Synthetic Data Generation for Network Placement Using Heuristic-Based Simulation
Open Access Depositedtraining an effective model requires high-quality placements, yet such placements are difficult to obtain without a trained model.This work addresses this training data gap by introducing a simulation framework for generating synthetic problem-solution pairs for service placement. By leveraging a range of naive and heuristic placement algorithms, the framework constructs a labeled dataset of service-to-network mappings, annotated with resource usage metrics such as CPU load and bandwidth consumption. These synthetic examples serve as training data for generative models aimed at learning ideal service placements. This work presents a scalable and customizable framework for generating training data that supports the development of generative models for network placement.
Determining how to optimally allocate service components within a fixed network topology, known as the node placement problem, is a critical problem within networking. Determining the ideal placement to optimize bandwidth usage and CPU requirements is an example of combinatorial optimization, a problem class known to be NP-hard due to the exponential number of placement possibilities. Recently, learning-based approaches, particularly generative models, have shown promise in approximating effective placement strategies. [1] However, there is a circular dependency issue
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Grb_gwu_0075M_17344.pdf | 2025-07-20 | Open Access |
|
|
Final_topology_generator.py | 2025-07-20 | Open Access |
|
|
|
sample-config_copy | 2025-07-20 | Open Access |
|