式的支持度及平均錯誤率。並進一步與Miriami[24]等人研究發表的兩個樣式比較,利用整體序列最佳並列排比,評估樣式間的差異性,以發掘出“新穎”的樣式。 / Before RNA sequences are translated into proteins, eukaryotes may produce different functional proteins from the same RNA sequences. It is due to influence of environment, second structure, specific substring pattern, etc. This mechanism is named alternative splicing. At present, there are still not enough research to judge causes and critical information of alternative splicing. We try to develop suitable data mining technologies to analyze large number of RNA sequences, and find out possible patterns affecting alternative splicing.
Basically, there are seven possible types of alternative splicing. We focus on “exon skipping” type. According to the analysis of exon skipping data, we propose two fault-tolerant data mining methods and procedures: “Full Sequence Pattern Mining (FSPM)” and “Inverted Repeat Pattern Mining (IRPM).” Full sequence pattern mining method can be applied to mine all fault-tolerant frequent substrings in the whole intron sequences, and then get consensus sequential patterns using ApproxMap method proposed by Kum[18].
Inverted repeat pattern mining method can be used to look for consenesus patterns with structure of inverted repeat. Because inverted repeat patterns are often appeared in biological sequences and such structural patterns may result in exon skipping. We could discover some important patterns by this method.
Finally, we mined patterns from two alternative splicing databsets “Avatar-120” and “ISIS-54”by above two proposed methods. The support and average fault number of mined patterns were discussed. These patterns were also used global alignment method as compared with two patterns (C / G-rich) discovered by Miriami[24]. Novel patterns measured by discrimination were reported.
|
Page generated in 0.022 seconds