Adaptive Model Update Algorithms for Remote Network Emulation

Remote network emulation is an approach that utilizes a remote parallel simulator to improve the scale and accuracy of network emulation for general users with no local access to high performance computing facilities. The remote network emulation approach involves using a large scale detailed packet level simulation at a remote site in conjunction with a local network emulation to quickly provide QoS predictions to real time applications. Model update algorithms are needed to keep the remote simulator and local emulator updated with each other´s status, while keeping the amount of communication required to an acceptable level. In this paper two adaptive model update algorithms are explored for triggering network status updates. The first algorithm dynamically adjusts update frequency according to the real network status between the remote simulator and emulator. The second algorithm triggers updates whenever the prediction error of the network emulation exceeds a preset threshold. Preliminary experimental results show that the two algorithms can improve emulation accuracy compared with a fixed update algorithm for the scenarios that were tested.