Smart controller over resource nodes
Download as PPTX, PDF
0 likes257 views
The document discusses a smart controller for managing resource nodes across clusters. The controller monitors nodes, performs cluster analysis, and enables switching between clusters. It monitors nodes using hierarchical and flat clustering. Cluster analysis involves prediction, communication, aggregation, and transmission phases to measure capacity, dependencies, and balance load. The controller can identify clusters based on membership profiles and study migratory behavior to predict switching between clusters.
Smart controller over resource nodes
- 1. Smart Controller Over Resource Nodes Ghazal Tashakor
- 2. Agenda Deployment Model Controller 1. Monitoring Nodes 2. Cluster Analysis 3. Switching in a multi-cluster graph References
- 3. Deployment Model Scheduling Jobs Controller Resource Discovery Aggregation nodes Leaf nodes Content-based Clustering/ XML Clustering Job1 Job2 Job3 Free Free Free Free Free Free Free Free Free Free Busy Busy TransmittingTimetotheBrokers. Transmittingphase AggregationTime CommunicationTime Communication phase Collaborative/Fuzzy Clustering Aggregation phase 1. Monitoring nodes (Predictors) 2. Cluster Analysis 3. Switching in multi-cluster graphs Load Balancing Resource Allocation Event Processing Job Management Brokers
- 4. Controller Cloud and Edge Resource nodes in collaboration and communication to work together needs a controller to do the smart cluster analysis and take the final decision for their workload elasticity. Also it needs monitoring strategy before analysis and Switching ability after that.
- 5. 1.Monitoring Nodes in: Multi-cluster combination model Multi-clustering techniques: 1.Hierarchical clustering for Cluster-head nodes. (Vertically) 2.Flat Clustering for non-head cluster nodes (Horizontally)
- 6. 2. Cluster Analysis phases 1. Prediction Phase: Cluster-head nodes 2.Communication phase: Receiving packets from non- head cluster nodes (Connectivity and mobility issues) 3. Aggregation Phase : Capacity and dependencies measurement for selected nodes/resources ( For load balancing and event processing issues) 4.Transmitting Phase: Transmit the aggregated information to the Brokers.
- 7. 3. Switching in a multi-cluster graph Identification clustering based on membership profiles could be a good method in standard fuzzy clustering algorithm to detect the user migration by checking the node migration between clusters. This is possible by studying the migratory behavior of user/nodes over time and achieve the statistical approaches for predicting the path.
- 8. References i. Switching regression models and fuzzy clustering, RJ Hathaway, JC Bezdek - Fuzzy Systems, IEEE Transactions , 1993 ii. Relative entropy collaborative fuzzy clustering method, M Zarinbal, MHF Zarandi, IB Turksen - Pattern Recognition, 2015 Elsevier iii. Collaborative clustering with the use of Fuzzy C-Means and its quantification, W Pedrycz, P Rai - Fuzzy Sets and Systems, 2008 Elsevier iv. Detecting the migration of mobile service customers using fuzzy clustering I Bose, X Chen - Information & Management, 2015 Elsevier v. Node Similarity-based Graph Clustering and Visualization M Erd辿lyi, J Abonyi - 7th International Symposium of Hungarian , 2006 vi. Node Similarity-based Graph Clustering and Visualization M Erd辿lyi, J Abonyi - 7th International Symposium of Hungarian , 2006 - researchgate.net vii. XML clustering: a review of structural approaches M Piernik, D Brzezinski, T Morzy - The Knowledge , 2015 - Cambridge Univ Press viii. Combining multiple clustering systems C Boulis, M Ostendorf - Knowledge Discovery in Databases: PKDD 2004, 2004 Springer ix. Toward Semantic XML Clustering. A Tagarelli, S Greco - Sdm, 2006 x. A cluster validity index for fuzzy clustering KL Wu, MS Yang - Pattern Recognition Letters, 2005 Elsevier xi. A Hierarchical Model for Minimum Entropy Data Partitioning W Wu, AB Lee, D Mumford 2003 xii. https://en.wikipedia.org/wiki/Hamming_distance