Anomaly Detection for Nodes Under the Cloud Computing Environment

2.1. Node Importance

A large number of research results (Malliaros et al., 2016; Yu et al., 2008; Zhao et al., 2009) show that the importance of nodes represents the degree of impact on node's performance when it failures.

Many researches have proposed related computing methods for node importance evaluation. Zhao Y.H et al. defined the Node Importance Contribution Matrix (NICM) based on the interrelationship between interconnected nodes. A node importance evaluation algorithm combining NICM and intermediary was proposed (Zhao et al., 2009). Liang Y.Y et al. proposed an algorithm for mining key nodes in a directed network based on association relationships (Liang et al., 2017). Ren Z.M et al. comprehensively considered the number of neighbors of a node and the closeness of connections between its neighbors, and proposed a node importance evaluation method based on neighbor information and clustering coefficient (Ren et al., 2013).

The above researches are all based on the mining of important nodes or key nodes. The nodes with high importance are more likely to be abnormal under the cloud computing environment in the case of abnormal occurrence. The nodes with high importance can be detected in time. The nodes with low importance have low probability of abnormal occurrence and have little impact on the cloud environment. So it is necessary to filter less important nodes to speed up anomaly detection.