There are a huge number of customers using Ericsson’s prepaid telecom charging system. This means that even the smallest optimization done in its management of load results in a big win when considering time and CPU usage. Ericsson wishes therefore to learn about how the load management can be improved considering speed, i.e. faster response to load changes.In this thesis work two subsystems of the prepaid telecom system are modeled. The first subsystem treats and prepares the telecom activity before it enters the second subsystem which checks if there is coverage for the action that the telecom activity asks for to be performed. A model which is an emulation of the subsystems is built using c++. It captures the timing of the real system not the logic or function. The c++ approach is compared to a Matlab approach which runs much slower in simulation. In addition, the model enables full view of the queue sizes and the rejects made by the regulator.Verification has been performed with self generated data based on synthetic test cases. Verification and analysis show that the model captures the intended behavior of load management. Information about different software parameters and its influence on the output is obtained. Different scenarios are tested and the outcome discussed.
Contents
1 INTRODUCTION
1.1 Background
1.2 Description of the problem
1.2.1 Output
1.2.2 Input
1.3 Objective
1.4 Method
1.5 Report Overview
2 DESCRIPTION OF CHARGING SYSTEM
2.1 Overview
2.2 Queues
2.3 INAP and TH
2.4 Traffic
2.4.1 Priority
2.5 Output handler
2.6 Regulation
2.6.1 Queue method
2.6.2 Latency time method
2.7 System configuration
2.7.1 CPU
2.7.2 Thread
2.7.3 Scheduling
2.8 Capacity
2.9 Simulation of the telecom charging system
2.9.1 Input
2.9.2 Output
2.10 Bottleneck
3 MODELING CHOICES
3.1 Modeling approaches
3.1.1 Simevents
3.1.2 Simulink and RealTime Workshop
3.1.3 C++
3.2 Limitations
3.2.1 Time of request
3.2.2 No external calls
3.2.3 Priority and Money control
3.2.4 Number of CPUs
3.3 Data generator
3.3.1 100% even method
3.3.2 Burst method
3.3.3 Poisson method
3.4 Scheduling and Time slices
4 VERIFICATION APPROACH
4.1 Basic verification cases
4.2 Real life situations
5 RESULT
5.1 Load at 75%
5.2 The behavior of the queue
5.3 Regulation
5.4 Input grouping
5.5 Thread distribution
6 CONCLUSIONS
6.1 Design
6.1.1 Matlab
6.1.2 C++
6.2 Knowledge of behavior
6.2.1 Regulation method
6.3 Analysis
6.4 Future work
7 REFERENCES
APPENDIX A, GLOSSARY
Mathworks
Session & request
APPENDIX B, POISSON DISTRIBUTION
Example
APPENDIX C, SIMULATIONS
Measurements
Soccer Goal
New Years Eve
Hardware Error
CPU
99%
Comments
APPENDIX D, SPECIFICATIONS
Author: Bjerre, Johan
Source: Linköping University
Download URL 2: Visit Now