2022 Fiscal Year Final Research Report
Network Topology-Aware Service Chaining in NFV Network
Project/Area Number |
21K21288
|
Research Category |
Grant-in-Aid for Research Activity Start-up
|
Allocation Type | Multi-year Fund |
Review Section |
1001:Information science, computer engineering, and related fields
|
Research Institution | Nara Institute of Science and Technology |
Principal Investigator |
Hara Takanori 奈良先端科学技術大学院大学, 先端科学技術研究科, 助教 (70907881)
|
Project Period (FY) |
2021-08-30 – 2023-03-31
|
Keywords | ネットワーク機能仮想化 / サービスチェイニング / 強化学習 / グラフニューラルネットワーク / 容量制約付き最短経路ツアー問題 |
Outline of Final Research Achievements |
This research project established (1) mathematical optimization and (2) machine learning based service chaining (SC) to realize the topology-aware SC in NFV networks. As for mathematical optimization, the Lagrangian heuristics has been proposed to efficiently solve the capacitated shortest path tour problem (CSPTP) based SC and function placement. Representative results showed that the Lagrangian heuristics achieves much smaller execution time while holding almost the same performance as the integer linear program. As for machine learning, the CSPTP-based SC adaptive to changes of service demand and network topology has been proposed by collaborating graph neural networks and deep reinforcement learning. Representative results showed that the proposed scheme achieves the efficient resource allocation adaptive to the changes of both service demand and network topology.
|
Free Research Field |
情報通信
|
Academic Significance and Societal Importance of the Research Achievements |
従来のSC問題では,最短経路問題を背景とした解法であったが,本研究課題では,VNFの実行順序の保証,単一のパスが同一リンクやノードを複数回利用するといったSC特有の課題に対して,CSPTPを導入することで,資源効率性と計算時間の観点から,CSPTPによるオンライン型SCの実現可能性を示した.更に,GNNを適用した学習モデルを適用することで,未知のネットワークトポロジやサービス需要の変化に追随可能なサービスチェイニングの実現性を示した.これらの成果はネットワークトポロジと特徴の効率的な利用方法の理解を深めるとともに,ネットワーク分野における資源割当や故障検知に貢献できると期待される.
|