A popularity prediction and dynamic data replication study for the ATLAS distributed data management / by Thomas Beermann. Wuppertal, July 5, 2017
Content
- Abstract
- Acknowledgements
- Contents
- I Introduction
- 1 The LHC, ATLAS and the WLCG
- 2 ATLAS Computing Model and Systems
- 3 Motivation
- II Prediction
- 4 Popularity Prediction
- 4.1 Introduction
- 4.2 System Design
- 4.3 Prediction Methods
- 4.4 Neural Network Prediction
- 4.5 Hybrid Solution
- 4.6 Implementation
- 5 Prediction Evaluation
- 5.1 Prediction Input Data
- 5.2 Data Access Pattern Examples
- 5.2.1 Highly popular project / datatypes
- 5.2.2 Popular project / datatypes
- 5.2.3 Unpopular project / datatypes
- 5.3 Evaluation
- 5.3.1 Evaluation Metric
- 5.3.2 Static Prediction
- 5.3.3 Linear Prediction
- 5.3.4 Neural Networks
- 5.3.5 Summary
- 5.3.6 Hybrid Prediction
- 5.4 Conclusion
- III Redistribution
- 6 Data Redistribution
- 7 Grid Simulator
- 8 Redistribution Evaluation
- IV Summary
- V Appendix