![Coordination between control layer AI and on-board AI in optical transport networks [Invited]](https://ksra.eu/wp-content/uploads/2020/08/Artificial-Superintelligence-In-Control-1024x536.jpg)
Abstract
In optical transport networks, the urgent demand for control efficiency and intelligence has become one of the most significant challenges for telecom operators. With the development in control technology, more attention has been paid to performance enhancement of the centralized controller of software-defined optical networks (SDONs). Meanwhile, machine learning (ML) is emerging as a promising technology to facilitate the intelligence of control planes in SDONs. Some research works have been conducted to use ML to solve problems in optical transport networks. However, it is still a challenge to deploy and use computing resources. On the one hand, computing resources can be deployed inside the centralized controller of an SDON to enable control layer artificial intelligence (AI). On the other hand, computing resources can also be deployed on the hardware board to enable onboard AI. The two-layer AI functions are able to meet different intelligent requirements in data and control layers in different scenarios. Therefore, coordination between them is an important issue. In this paper, a novel control architecture based on an SDON is proposed, and it can support control layer AI and on-board AI simultaneously. Particularly, on-board AI is proposed based on edge computing to support various ML applications. To evaluate the proposed architecture, we develop an experimental testbed and demonstrate a typical use case, i.e., alarm information prediction. Experimental results show that coordination and cross-layer optimization between control layer AI and on-board AI can be achieved. However, there is much space for research in this area, and we envision some open issues.
-
IEEE Keywords
- Artificial intelligence ,
- Optical fiber networks,
- Computer architecture,
- Maximum likelihood estimation ,
- Protocols,
- Biomedical optical imaging
Introduction
With the breakthrough of artificial intelligence (AI) in recent years [1,2], AI has been applied in many fields, such as computer science, finance, trade, medicine, diagnosis, heavy industry, transportation, etc. As a representative technique ofAI, machine learning (ML) has become a hot topic recently.ML was initially studied in 1988 [3], mainly targeting pattern recognition and computing learning theory. Now, ML is being widely used in various fields. In the field of image classification, for example, ML achieves and further exceeds human-level recognition capability [4–6]. In gaming, AlphaGo [2], developed by Google’s DeepMind Research Group, beat the world’s champion. Meanwhile, ML has also recently been used in data mining [7] and natural language processing [8].In optical transport networks, ML is emerging as an advanced tool for dealing with complex issues from two different perspectives. For optical transmission, ML is capable of estimating fiber linear/nonlinear impairment according to signal and device parameters [9]. For example, both unsupervised and supervised ML methods are applied to improve the performance of the optical communication system based on the nonlinear Fourier transform [10]. ML can also be used to estimate the nonlinear noise and thereby monitor the optical signal-to-noise ratio (OSNR) [11]. Fiber-inducedintra- and inter-channel nonlinearities are tackled using blind nonlinear equalization by unsupervised ML-based clustering∼46 Gb/s single-channel and∼20 Gb/s multichannel coherent multicarrier signals [12]. The pre-distortion scheme based on a clustering algorithm of ML can mitigate nonlinear impairments for a VLC system [13]. For optical networks, ML can deal with issues such as traffic load prediction, failure location, and resource allocation [14]. For example, the network management system can flexibly manage resources in the backbone IP/optical network by using machine learning to predict traffic flows in the short- and long-term [15]. A deep-learning-based failure prediction algorithm is proposed, which constructs the data set based on data augmentation for data training [16]. A resource-allocation method based on reinforcement learning is proposed for multimodal optical networks [17]. Besides, ML can also detect intrusions in the control plane of a software-defined optical network (SDON)[18]. Simulation results show that the accuracy of an intrusion detection scheme can reach more than 85%. The research shows that ML can achieve good performance in dealing with different problems in optical networks.
However, ways to deploy AI functions in optical transport networks remain an open issue. Specifically, how to implement training, testing, and application of the AI model are three major issues. First, in the training process, the ML engine needs to be fed with a large amount of data, which means that the components that have ML capability require storage resources to save data sets and computing resources to update the parameters of the models (such as the forwarding transmission of the neural network). Second, for the testing process, because the testing data set is much smaller than the training data set, the testing process does not require too many storages and computing resources. Finally, in some cases, AI com-ponent should have the ability to make real-time responses. Therefore, AI modules are usually located on a resource-rich device such as the central controller of an SDON, which can be considered as control layer AI. However, the AI module of the central controller part cannot deal with the local problems of each network element efficiently unless all equipment can report their local data to the central controller in real-time. Because such many-to-one synchronization will incur a heavy workload to the central controller, the controller side AI is not an ideal place for solving equipment-side problems. Thus, it is necessary to deploy AI capability into the data plane in a distributed way. Up to now, few works have studied how to deploy AI functions in optical transport nodes to enable equipment-level applications.
This paper introduces the concept of deploying on-boardAI into optical networks to enable AI in transport network equipment. The architecture of on-board AI is designed based on edge computing, and the collaboration mode between control layer AI and on-board AI is studied. Control layer AIis deployed with an SDON controller, and onboard AI is deployed with optical transport equipment. An experimental testbed is built to evaluate performance between the two-layer AI. On-board AI is performed on a circuit board within AI-specific chip, along with other necessary components like memory, express card, etc. We also study the potential applications based on onboard AI, even the coordination between control layer AI and on-board AI. This paper is an extension of [19]. The major new contribution of this paper is that we describe the process of training/testing the AI model and the method of selecting the best model from the AI model library in the collaboration mode between control layer AI and-board AI in more detail. Meanwhile, we develop an experimental test platform for evaluating the proposed collaboration mode. The experimental test platform takes the ML-based traffic prediction function as an example. The embedded-board DP-8020 is used as the onboard AI to evaluate the efficiency of the training/testing model process in the collaboration mode. The experimental results prove that the proposed collaboration mode can improve the efficiency of AI control functions in optical networks. The rest of this paper is organized as follows. In Section 2, we introduce the architecture of self-optimized optical net-works (SOONs), which is designed based on SDONs and supports AI. Detailed architecture is also given to show how the two-layer AI is deployed. In Section 3, we introduce the collaboration mode between the control layer AI and on-boardAI. In Section 4, we perform experiments in two scenarios and discuss the results. We first build an experimental testbed to evaluate the network architecture we proposed. We also demonstrate a typical use case. Section 5 discusses the open issues in the proposed network architecture, and Section 6offers a conclusion.
Conclusion
In this paper, we discuss on-board AI based on edge computing together with AI-based on an SDON controller. Both should be conducted collaboratively to complete cross-layer optimization. We integrate onboard AI into optical transport networks to enable the deployment of equipment-level applications. Then, we propose a collaboration mode of joint optimization between control layer AI and on-board AI. The mode completes the data collection, ML algorithm model selection, and optimization through the interaction between the SDONcontroller and the device. The experimental results show that the computational efficiency of the model training is improved by 12.5% through the collaborative mode of the control layerAI and the on-board AI. Taking the ML application of the alarm prediction as an example, the correct prediction rate of the alarm in the network can reach as high as 99%.
About KSRA
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:
”Coordination”]
Bibliography
author
Year
2020
Title
Coordination between control layer AI and on-board AI in optical transport networks [Invited]
Publish in
in IEEE/OSA Journal of Optical Communications and Networking, vol. 12, no. 1, pp. A49-A57, January 2020,
Doi
10.1364/JOCN.12.000A49.
PDF reference and original file: Click here
Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.
Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.
