Fault Detection Mechanism In The Cloud Environment

Cloud computing is the product of distributed processing, parallel processing and grid computing. It has been one of the most popular technologies and brings a profound influence to our life.Based on the history of the cloud computing, the thesis introduces the Hadoop and its tolerance mechanism based on the Google cloud environment. The Hadoop is open source platform of MapReduce. After summarizing the usual heartbeat expiry time algorithms we propose an adaptive algorithm based on weight, which assumes that the arrival time of the heartbeat information almost remain unchanged within a short time. In this thesis we take use of linear prediction model, of which the data is little and the computation is simple. The algorithm takes use of the linear prediction model to reduce the amount of data and simplify the operation. The expiry time is related to weight coefficient which varies in a certain range. The experiment verifies the effectiveness of the algorithm.The thesis analyses collaborative fault detection mechanism based on the cloud computing environment. Due to the high heterogeneity and the high dispersed location, the cloud computing seems easier to make mistakes compared with the traditional computing platform. The thesis designs chord ring and collaborative judgment method voted by multiple groups, which can be used in different cloud environment. The former divides nodes into three types according to the different characteristics of every node in group. We choose the group’s credibility as final judgment standard. Unlike the other methods, the Chord ring model forms a full connection no longer between each node but take the form of judgment one by one. First, each node is coded by DHT (Distributed Hash Table) and the adjacent one will be neighbor. Every node only checks the adjacent neighbor, which leads to reliability and robustness of the cloud computing system.