Abstract
在即時資料庫的研究中，現有的即時排程都以集中式資料庫為研究環境，然而網際網路已發展成熟，分散式資料庫環境有其必要的研究目的，但若以現有的即時排程用於分散式的資料庫環境下，將產生兩個問題：一是現有的即時排程會給遠端交易較低的優先權，二是不能有效控制網路傳輸的延遲會導致即時排程有誤判的情形。因此我們提出FHR即時排程，適用於分散式資料庫的環境□，這個即時排程不但可解決現有即時排程面對分散式資料庫環境所產生的問題，同時也保存現有即時排程的優點，如下所示：(1)考慮到交易的價值是依時間變 動的特性，(2)考慮到交易間的價值衝突而以淨價值決定交易的價值，作為即時排程優先權 指定的基礎，(3)考慮到交易的報酬比例，而以交易剩下的執行時間最短作為第一考量，(4)可依照每個系統的環境，在交易的價值與交易的緊急程度間動態調整優先權大小。主要觀點是讓遠端交易擁有較高的優先權，也就是有較高的機會讓讓遠端交易在截止時間之前完成，而當地交易又可利用遠端交易在等待遠端資料的時間，獲得執行的機會。在我們建立的模擬分析環境，採用M.J. Carey所提的方式加以擴充，而傳輸延遲時間(communicationdelay) 方面，現今很多研究報告為了簡化模擬環境的複雜度均採用固定值，但是我們為了符合現有網路環境的情況，採用動態的傳輸延遲時間，FHR即時排程在模擬分析中都比起現有的即時排程表現優異，是一個適用於分散式資料庫環境下的即時排程。In real-time database research, the current real-time scheduleruses centrali-zed database as its research environment.However, with advance development oof the Internet, distributeddatabase environment also requires research. Two problems arisewhen using current real-time scheduler under distributed datab-ase environment. One problem is current real-time schedulergives low priorityto remote transaction; two is unable toeffectively control the communicationdelay through the network,thus it generates miscalculation priority of transaction inreal-time scheduler. The above two problems have led us tointrod-uce FHR scheduler.FHR scheduler is ideal in distributeddatabase environment.This real-time scheduler not only can solvethe problems faced incurrent real-time scheduler, but it alsocan maintain the advantage of thecurrent real-timescheduler.These advantages are listed below. (1) Itconsiders the propertyof time-variant value on the transaction. (2) It considers thevalue conflict in transactions. The transaction value isdetermined by the net value. This is the base for determiningpriority in real-time scheduler. (3) When calcul-atingtransaction reward ratio, the shortestremaining execution timeis the first consideration.(4) Priority is adjustedaccording tothe transaction value and urgency in each system. Theobjectiveis to give high priority to remote transaction. Thatis,to give higher chancefor remote transaction to be comple- tedbefore the dealine. however, local transaction can utilize thewaiting time of remote transaction to obtain exec-utionopportunity. In our simulation model, FHR scheduler outperformsthecurrent real-time scheduler. It is a comp-atible schedulerin distributeddatabase environment.