在行動環境下以行動支援主機為主的多點因果遞送演算法

Abstract

在行動計算環境 (Mobile computing environment)下發展應用軟體是一 個蓬勃發展的趨勢。在這種環境下,原有的網路裝有行動支援主機 (Mobile support station, MSS),它可以透過無線電波直接與行動主機 (Mobile host, MH)通訊。相對於傳統的分散式系統,行動主機面臨了一些 新的限制: 較小的無線網路頻寬,有限的電源供應以及較低的計算能力。 對於許多的分散式軟體而言,多點因果遞送 (Causal multicast)是非常重 要的。在行動環境下,由於以上的限制,每一個動支援主機皆當作其所屬行 動主機之代理者,我們假設行動支援主機與行動主機之間的頻道 (channel)其訊息傳遞是依序的 (FIFO),因此,維持行動支援主機間的因果 遞送亦即維持了行動主機間的因果遞送,如此可以大量降低訊息空間之負 擔 (overhead)。我們的論文利用前述的想法,提出了以行動支援主機為主 (MSS-based)的多點因果遞送之演算法,它亦能處理動態改變的群組。這個 演算法不但僅需較小的訊息空間負擔,而且這訊息的大小與行動主機的數 目無關,我們認為這個演算法是具有可擴增性的(scalable)而且能很容易 處理主機連線/斷線(connection/ disconnection)的問題,因此它適合行 動計算的環境。 There is a growing trend in developing applications in mobilecomputingenvironment. In this environment, thestatic network is augmented with mobile support stations(MSSs)that couldcommunicate directly with the mobile hosts(MHs) via wireless media. Compared to the traditional distributed system, a mobile hosts in the systemfaces a new set of constraints: low bandwidth of wireless links, tightconstraints on power consumption and a significantly lower computing power. Causal multicast is required for several distributed applications.In mobilecomputing environment, to meet the constraints, each MSS acts as a proxy forits local MHs. We assume the wireless channel between an MSS and its local MHs is FIFO. Thus, maintaining causal ordering among MSSs alsomaintains causal ordering among MHs. Message space overhead canthus be reduced greatly. In this thesis, we presented an MSS-basedalgorithmthat uses this knowledge to enforce causal ordering among MHs. This algorithmalso handles dynamically changing groups. Message space overheadis not only small but is also independent of the number of mobile hosts.We claim that ouralgorithm is scalable and can handle it connections/disconnectionseasily. Hence,it is suitable for mobile computing environment.