SDN-enabled Application-aware Network Control Architectures and their Performance Assessment (DFG SDN-App)Phase 2: Design and Evaluation of Flexible Programmable Hybrid Real-time Networks with Hard and Soft Real-time Guarantees

软件定义网络(SDN)标志着一种根本性的范式转变，它允许在当今的通信网络中集成新的概念，如应用感知资源管理。这些概念尤其改善了面向用户的多媒体应用的体验质量(QOE)。这一概念的可行性在DFG SDN-App第一阶段得到了论证。第一阶段体系结构从两个方面研究了应用感知管理和控制的实现：考虑应用控制平面上的网络信息提高了网络资源的使用；考虑网络控制平面上的应用需求严格提高了应用性能和质量。因此，DFG SDN-App第二阶段扩展了所考虑的场景和体系结构：除了照顾面向最终用户的多媒体应用和QOE(第一阶段)之外，它新集成了时间关键型(工业)服务及其变化的需求-终端用户不起主导作用的情况。具体而言，阶段2侧重于时间关键型服务的硬实时和软实时需求，这些服务具有动态变化的需求，如从智能制造到控制设备。硬实时约束要求例如最大端到延迟和最大延迟抖动；软实时约束体现在随机保证中，例如分组丢失和延迟。现有方法存在未使用的、因此浪费的网络资源或处理动态变化的需求和网络重新配置的复杂性。第二阶段的目标是实现一个混合实时网络，在保证硬实时应用和软实时应用的同时，同时托管多媒体应用。特别是，时间敏感型网络(TSN)和网络可编程性(P4)等概念被联合调查和分析。为了评估整个体系结构，不仅采用了传统的衡量标准，还研究了网络灵活性等新的衡量标准，以量化混合实时网络在时间和成本约束下不断变化的需求下的收益。总体目标是选定用例(工业、数据中心、广域网)的指导方针，将阶段1和阶段2的结果结合在一起。阶段1向工业和实时网络的扩展对于支持当今应用程序的动态和更快增长的需求非常重要。灵活的体系结构在重新配置时提供保证，将允许实现可同时托管所有应用类型的混合网络。由于能够为所有类型的网络场景编程和使用标准硬件，网络运营商将在未来节省运营和资本支出。

Software Defined Networking (SDN) marks a fundamental paradigm shift that allows integrating new concepts such as application-aware resource management in today’s communication networks. Such concepts, in particular, improve the Quality-of-Experience (QoE) of user-oriented multimedia applications. The feasibility of such concept was demonstrated in Phase 1 of DFG SDN-App. The Phase 1 architecture investigated the realization of application-aware management and control on two fronts: taking network information on the application control plane into account improves the use of network resources; considering application demands on the network control plane rigorously improves application performance and QoE.Phase 2 of DFG SDN-App consequently extends the considered scenario and architecture: besides taking care of end user-oriented multimedia applications and QoE (Phase 1), it newly integrates time-critical (industrial) services and their changing requirements – cases where end users do not play a dominant role. In particular, Phase 2 focuses on hard as well as soft real-time requirements of time-critical services with dynamically changing demands like smart manufacturing to control devices. Hard real-time constraints require, e.g., a maximum end-to-delay and maximum delay jitter; soft real-time constraints manifest in stochastic guarantees, e.g., on packet loss and delay. Existing approaches suffer from unused, hence wasted network resources or the complexity to address dynamically changing demands and network reconfigurations. Real-time constraints are in trade-off to flexibility.Applying the methodologies from analysis, simulation, to measurement and testbed implementations, the goal of Phase 2 is to realize a hybrid real-time network that provides guarantees to hard and soft real-time applications, while at the same time hosting multimedia applications. In particular, concepts such as Time Sensitive Networking (TSN) and network programmability (P4) are investigated and analyzed jointly. To evaluate the overall architecture, not only traditional measures are applied, but also new measures such as network flexibility are investigated to quantify the benefits of hybrid real-time networks under changing demands with time and cost constraints. The overall target are guidelines for selected use cases (industrial, data center, wide area network), that combine the results of Phase 1 and 2.An extension of Phase 1 towards industrial and real-time networks is important to sustain the dynamically and faster growing demands of today’s applications. A flexible architecture that provides guarantees in face of reconfigurations will allow realizing hybrid networks that can host all application types simultaneously. With the ability to program and use standard hardware for all kind of network scenarios, network operators will save operation and capital expenditures in the future.