Lease Time的預測及調整 — 以Jini為基礎的研究

• Main Authors: Chun-Fu Cheng, 程俊富
• Other Authors: Hsing Mei
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/93277086072845635987

碩士 === 輔仁大學 === 資訊工程學系 === 90 === Lease is a time-based mechanism that allows services to be granted, modified, or denied. A user can negotiate with service provider to get a lease. After a user gets a period of time granted by the service provider, that user has the right to get the service. Later, a user may send renewal messages to service provider when the previous granted lease is expired. Service providers may not know how long user will be served. Sometimes, even the leasing clients may not know how long they want to be served. In this thesis, we proposed a mechanism to predict the suitable lease length for different type of services. If a short lease time is granted, the leasing client will request to renew many times in order to finish the service. This will in turn increase the system overhead. If a long lease time is granted, resources will occupy by certain clients exclusively for a longer period of time. Hence, we suggest a dynamic adjustable lease scheme (DALS) to predict suitable lease time for the leasing clients. The filter extreme history approach can get a balance between first response and execution time, total waiting time, and total number of renewal in most cases. Through our testbed measurements, the filter extreme history approach can reduce up to 70% on total waiting time than maximum history approach, 19% than average history approach, and 44% than most frequently history approach.