Results of the DYNAMO (Dynamic MEC Orchestration of Cellular Networks) experiment in the Fed4FIRE+ testbeds

The main objective of the DYNAMO Fed4FIRE+ experiment was to perform Network Function Virtualization (NFV) Management and Network Orchestration (MANO) of a cellular network on top of cloud infrastructures, exploring one of the key enabling technologies for 5G systems and beyond. DYNAMO used cloud and radio access facilities at the IRIS testbed and cloud facilities at the University of Vigo (UVIGO) to deploy an end-to-end (E2E) cellular network and perform elastic changes on it if needed. The geographic distance in between facilitated the setup of a realistic Multi-Access Edge Computing (MEC) use case, where the virtual Evolved Packet Core (vEPC) was deployed at UVIGO (Spain) and the access network, i.e., the User Equipment (UE), the e-Node-B (eNB) and edge cloud, were implemented on IRIS testbed (Ireland). While the initial deployment of the E2E cellular network may be considered as static, DYNAMO showcases the elasticity that an E2E cellular network may need in runtime. Hence, we presented a use case consisting of a latency sensitive E2E cellular network (network slice), where the endpoint of the UE connection was initially located in the core (UVIGO) but then migrated to the edge (IRIS), in case the UE's latency ranges were unacceptable. In this regard, the UE reported the experienced latency to Open Network Automation Platform (ONAP), which is responsible to trigger specific policy-driven control actions if a predefined Service-Level Agreement (SLA) is violated. This datased includes the reports provided by the UE to ONAP. As a result, the endpoint of the data plane of the UE is automatically moved to the access network (IRIS) thus reducing significantly the latency for the UE. For the access part of the network, we implemented one srsLTE e-Node-B (eNB), one srsLTE User Equipment (UE) and a Devstack (Edge Cloud) in virtual machines on IRIS testbed. In addition, we also implemented an SDN switch controlled by an ONOS SDN controller. For the core part of the network, we considered a disaggregated vEPC from Open Air Interface (OAI) on a Devstack (Core Cloud) at UVIGO. DYNAMO has succeeded in the integration of a broad set of network elements and technologies between the two different domains (UVIGO and IRIS testbed) and fulfilled all initial objectives: (i) establishing communication between ONAP and IRIS testbed to deploy generic VNFs on core and edge clouds, (ii), deployment of an E2E cellular network with UE and eNB in IRIS and the vEPC at UVIGO, (iii), sending telemetry of the UE to ONAP and (iv) designing and testing closed-loop control actions in ONAP to migrate the data plane of the UE to the edge in case of unsatisfactorily SLA. DYNAMO paves the way to a broad set of future 5G experiments that will require resource orchestration, such as the deployment of network slices or the automatic scheduling of services in the limited resources of Edge Clouds. This repository contains the information sent from the UE to ONAP, in order to decide if the latency between the UE and the PGW is OK or if an action has to be considered to reduce such latency.DYNAMO Fed4FIRE+实验的核心目标，是在云基础设施之上开展蜂窝网络的网络功能虚拟化（Network Function Virtualization, NFV）管理与网络编排（Network Orchestration, MANO）工作，探索面向5G及后续演进系统的关键使能技术之一。本项目依托IRIS试验床的云资源与无线接入设施，以及维戈大学（University of Vigo, UVIGO）的云基础设施，部署了一套端到端（end-to-end, E2E）蜂窝网络，并可根据实际需求对网络进行弹性调整。两地的地理跨度为搭建真实的多接入边缘计算（Multi-Access Edge Computing, MEC）用例提供了理想条件：虚拟演进分组核心（virtual Evolved Packet Core, vEPC）部署于西班牙维戈大学，而接入网部分——包括用户设备（User Equipment, UE）、演进型基站（e-Node-B, eNB）与边缘云——则搭建于爱尔兰的IRIS试验床。尽管端到端蜂窝网络的初始部署可视为静态配置，但DYNAMO项目展示了端到端蜂窝网络在运行时所需的弹性扩展能力。为此，我们设计了一个对延迟敏感的端到端蜂窝网络（网络切片）用例：初始状态下，用户设备连接的端点位于核心网（维戈大学）；当用户设备的延迟阈值超出可接受范围时，该连接端点将迁移至边缘节点（IRIS试验床）。在此场景中，用户设备会将实测延迟上报至开放网络自动化平台（Open Network Automation Platform, ONAP）；若预先定义的服务水平协议（Service-Level Agreement, SLA）遭到违反，ONAP将触发相应的策略驱动控制动作。本数据集包含用户设备向ONAP上报的全部监测报告。最终，用户设备数据平面的连接端点会自动迁移至接入网（IRIS试验床），从而大幅降低用户设备的端到端延迟。针对网络接入部分，我们在IRIS试验床的虚拟机中部署了一套srsLTE演进型基站、一套srsLTE用户设备，以及一套Devstack（边缘云）；此外还搭建了由ONOS SDN控制器管控的软件定义网络（Software Defined Network, SDN）交换机。针对网络核心部分，我们在维戈大学的Devstack（核心云）上部署了源自开放无线接入接口（Open Air Interface, OAI）的解耦式虚拟演进分组核心。DYNAMO项目成功实现了两个不同域（维戈大学与IRIS试验床）间多类网络组件与技术的集成，并达成了全部初始目标：(i) 实现ONAP与IRIS试验床的通信，以在核心云与边缘云部署通用虚拟网络功能（Virtual Network Function, VNF）；(ii) 在IRIS试验床部署搭载用户设备与演进型基站的端到端蜂窝网络，并在维戈大学部署虚拟演进分组核心；(iii) 将用户设备的遥测数据上报至ONAP；(iv) 在ONAP中设计并测试闭环控制动作，当服务水平协议无法满足要求时，将用户设备的数据平面迁移至边缘节点。DYNAMO项目为大量后续5G实验铺平了道路，这些实验将需要资源编排能力，例如网络切片的部署，或是在边缘云有限资源中自动调度服务。本仓库包含用户设备向ONAP发送的监测信息，用于判定用户设备与分组数据网关（Packet Data Network Gateway, PGW）之间的延迟是否处于合理范围，或是需要采取相应动作以降低此类延迟。