Decentralized Architecture for Energy-aware Service Assembly

This repository contains the simulation code to replicate the results of the experimantation of the paper "Decentralized Architecture for Energy-aware Service Assembly" accepted at ECSA 2020.

Instructions

1. Download the Java Project "Energy-Aware-Service-Assembly" and import it in your IDE as a Java project
2. Link the provided libraries in "ext-lib" to the project
3. Input the program argument "configs/mida-assembly-config.txt" (i.e., the configuration file of PeerSim)
4. The main class to run the experiments is "peersim.Simulator"

Configuration Parameters

The file configs/mida-assembly-config.txt contains the configuration parameters for the simulation. The main parameters are:

• NETWORK_SIZE: The number of services to assemble
• TYPES: Number of types of services
• M: learning window of the algorithm proposed in [1]
• STRATEGY: The selection criteria that the nodes adopt. We implemented multiple strategies:
  1. emergent (QoS-aware with learning - see [1])
  2. random
  3. greedy (Greedy on QoS)
  4. shaerf (QoS-aware with learning - see [2])
  5. individual_energy (Energy-aware Local - ECSA2020)
  6. overall_energy (Energy-aware Overall - ECSA2020)
  7. fair_energy (Energy-aware Learning - ECSA2020)

References

[1] M. D’Angelo, M. Caporuscio, V. Grassi, and R. Mirandola, ‘Decentralized learning for self-adaptive QoS-aware service assembly’, Future generations computer systems, vol. 108, pp. 210–227, 2020. https://doi.org/10.1016/j.future.2020.02.027.

[2] A. Schaerf, Y. Shoham, and M. Tennenholtz. 1995. Adaptive load balancing: a study in multi-agent learning. J. Artif. Int. Res. 2, 1 (August 1994), 475–500.

Contact

If you have additional questions do not hesitate to contact me: https://lnu.se/en/staff/mirko.dangelo/