Packet-optical node transceiver frequency allocation.
In an optical network scenario which consists of multiple nodes (whiteboxes) at its edges and ROADMs in-between, the coherent transceiver average laser configuration time is improved. The process is evaluated according to a testbed setup. This is facilitated in the appropriate lab equipment (or via simulation when required).
For that purpose, a software agent (Netconf server) residing at the whiteboxes, is developed receiving input from the Software-Defined Networking (SDN) packet controller (PacketCTL - a Netconf client). Then, configuration of the local transceiver laser frequencies of the controlled pluggable devices takes place, for facilitating the connectivity in-between the ROADM network. Also, the agent records and reports back telemetry data (feedback) which is used by the PacketCTL's resource-allocating mechanism to improve efficiency within the network topology.
PacketCTL is able to improve the overall control plane configuration time, by utilising an internal Feed-forward Neural Network (FNN) which is based on Q-Learning (QL). This Reinforcement Learning (RL) mechanism adapts to the current operating environment by efficiently allocating frequency slots to the pluggables, decreasing the average laser configuration times. To achieve this, it exploits the previous configuration times as feedback input while in learning state.
First, compile wopen_sim.cpp. This is the simple simulator that configures the laser frequency of the pluggable. Next, install both the pyang library (e.g., v2.6.1) and the plugin pyangbind (e.g., v0.8.6). This represents the YANG instance by a Python class instance. Replace paths with the correct local ones.
% pyang --plugindir /opt/homebrew/lib/python3.12/site-packages/pyangbind/plugin -f pybind \ -o /stuff/coding/astrape_nc/modules/pluggable_config.py \ /stuff/coding/astrape_nc/modules/pluggable-config@2024-11-19.yang
The procedure of setting a specific laser frequency on a pluggable device is depicted as a sequence in the next figure. Initially, a request is generated from the HrCTL, including a pair of whiteboxes. The local DB of PacketCTL contains previous feedback values (configuration times) per transceiver ID in each whitebox. An RL method based on QL is applied and the transceiver frequency slots are chosen to fulfil the request. Next, it launches system threads to connect concurrently to both whiteboxes via the Netconf protocol, using an XML RPC edit-config structure that contains the pluggable ID and the chosen slot. After successful configuration, the whitebox returns embedded in the RPC XML OK message, the actual configuration time as feedback. Finally, the controller updates its local DB accordingly.
Configure the system, train the ML agent (multiple attempts to produce efficient results), launch some optical nodes with separate shell processes, execute the packet controller. Connectivity is realised via system socket files.
0.1 2025-03-03 Initial release.