Service Function Chain Migration in LEO Satellite Networks

With the advancements of software defined network(SDN)and network function virtualization(NFV),service function chain(SFC)placement becomes a crucial enabler for flexible resource scheduling in low earth orbit(LEO)satellite networks.While due to the scarcity of bandwidth resources and dynamic topology of LEO satellites,the static SFC placement schemes may cause performance degradation,resource waste and even service failure.In this paper,we consider migration and establish an online migration model,especially considering the dynamic topology.Given the scarcity of bandwidth resources,the model aims to maximize the total number of accepted SFCs while incurring as little bandwidth cost of SFC transmission and migration as possible.Due to its NP-hardness,we propose a heuristic minimized dynamic SFC migration(MDSM)algorithm that only triggers the migration procedure when new SFCs are rejected.Simulation results demonstrate that MDSM achieves a performance close to the upper bound with lower complexity.

SFC Design and VNF Placement Based on Traffic Volume Scaling and VNF Dependency in 5G Networks

The development of Fifth-Generation(5G)mobile communication technology has remarkably promoted the spread of the Internet of Things(IoT)applications.As a promising paradigm for IoT,edge computing can process the amount of data generated by mobile intelligent devices in less time response.Network Function Virtualization(NFV)that decouples network functions from dedicated hardware is an important architecture to implement edge computing,deploying heterogeneous Virtual Network Functions(VNF)(such as computer vision,natural language processing,intelligent control,etc.)on the edge service nodes.With the NFV MANO(Management and Orchestration)framework,a Service Function Chain(SFC)that contains a set of ordered VNFs can be constructed and placed in the network to offer a customized network service.However,the procedure of NFV orchestration faces a technical challenge in minimizing the network cost of VNF placement due to the complexity of the changing effect of traffic volume and the dependency on theVNFrelationship.To this end,we jointly optimize SFC design and VNF placement to minimize resource cost while taking account of VNF dependency and traffic volume scaling.First,the problem is formulated as an Integer Linear Programming(ILP)model and proved NPhard by reduction from Hamiltonian Cycle problem.Then we proposed an efficient heuristic algorithm called Traffic Aware and Interdependent VNF Placement(TAIVP)to solve the problem.Compared with the benchmark algorithms,emulation results show that our algorithm can reduce network cost by 10.2%and increase service request acceptance rate by 7.6%on average.

基于深度强化学习的能源高效VNF放置和链接方法

网络功能虚拟化(NFV)技术的出现使得网络功能由虚拟网络功能(VNF)提供,从而提高网络的灵活性,可扩展性和成本效益.然而,NFV面临一个重要挑战是,如何有效地将VNF放置不同的网络位置并链接起来引导流量,同时最大限度减少能源消耗.此外,面对网络服务质量要求,提高服务接受率对于网络性能也是至关重要的.为了解决这些问题,本文研究了NFV中的VNF放置和链接(VNFPC),以最大化服务接受率同时权衡优化能源消耗.因此,在NFV中设计了一种基于Actor-Critic深度强化学习(DRL)的能源高效的VNFPC方法,称为ACDRLVNFPC.该方法应用了适应性共享方案,通过在多服务之间共享同类型VNF和多VNF共享同一个服务器来实现节能.实验结果表明,提出的算法有效权衡了能耗和服务接受率,并且,在执行时间方面也得到了优化.与基准算法相比,ACDRL-VNFPC在服务接受率,能耗和执行时间方面性能分别提高了2.39%,14.93%和16.16%.

基于探路者算法的VNF放置与调度联合优化

虚拟网络功能放置和虚拟网络功能调度是近年来研究的热点,二者紧密关联缺一不可。将两个环节统一考虑、全局寻优既是技术发展趋势,也是网络功能虚拟化落地面临的艰巨挑战之一。然而,现有研究多将二者分立考虑,仅着眼单一问题的研究缺乏整体观。论文提出了一种动态的放置调度联合优化模型,在满足相关约束下,最小化服务功能链的完成时间。提出了一种改进的探路者算法解决该问题,引入了群体划分的机制增强全局搜索能力避免陷入局部最优。进一步设计了个体交叉和变异机制,减小对个体解的破坏,增强了局部搜索能力以及种群的多样性。此外,精英保留策略加快了算法的收敛。实验数据表明,论文提出的算法相较于各种主流进化算法在服务功能链完成时间评价指标上均有显著优势。

Virtualizing Network and Service Functions: Impact on ICT Transformation and Standardization

Virtualization of network/service functions means time sharing network/service（and affiliated）resources in a hyper speed manner.The concept of time sharing was popularized in the 1970s with mainframe computing.The same concept has recently resurfaced under the guise of cloud computing and virtualized computing.Although cloud computing was originally used in IT for server virtualization,the ICT industry is taking a new look at virtualization.This paradigm shift is shaking up the computing,storage,networking,and ser vice industries.The hope is that virtualizing and automating configuration and service management/orchestration will save both capes and opex for network transformation.A complimentary trend is the separation（over an open interface）of control and transmission.This is commonly referred to as software defined networking（SDN）.This paper reviews trends in network/service functions,efforts to standardize these functions,and required management and orchestration.

可靠性感知的边缘计算VNF实例放置

为了解决日益增长的延迟敏感型应用程序和用户需求与计算资源受限的冲突,移动边缘计算(Mobile Edge Computing,MEC)已经成为一种很有前途的计算范式。服务提供商通过在边缘环境中部署虚拟化网络功能(Virtual Network Functions,VNF),为用户提供更加高效和可扩展性的服务供应链(Service Function Chain,SFC)来满足用户请求。若在提供服务过程中出现不可靠的服务或严重的服务失败,可能导致用户的巨大损失,所以网络服务提供商必须保证提供持续可靠的服务。针对该问题,考虑了边缘服务器的可靠性,利用计算统一设备架构(Computational Unified Device Architecture,CUDA)支持的门控循环单元(Gate Recurrent Unite,GRU)来预测VNF实例是否可用,通过预测结果,提前对VNF进行备份,避免了过度冗余备份造成的成本过高问题。考虑服务器的存储资源有限,提出了基于VNF实例可用性的放置(RVP)算法,优化服务提供商的成本。最后对提出的算法进行了性能评估,实验结果验证了RVP算法的优越性。

基于VNF实例共享的服务功能链部署算法

针对服务功能链(SFC)部署过程中存在虚拟网络功能(VNF)实例部署成本和转发路径成本难以权衡的问题,提出了基于VNF实例共享的SFC部署算法。首先针对多链SFC建立VNF和虚拟链路映射模型,并预估路径部署长度上限,保证SFC时延需求;其次,在路径部署长度限制范围内,尽可能使VNF实例共享最大化,以平衡链路转发成本和VNF部署成本,最终得到SFC部署策略。与已有的SPH(shortest path heuristic)和GUS(greedy on used server)部署算法相比,所提算法所得的总运营成本分别降低6.6%和12.15%,且当SFC数量增多时,该算法的服务接受率可达89.33%。仿真实验结果表明,提出算法可以在保证用户服务质量的同时有效降低SFC部署成本。

基于特征权重感知的VNF资源需求预测方法

虚拟网络功能(Virtual Network Function,VNF)以服务功能链(Service Function Chain,SFC)的形式提供服务,能够满足不同服务的性能需求。由于网络具有动态性,为VNF实例分配固定资源会导致VNF实例的资源过多或者不足的问题。以往的研究对于VNF配置文件相关网络负载特征的重要性未做区分,因此,提出了一种基于特征权重感知的动态VNF资源需求预测方法。首先,使用ECANet学习VNF特征的权重值,以此来减少无用特征对模型预测结果的消极影响。其次,由于VNF配置文件数据集具有结构化特性,构建VNF资源预测模型时需要考虑以加强特征交互的方式来挖掘特征间深层的相互关系,提出使用深度特征交互网络(Deep Feature-Interactive Network,DIN)增强网络负载特征与VNF性能特征之间的交互能力,提高模型预测精度。最后,在基准数据集上将所提方法与同类方法进行对比实验,发现其在预测的有效性与精确性上更具优势。

Optimize the Deployment and Integration for Multicast-Oriented Virtual Network Function Tree

Due to the development of network technology,the number of users is increasing rapidly,and the demand for emerging multicast services is becoming more and more abundant,traffic data is increasing day by day,network nodes are becoming denser,network topology is becoming more complex,and operators’equipment operation and maintenance costs are increasing.Network functions virtualization multicast issues include building a traffic forwarding topology,deploying the required functions,and directing traffic.Combining the two is still a problem to be studied in depth at present,and this paper proposes a two-stage solution where the decisions of these two stages are interdependent.Specifically,this paper decouples multicast traffic forwarding and function delivery.The minimum spanning tree of traffic forwarding is constructed by Steiner tree,and the traffic forwarding is realized by Viterbi-algorithm.Use a general topology network to examine network cost and service performance.Simulation results show that this method can reduce overhead and delay and optimize user experience.

Network Function Virtualization Technology:Progress and Standardization

Network innovation and business transformation are both necessary for telecom operators to adapt to new situations, but operators face challenges in terms of network bearer complexity, business centralization, and IT/CT integration. Network function virtualization （NFV） may inspire new development ideas, but many doubts still exist within industry, especially about how to introduce NFV into an operator＇ s network. This article describes the latest progress in NFV standardization, NFV requirements and hot technology issues, and typical NFV applications in an operator networks.

基于VNF的网络服务冲突检测与避免机制

针对如何构建服务功能链(Service Function Chain,SFC)并进行资源分配为用户提供满意服务的问题,本文提出了一种包括候选路径构建、依赖冲突检测与避免以及虚拟网络功能(Virtual Network Function,VNF)部署的机制.首先,为了给SFC部署提供充足的资源,提出二级筛选及最优化选取的候选路径构建规则,为服务提供预选路径.其次,在SFC构建过程中,检测复用性与依赖关系之间的冲突,将依赖关系划分二元组后进行冲突判断,若产生冲突则进行等价类划分,给出冲突集合.然后,提出基于冲突集合以及LFGL(Least-First-Greatest-Last)原则的VNF部署规则,以最大化链路剩余带宽,保证端到端延迟.最后,在进行服务递交时,检测VNF流入流出比对数据量的影响,若产生冲突则进行冲突避免,若无法成功避免则执行规避策略.最后基于小型和大型两种网络拓扑对仿真系统进行性能评价.实验结果表明,本文设计的机制在复用率、时延、部署成功率方面所表现出的性能均优于对比算法.

AdaptNF: Adaptive service chain scheduling with stateless migration and NF consolidation

The combination of network function virtualization and software-defined networking allows various network functions to process flows according to their characteristics and requirements.Due to the highly dynamic nature of the workload,the network infrastructure needs to properly schedule the underlying resources in order to respond to workload changes in a timely manner.However,the existing NFV platform lacks a comprehensive solution for how to scale under workload variation,which may seriously hurt the overall system performance.To improve the scalability of the NFV platform and ensure consistent high performance under dynamic workloads,we propose AdaptNF,a novel NFV platform designed to support a combination of course-grained and fine-grained resource scheduling strategies.To deal with resource imbalance,which is the essential scheduling problem that leads to insufficient NFV performance,AdaptNF adopts a novel algorithm that can efficiently balance the workload among multiple network function instances through stateless flow migration.Our controlled experiments show that the AdaptNF scheme can optimize resource allocation and ensure outstanding performance after scaling.In terms of network throughput and latency,AdaptNF significantly improves the performance of the underlying NFV platform.

基于资源感知的多域服务功能链编排成本优化

网络功能虚拟化技术的兴起使得实例化为服务功能链(SFC)的网络服务能够共享基底网络,缓解了传统网络体系结构僵化的问题.然而,网络中大量服务请求给多域SFC编排带来了新的挑战.首先由于域内网络资源信息及内部策略的保密性,使得多域SFC的编排更为复杂.其次多域SFC编排要确定最佳候选编排域集,先前的研究较少考虑域间负载的均衡性,对服务接受率造成了消极影响.此外跨网络域编排服务请求对服务的成本和响应时间提出了更严格的要求.为解决上述挑战,在本文中,我们首先针对多域网络隐私性需求,提出了域级图的构造方法;然后基于域间负载均衡提出了域权重的计算方法进行SFC编排域的选择;最后,针对多域网络成本和响应时间需求,提出编排算法.实验结果表明,提出的算法有效地权衡了平均服务成本和接受率,并且在服务平均响应时间方面也得到了优化.

一种基于近端策略优化的服务功能链部署算法

针对网络功能虚拟化(NFV)环境下高维度服务功能链(SFC)部署的高可靠低成本问题,该文提出了一种基于近端策略优化的服务功能链部署算法(PPO-ISRC)。首先综合考虑底层物理服务器特征和服务功能链特征,将服务功能链部署建模为马尔可夫决策过程,然后,以最大化服务率和最小化资源消耗为优化目标设置奖励函数,最后,采用近端策略优化方法对服务功能链部署策略求解。仿真实验结果表明,与启发式算法(FFD)和深度确定性策略梯度算法(DDPG)相比,所提算法具有收敛速度快,稳定性高的特点。在满足服务质量的要求下,降低了部署成本,并提高了网络服务可靠性。

航空信息网络服务功能链请求调度策略

为解决航空信息网络场景下业务突发、拓扑结构动态变化以及资源有限所带来的资源利用效率低下问题,针对服务功能链请求(Service Function Chain Requirement,SFCR)高效调度的难题,提出一种网络功能虚拟化的服务功能链调度算法。算法依据SFCR与平台的相关性进行映射,提高服务器处理效率并有效减少虚拟网络功能实例化数量及整合平台资源;综合考虑时延和流量影响对网络功能实例进行整合迁移,降低网络能耗同时提升网络资源利用率。仿真结果表明,在满足时延需求条件下,新方法在运行平台数量、SFCR接受率、网络资源消耗等方面具有较好的优化性能,适用于解决航空信息网络场景下的SFCR调度问题。

一种基于主-备份节点同步映射的VNF热备份机制

网络功能虚拟化实现了网络功能与专属硬件平台的分离,虚拟网络功能(Virtual Network Function,VNF)通过动态组合形成服务功能链,实现同一硬件平台部署不同的功能服务。针对软件定义网络功能服务可靠性和物理网络资源效率问题,提出了一种基于主备份节点同步映射的VNF节点热备份机制。首先,选择满足节点资源约束的物理节点为主备份节点候选映射对;然后,选择满足带宽约束的最短物理链路为服务路径,实现节点与链路资源总开销的优化。实验结果表明,与传统的备份机制相比,该方法能够满足电信级高可靠性的服务需求,降低了VNF节点备份资源总开销,提高了服务请求的接受率和底层物理网络资源的利用率。

特征关联的服务功能链可靠性优化算法

针对优化服务功能链(Service Function Chain,SFC)可靠性时资源利用率低的问题,提出了特征关联的服务功能链可靠性优化算法。首先,从底层节点的资源可用率、周边资源密集度、全网中心性评价其作为虚拟网络功能(Virtual Network Function,VNF)实例备选节点的优势,从VNF实例的节点度数与带宽流量评价其作为备选对象的优势。其次,逐个对优越性最高的VNF实例选择备份底层节点,直到所有SFC的可靠性满足客户要求。在为VNF实例选择备份底层节点时,提出了基于资源特征关联感知的备份节点选择算法,将当前VNF实例所映射的底层节点与其他待备份VNF实例的底层节点的邻接节点进行关联,为当前VNF实例选择最优备份底层节点的同时,减少为其他待备份VNF实例选择备份底层节点时链路的开销。仿真结果表明,相比于已有算法,该SFC可靠性优化算法下备份实例数降低了约29%,备份带宽资源消耗降低了约42%,SFC接受率提升了约12%。

移动边缘网络中基于用户移动的服务功能链迁移策略

移动边缘计算(Mobile Edge Computing,MEC)通过在网络边缘部署服务器,提供计算和存储资源,可为用户提供超低时延和高带宽业务。网络功能虚拟化(Network Function Virtualization,NFV)与MEC技术相结合,可在MEC服务器上提供服务功能链(Service Function Chain,SFC),提升用户的业务体验。为了保证移动用户的服务质量,需要在用户跨基站移动时将SFC迁移到合适的边缘服务器上。主要以最小化用户服务的端到端时延和运行成本为目标,提出了MEC网络中具有资源容量约束的SFC迁移策略,以实现移动用户业务的无缝迁移。仿真结果表明,与现有方案相比,该策略具有更好的有效性和高效性。

基于5G端到端网络切片技术的配网差动保护方法

在当前配网差动保护过程中,获取差动保护判据时仅能对单一参数进行处理,导致差动保护效果受故障位置影响较大,对此提出基于5G端到端网络切片技术的配网差动保护方法。根据“充分式”思想进行配网充分式故障识别与分析,选用5G端到端网络切片技术进行配网端到端网络切片处理,获取配网通信网络部分判据,采集复合式判据,设定差动保护启动单元,实现配网差动保护。仿真实验结果证实,无论故障点在保护区内或是保护区外,其使用效果均不会受到故障点位置的影响,保护装置可及时启动保护单元,维护配网运行的稳定性。

基于特征选择的VNF资源需求预测方法

针对在网络切片场景下以往的VNF(虚拟网络功能)资源分配策略无法满足动态的资源需求,很容易导致资源分配不足或过度分配的问题,提出了一种基于两阶段算法(two-stage algorithm,TSA)的VNF资源需求预测方法。该方法首先基于数据特征筛选出与预测目标高度相关的候选特征集,然后利用贪婪式前向搜索策略对候选特征集进一步筛选获得最优特征集,最终训练出不同类型的预测模型。仿真结果表明,基于该方法所训练的模型可以获得更好的预测性能,同时该方法的可扩展性较好,训练好的模型可以直接集成到现有的VNF部署算法中应用。