1 .. This work is licensed under a Creative Commons Attribution 4.0 International License.
2 .. http://creativecommons.org/licenses/by/4.0
3 .. (c) Anuket, The Linux Foundation, BIT Mesra, VTU and Others.
6 ==============================
7 AI/ML Models for NFV Usecases.
8 ==============================
10 This document describes all the models created by Anuket-Thoth project.
12 *********************************
13 1. Failure prediction (FP) models
14 *********************************
16 a. Summary of the VM Failure-Prediction models.
17 ===============================================
18 We have developed Neural Network models for predicting failures in
19 virtual machines (VMs) used in network function virtualization (NFV)
20 environments by analysing VNF data. The data used to build these models
21 are provided by Orange Labs, and the VMs are based on project Clearwater.
23 The links for the data are:
25 * Processed Data: https://drive.google.com/drive/folders/1crrVZMJwf00MP5qM7nmVEqFsatOAShla
26 * Raw: https://drive.google.com/file/d/1QVipyoWPD1_4W_QXWzxEla4b88EWo5X5/view?usp=drivesdk
27 * Raw Source: https://www.kaggle.com/datasets/imenbenyahia/clearwatervnf-virtual-ip-multimedia-ip-system
30 These models are found under *models* directory. In the below table, only the jupyter-notebooks reference is given, which can be found in *models/failure_prediction/jnotebooks* folder . The corresponding python file can be found in *models/failure_prediction/python* folder.
32 .. list-table:: Summary of Failure Prediction Models.
43 - Simplest (implementation-wise) case.
47 - Convolutional Neural Network. Poorest among the Neural-network based models.
51 - Basic Long Short-Term Memory. Better than CNN.
54 - LSTM_attention.ipynb
55 - The attention mechanism distributes weights accordingly. Performance is similar to correlation LSTM.
58 - LSTM_correlation.ipynb
60 * - Correlation with Stacked LSTM
62 - stacked_LSTM_correlation.ipynb
63 - Second best performance.
64 * - Correlation with Bi-LSTM
66 - Bi_LSTMstacked_LSTM_Correlation.ipynb
67 - Third best performance.