METRO-HAUL: Network Slicing for Improving Public Safety

Metro-Haul aims at architecting and designing cost-effective, energy efficient, agile and programmable metro networks that are scalable for 5G access and future requirements, encompassing the design of all-optical metro nodes, including full compute and storage capabilities, which interface effectively with both 5G access and multi-Tbit/s elastic core networks.

Demo Overview

The Network Slicing for Improving Public Safety Demo intends to showcase low-latency object detection and tracking in a multi-layer metro optical network scenario for the realization of a public safety use-case. This experimental demonstration presents a system for the flexible deployment of a network slice instance, implemented in terms of an ETSI NFV Network Service, in order to support the high-bandwidth low latency public safety application over a next-generation metro network.

In this demonstration, security is provided by real-time object recognition and tracking. Simultaneous real-time access to the data of fixed and controllable PTZ (pan-tilt-zoom) cameras allows tracking of objects and persons as well as the automatic recognition of critical events, which encompass areas larger than the field of view of a single camera. Dynamically provided network slices with high bandwidth will enable transmission of the video footage of several hundreds of cameras, while the optimum distribution of video analytics computation and storage will reduce the perceived latency.

Architecture

The scope of the project is the metro area, with typically 20 to 200 km distance between the nodes. Metro core edge nodes (MCEN) are connected to the core network, while access metro edge nodes (AMEN) are connected in a line to form a horseshoe. At the AMENs, edge data centers (DC) are located, to enable time critical computations close to the end-user, while less latency-sensitive and larger storage requiring functions can be placed in regional or core DC. For the demonstration, a semi-filterless architecture is chosen, where add/drop nodes are realised by splitters and couplers with wavelength blockers (WB) in between. Besides enabling high data rates, e.g. 200 Gbit/s net rate at ~30 GBaud with dual polarisation (DP) 16QAM, coherent transmission allows the filterless selection and detection of dropped channels by only adjusting the wavelength of the local oscillator laser. The transponders integrated in the demonstration is a commercially available CloudConnect QuadFlexTM dual-wavelength ADVA transponder, capable of 100 Gbit/s to 200 Gbit/s per wavelength.It features an OpenConfig interface for software defined control, while vendor B deploys an OpenConfig agent developed within the project. WB based ROADMs will be included based on LCoS technology and EDFAs to cope with the loss of the WB and the transmission fiber. The LCoS-based device will be integrated in ADVA’s open line system (OLS). The ADVA OLS offers a north bound Transport API interface (TAPI).

Deployment

A smart city video surveillance application requires multiple IP cameras distributed acorss the entire city and connected to several recording servers. One sever might have 100 to  250 cameras connected, either fixed or thermal cameras to enable surveillance at night and PTZ cameras to follow objects. In a city-wide installation, the controlling hard and software will be located in a central data centre. In a larger urban area, this central data center will not be in the same network node as the servers but distributed across a city and connected by the optical metro network. The servers act as core system slaves (CSS) for the video management system and device managers (DM) to control the cameras. They are connected to a core system master (CSM) server, for management and video analytics. A control centre client requests video footage from the individual servers and controls cameras manually. Live or archive video could be requested from multiple servers simultaneously, requiring high bandwith through the optical network. For automated and manual control of the cameras, triggered by the analytics or the user, a low latency connection is inevitalble. Analytics can range from simple motion detection to face and even behaviour recognition. All these functions of video surveillance require various amounts of bandwidth, storage and computational power.

Results

Metro-Haul has successfully demonstrated the benefits of network slicing for improving public safety. Specifically, the project has demonstrated the full workflow of planning, orchestration, deployment, and running a network slice including compute resources as well as connectivity one over an edge-computing enabled metro optical networking test-bed at Fraunhofer HHI. The demo starts with a service request from the network planer towards the OSM orchestrator. The orchestrator then requests two virtual network function (VNF), which run different pieces of the video management system including the video analytics engine and data manager, as well as a connectivity slice from the Parent Controller. Then VNFs are deployed by the OpenStack instances running on the two edge computing enabled nodes (i.e., MCEN and AMEN). Moreover, the controller requests optical network connectivity from the SDN controller (i.e., ONOS) of the network. ONOS proceeds to requests the connectivity from the OLS controller, which manages the ROADMs, and the two Quadfelx optical transponders. As soon as the compute and network slices run, the video management system initiate and the surveillance system triggers. WE measured different parameters including the end-to-end service establishment time, end-to-end latency in the connectivity layer (L2 switch 1 -> optical transponder 1 -> metro network -> optical transponder 2 -> L2 switch 2), considering two different lengths of optical fibers, as well as the accuracy of the video analytics engines running on the deployed VNF and the PTZ camera. 

5G Empowerment

This demonstration showcased that future edge-computing enabled by metro optical networks can provide high capacity connectivity across metro networks with low latency that allows, for instance, distributing different functions of a video management system that could manage hundreds of a camera across a city or beyond.

One of the key requirements to perform intelligent video surveillance services enabled by video analytics capabilities is the availability of powerful computation resources in the surveillance zone. However, this is not the practically viable solution in most cases. Therefore, the computational task should be outsourced to some remote Edge Compute Node and the analytics results should then become available in the surveillance zone within a fraction of second. In addition, transporting several high-definition video streams across a metro network necessitates the availability of high bandwidth connectivity that can be easily supported by optical networks. This is exactly the focus of this demo, where we show the potential that can be brought about by 5G networks supported by smart metro optical network.

Location:  Berlin (Germany)

Dates: Q3-2020

Vertical Partners: See-Tech

Partners involved: BT, TIM, CTTC, UPC, CNIT, NAUDIT, OLC, LEX, HHI, TUE, TEI, ADVA, NOKIA, STEC

METRO-HAUL Website | @MetroHaul