Distributed AI training often suffers from network faults. Network faults, especially at the last hop between a switch and a host, result in loss of connectivity, resulting in training job stalls and eventual failure. This is typically managed through a fail-stop mechanism, followed by a restart, incurring significant inefficiencies. We present ReCCL, the first network fault-tolerant collective communication library (CCL) that allows training progress to be preserved by seamlessly failing over to alternate paths when a network fault occurs. During failover, ReCCL keeps communication states synchronized while using dynamic channel load balancing and intra-host GPU routing to improve communication performance. Our evaluations demonstrate that ReCCL can perform failover seamlessly with minimal performance losses. Additionally, our simulations also demonstrate that failover can be effectively used to achieve significant savings in GPU hours for large-scale distributed AI training workloads.

Virtual Radio Access Network (vRAN) is an emerging approach offered by cloud providers to accelerate 5G services deployment. Despite significant microsecond-scale traffic variations, vRAN instances are provisioned based on peak load to meet strict latency requirements, leading to significant resource waste. Conceivably, vRAN can share CPUs with other applications to increase CPU utilization. Yet, existing sharing solutions require modifications to vRAN source code, hindering their deployment on public clouds. We present Aquifer, a microsecond-scale scheduler providing transparent CPU sharing for vRAN workloads. Our key observation is a common producer-consumer task execution pattern in mainstream vRAN implementations. We exploit this pattern to reclaim CPU cores from worker threads only at the boundary of processing different tasks. This guarantees run-to-completion task processing, which is critical for vRAN to achieve low latency and stability. Aquifer intercepts system calls invoked by vRAN at the OS layer to achieve transparent load monitoring and core reallocation. Aquifer employs a set of system-level optimizations on thread state detection, signal transmission and core selection, which reduces the scheduling cycle to 2$\mu s$. Experimental results show that Aquifer reclaims up to 88.31% of wasted CPU resources for two mainstream vRAN implementations, FlexRAN and OAI, without any source code modifications.

Network function virtualization (NFV) is a promising technology that decouples network functions from hardware. Connecting virtual network functions (VNFs) in series to form a service function chain (SFC) can flexibly orchestrate and expand network functions. However, there are higher availability requirements for SFCs. This paper aims to solve the SFC placement problem under availability and resource constraints. This paper proposes the sideway cross (SC) backup model, which considers the availability of both VNFs and physical machines (PMs) in a data center. The SC model cross-arranges the backup instances of VNFs to guarantee availability and optimize resource consumption. Then, this paper proposes the heuristic meteor shower optimization (MSO) algorithm to place SFCs. Compared to traditional heuristic algorithms, MSO can improve the execution time by approximately 200%. Combined with the SC backup model, MSO can effectively improve the availability and resource overhead. The evaluation results show that the proposed approach can guarantee higher availability and consumes fewer resources. The proposed approach only needs 75% of the resources to achieve the same availability as the state-of-the-art models.

Network Function Virtualization (NFV) is a promising paradigm that separates network functions from proprietary devices. Network service in NFV-enabled networks is achieved as a Service Function Chain (SFC), consisting of a series of ordered Virtual Network Functions (VNFs). However, migration of VNFs for more flexible services within dynamic networks is a key challenge. Current VNF migration studies mainly focus on single VNF migration decisions without considering the sharing and concurrent migration of VNF Instance (VNFI). In this paper, we assume that each deployed VNFI is used by multiple SFCs and deal with the optimal location allocation for the concurrent migration of VNFIs based on the actual network situation. We first formalize the VNF migration and SFC reconfiguration problem as a mathematical model, which aims to minimize the end-to-end delay for all affected services and to guarantee network load balancing after the migration simultaneously. To this end, we prove the NP-hardness of this problem and propose the Improved Hybrid Genetic Evolution (IHGE) algorithm to address it. Besides, to reduce the computation overhead of IHGE for large-scale networks, a multi-stage heuristic algorithm based on optimal order (MSH-OR) is designed. Finally, we perform a side-by-side comparison with prior algorithms. Extensive evaluation shows that the proposed approaches can effectively reduce the average delay for different scale networks while ensuring network load balancing.

Network Function Virtualization (NFV) brings flexible provisioning and great convenience for enterprises outsource their network functions to the Data Center Networks (DCNs). Network service in NFV is deployed as a Service Function Chain (SFC), which includes an ordered set of Virtual Network Functions (VNFs). However, in one SFC, the SFC delay increases linearly as the length of SFC increases. SFC parallelism can achieve high performance of SFC. In this article, we focus on the parallelized SFC placement problem in DCN considering availability guarantee and resource optimization. Firstly, we define the parallelized SFC and propose a multi-flow backup model. The parallelized SFC consists of multiple parallelized sub-SFCs. We split large data flow into multiple small sub-flows, each of them can be transmitted in one sub-SFC. The backup model provides backup sub-SFCs for working sub-SFCs to improve availability. Finally, we design three placement strategies and a Hybrid Placement Algorithm (HPA) aimed at mapping SFCs to DCN. Evaluation results show that our proposed solutions outperform the related work. We can reduce SFC delay (30%) and optimize link consumption (reduce 40%) while guaranteeing availability (99.999%).

Since the advent of network function virtualization (NFV), cloud service providers (CSPs) can implement traditional dedicated network devices as software and flexibly instantiate network functions (NFs) on common off-the-shelf servers. NFV technology enables CSPs to deploy their NFs to a cloud data center in the form of virtual network functions (VNFs) without costly capital expenditures and operating expenses. However, it is an essential but intractable issue for CSPs to devise a suitable VNF placement scheme to optimize network resource consumption and improve network performance. In this article, we focus on the VNF placement problem for mapping users’ service function chain requests (SFCRs) in cloud networks. To enhance network resource utilization, we consider the fundamental resource overheads and implementation method of VNFs. The VNF placement problem is formulated as an integer linear programming model with the aim of minimizing the total network resource consumption while guaranteeing the delay requirements of SFCRs. We devise a two-phase optimization solution (TPOS) to solve the problem. TPOS contains a mapping phase to map SFCRs on servers and an adjustment phase to optimize the placement of VNFs and VNF requests. Evaluation results demonstrate that TPOS can derive near-optimal server resource consumption and significantly enhance network resource utilization. TPOS can guarantee the delay requirements of SFCRs and outperform contrastive schemes in terms of activated servers, SFCR acceptance ratio, and average VNF utilization.

Nowadays, network softwareization is an emerging techno-economic transformation trend that significantly impacts how enterprises deploy their network services. As an essential technology in this trend, Network Function Virtualization (NFV) enables scalable and inexpensive network services by flexibly instantiating Virtualized Network Functions (VNFs) on commercial-off-the-shelf devices. In this paper, we focus on the VNF placement problem in NFV-enabled networks, aiming to maximize the number of accepted Service Function Chain Requests (SFCRs) while guaranteeing their delay requirements. To improve resource utilization efficiency, we take account of Fundamental Resource Overheads (FROs) and the shareability of VNF instances. We mathematically formulate the VNF placement problem and propose the Throughput Optimization and Delay Guarantee (TO-DG) heuristic solution, consisting of an affinity-based SFCR mapping algorithm and a VNF request adjustment algorithm. The evaluation results show that the performance of TO-DG is near to results derived by ILP solver for small scale problems. Moreover, TO-DG obtains higher network throughput than contrasting schemes in different scenarios and significantly improves network resource utilization.

The next-generation network system is envisioned to be a multi-service network supporting different applications with multiple requirements. In this vision, Network SLicing (NSL) is considered a key mechanism to create multiple virtual networks over the same physical infrastructure. However, it is a challenging problem to deploy NSL with great flexibility and full automation. In this paper, we propose a novel lightweight SOA-based NSL creation system, which includes the design domain, execution domain, and infrastructure domain. The design domain provides an easily-operating service design environment. Besides, we define a detailed description of NSL named Network SLicing Descriptor (NSLD), illustrating the resource and requirements for designing network services. The execution domain is a service execution environment with multi-tenancy support. This domain based on micro service architecture is distributed and self-organized. The infrastructure domain contains multiple network infrastructure resource domains that enable micro services. With these components, tenants can create NSL freely via an intuitive and easy-to-use UI on a web browser. We also implement an IoT NSL scenario to solve the multi-flow transmission problem. After adding sub-services and QoS policy into the design domain, the scenario can run automatically at the execution domain.

Software-Defined Networking (SDN), providing flexible controlling and monitoring mechanisms that simplifies network management, is becoming prevalent in recent years. However, replacing all legacy network devices with SDN-capable devices is cost-prohibitive. One practical approach for the SDN deployment is to incrementally upgrade a few legacy devices to SDN devices. The network, which consists of legacy and SDN devices, is called a hybrid SDN. Existing hybrid SDN deployment schemes do not consider the security impact of device deployment. They use the same type of devices to upgrade, and upgraded devices could be compromised if an attacker controls one SDN device by leveraging its vulnerabilities.In this paper, we consider this security issue in the hybrid SDN deployment and present the Secure and Cost-effective Switch Upgrade (SCESU) problem. The SCESU problem aims to upgrade a few network devices to satisfy the security requirement by using multiple SDN switch types with a minimal upgrade cost. The complexity of the SCESU problem comes from common vulnerabilities shared among different types of SDN devices and attack propagations among network nodes. To efficiently solve the problem, we propose the BAGUETTE algorithm to judiciously choose and upgrade critical legacy switches with selected SDN devices. Simulation results show that BAGUETTE achieves up to about 92.1% security enhancement compared with legacy network and reduces to 11.1% cost of the securest deployment.

Network Function Virtualization (NFV) brings flexible provisioning and great convenience for enterprises outsource their network functions to the Data Center Networks (DCNs). Network service in NFV is deployed as a service chain, also known as Service Function Chain (SFC), which includes an ordered set of Virtual Network Functions (VNFs). However, there might be some small size flows being queued behind the large flows in one SFC, resulting in network congestion and high SFC delay. In this paper, we focus on the parallelized SFC placement problem in DCN considering availability guarantee and resource optimization. Firstly, we define the parallelized SFC and propose a multi-flow backup model. Then, we design three placement strategies and a Hybrid Placement Algorithm (HPA) aiming at mapping SFCs to DCN. Compared with the existing approaches, our proposed solutions can reduce SFC delay and optimize link consumption while guaranteeing availability.

Mobile Edge Computing (MEC) is an emerging architecture that supports ultra-low latency and high-bandwidth services by deploying servers at the edge of the network to provide computing and storage resources. Recent studies tend to combine (Network Function Virtualization) NFV with MEC and deploy (Virtualized Network Functions) VNFs on MEC servers to achieve fast access to the edge user equipment (UE). However, to guarantee the QoS requirements of mobile users, it is necessary to migrate VNFs to an advisable edge server when users move across Base Stations (BS). How to choose the target BS for VNFs migration? How to select the path for VNF data migration? How to ensure the QoS of user services during the migration process? To solve these issues, we study the seamless VNFs migration problem in mobile edge networks and formulate it as an ILP model, which aims to minimize the migration delay and cost. Then we propose a migration algorithm based on Dijkstra (MBD) to obtain the migration destination BS and migration paths. We implement the mathematical model in Gurobi and design a Greedy algorithm to compare the performance with the MBD algorithm. The experiment results show the effectiveness and efficiency of our algorithm.

Network function virtualization (NFV) plays a vital role in 5G mobile networks. Concatenating virtual network functions (VNFs) into service function chains (SFCs) provides flexible and diverse network support for intelligent applications. However, the mobile network connection is very unreliable. A reasonable SFC composition mechanism is essential for stable service providing. This paper proposes a high availability service function chain composition approach, TSFCC. TSFCC includes real-time road marking strategy, bi-composition mechanism, and VNF reallocation mechanism. Evaluation results prove that TSFCC can adapt to the mobile network environment and provide users with efficient and highly available SFC service.

Network function virtualization (NFV) and mobile edge computing (MEC) enable internet service providers (ISPs) to deploy service function chains (SFCs) to achieve the convenience and performance benefit without incurring high service delay, capital expenditures, and operating expenses. In MEC-NFV networks, network services are deployed in the form of service function chains (SFCs), each consisting of an ordered set of virtual network functions (VNFs). In this paper, we focus on the VNF placement problem in MEC-NFV enabled networks, aiming to optimize the throughput of SFC requests (SFCRs). First, we involve the sharing mechanism of VNF instances in the problem formulations, which can improve network resource utilization and save more node resources. Then we formulate the problem mathematically and propose a correlation-based mapping algorithm to map SFCRs in the network. Moreover, we design an adjustment algorithm to optimize the mapped SFCRs. Evaluation results show that our proposed solution efficiently improves the throughput of SFCRs compared with the benchmarks.

Mobile Edge Computing (MEC) is an emerging network architecture that provides computing capabilities at the edge of the mobile network. Recent approaches tend to deploy MEC applications in the Network Function Virtualization (NFV) network. In the MEC-NFV environment, mobile network services are deployed as service chains, also known as Service Function Chains (SFCs). In this paper, we focus on the SFC placement problem in the MEC-NFV environment while guaranteeing availability. We design a backup model to improve SFC availability. Besides, we propose a Dynamic Programming (DP)-based algorithm to place SFC. Evaluation results show that our proposed solutions outperform the existing approaches in terms of availability guarantee and resource optimization.

Network Function Virtualization (NFV) is a promising technology that decouples network functions from the physical device on which they deployed. Network service in NFV is deployed as Service Function Chain (SFC) that consists of an ordered set of Virtual Network Functions (VNFs). In this paper, we focus on the VNF placement problem in data center networks considering availability guarantee and resource optimization. Firstly, we define an availability model that takes both physical device failures and VNF failures into consideration when evaluating the availability of SFC. Secondly, we propose a novel Joint Path-VNF (JPV) backup model that combines path backup and VNF backup in a joint way. In the JPV backup model, resource consumption can be effectively reduced. Finally, we design an Affinity-Based Algorithm (ABA) to reduce physical link consumption when map VNFs. The evaluation results show that ABA and JPV can achieve better availability improvement (99.99%) with less resource (reduce 40% PL consumption).

Mobile Edge Computing (MEC) is a promising network architecture that pushes network control and mobile computing to the network edge. Recent studies propose to deploy MEC applications in the Network Function Virtualization (NFV) environment. The mobile network service in NFV is deployed as a Service Function Chaining (SFC). In the dynamic and resource-limited mobile network, SFC placement aiming at optimizing resource utilization is a challenging problem. In this article, we solve the SFC placement problem in the MEC-NFV environment. We formulate the SFC placement problem as a weighted graph matching problem, including two sub-problems: a graph matching problem and an SFC mapping problem. To efficiently solve the graph matching problem, we propose a Linear Programming–(LP) based approach to calculate the similarity between VNFs and physical nodes. Based on the similarity, we design a Hungarian-based algorithm to solve the SFC mapping problem. Evaluation results show that our proposed LP-based solutions outperform the heuristic algorithms in terms of execution time and resource utilization.

With the increasing scale of cloud datacenters and rapid development of virtualization technologies, many cloud-based services have been deployed to meet requirements. Virtual machines (VMs) are placed on physical servers, and often provide virtual environment for cloud services. Therefore, virtual machines placement (VMP) problem has gradually attracted many attentions. It is meaningful that how to effectively and efficiently place VMs on servers to guarantee the service reliability and reduce the bandwidth consumption. In this paper, we first formulate VMP with a reliability model and a bandwidth consumption model, and analyse its complexity. Then we propose a VMP optimization approach to solve the problem and prove its effectiveness and efficiency. The core algorithm of our approach is an approximation algorithm to get VM partitions under the constraint of a specified reliability parameter. Then placement problem is transformed into matching problem between VM partitions with physical servers. Finally, the evaluation results show the effectiveness of the proposed approach and performance advancement over the existing approaches.

Mobile Edge Computing (MEC) is a promising network architecture that pushes network control and mobile computing to the network edge. Recent studies propose to deploy MEC applications in the Network Function Virtualization (NFV) environment. The mobile network service in NFV is deployed as a Service Function Chains (SFC). In this paper, we solve the SFC placement problem in a MEC-NFV environment. We formulate the SFC placement problem as a weighted graph matching problem, including two sub-problems: a graph matching problem and a SFC mapping problem. To efficiently solve the graph matching problem, we propose a linear programming-based approach to calculate the similarity between physical nodes and VNFs. Based on the similarity, we design a Hungarian-based placement algorithm to solve the SFC mapping problem. Evaluation results show that our proposed solutions outperform the greedy algorithm in terms of execution time and resource utilization.

Network Function Virtualization (NFV) brings great flexibility and scalability to the deployment of network services by decoupling network functions from dedicated devices, which has attracted more attention from both academia and industry. Network services in NFV are deployed in the form of Service Function Chain (SFC), which consists of multiple ordered Virtual Network Functions (VNFs). However, how to effectively place VNFs remains a problem to be solved. In this paper, we investigate joint correlation-aware VNF selection and placement problem. We first formulate the problem as an Integer Linear Programming (ILP) problem and propose a method based on self-learning matrix to partition VNF correlation. Then, we design a Joint Correlation-aware VNF Placement (JCVP) algorithm based on Dynamic Programming to transform the problem into several VNF mapping subproblems. Extensive simulation results show that compared with the previous algorithms our approach has better performance in link occupancy, SFC acceptance, and VNF utilization rate.

Although network function virtualization (NFV) is a promising approach for providing elastic network functions, it faces several challenges. A critical but difficult issue for the service and network providers is deciding where to instantiate a list of virtual network functions (VNFs), namely VNF placement problem. In this paper, we investigate the VNF placement, for the purpose of resource and network traffic consumption minimization. Moreover, we consider the arrival rates of users' requests for different types of service function chains (SFCs). This allows the placement scheme to adapt to users' time-varying requests and improve the network resource utilization. Then we formulate the VNF placement problem as a jointly constrained optimization problem. Afterwards, we propose an approach called joint optimization resource and traffic consumption (JORTC) with enhanced biogeography-based (EBBO) optimization algorithm to resolve the VNF placement problem. Finally, the evaluation results show the effectiveness of J-ORTC approach and performance advancement over the benchmarks.

Network Function Virtualization (NFV) is an emerging technology decouples network functions from hardware. Network service in NFV is deployed as a service chain, also known as Service Function Chain (SFC). SFC consists of an ordered set of Virtual Network Functions (VNFs). However, VNFs bring new challenges in providing network services with availability guarantee. In addition, in a customizable and dynamic NFV-enabled network, the composition and mapping of service chain are different from that of a traditional network. In this paper, we define an availability model that takes both hardware and VNF failures into consideration. Then we propose Joint Path-VNF backup model to combine path and VNF backup in a joint way. And a priority-based algorithm is designed for service chain composition and mapping. Simulation results show that our proposed solutions can reduce resource consumption while guaranteeing availability.

Mobile edge networks benefit cloud applications in particularly providing a shorter response latency for mobile terminals. However, the cumulative increase of mobile terminals and emerging cloud applications, poses new challenges for edge networks. To combat this issue, the promising idea of mobile micro-clouds (MMCs) is proposed to enhance edges and the cloud. In this paper, we investigate the problem of virtual cluster (VC) placement in MMCs, to minimize the average response latency with various requests among multiple cloud applications. Then a hybrid swarm intelligence approach is proposed to optimize VC placement scheme, for a trade-off between the average response latency and the overall VC placement cost. The preliminary evaluation results show the effectiveness and efficiency of our approach.

Mobile Edge Computing (MEC) is a promising architecture where network services are deployed to the network edge. Recent studies tend to deploy Network Function Virtualization (NFV) services to MEC. Network services in NFV are deployed as Service Function Chains (SFCs). In this paper, we mainly focus on the SFC placement problem in a MEC-NFV environment, which is different from the data center network. Firstly, we formulate this problem as a weighted graph matching problem consisting of graph matching and SFC mapping. Then, we propose a linear programming-based approach to match the edge network and SFC. Finally, we design a Hungarian-based placement algorithm to map SFC in the edge network. A heuristic-based greedy algorithm is also designed to compare the performance. Evaluation results show that our proposed solutions outperform the greedy algorithm in terms of execution time.

With the explosive growth of large scale services, traditional mobile networks have become increasingly unable to guarantee the efficient operation of services. However in the fifth generation (5G), supported by Software Defined Network (SDN) and Network Function Virtualization (NFV), network slicing technology [1] makes mobile networks more intelligent and flexible. 5G network slicing allows a set of logically independent virtual networks to be created on a common physical infrastructure and provides appropriate monitoring, management and resource allocation for a variety of different types of communication services [2].

Network Function Virtualization (NFV) is a new network paradigm that decouples network functions from dedicated hardware. Network services in NFV are deployed as service chains, also known as Service Function Chains (SFCs). SFC consists of an ordered set of Virtual Network Functions (VNFs). One of the main challenge when deploying SFC is to efficiently make use of the resource. In this paper, we focus on the SFC composition and mapping considering resource optimization. We formulate the SFC composition and mapping problem as a weighted graph matching problem. Then we propose a Hungarian based algorithm to solve the SFC composition and mapping problem in a coordinated way.

With the emergency of IoT, there are many IoT network slices with different network requirements. Most of the current IoT system are specific and non-programmable and therefore their slices are difficult to reuse. It is difficult to meet different QoS requirements especially in IoT system because there are plenty of IoT sensors in IoT system. In this paper, we propose a novel IoT network slicing creation system which based on two emerging SDN and NFV technologies. It provides an easily-operating service creation environment and a service execution environment based on micro service architecture. We implement an IoT muti-flow transmission scenario. After adding subservices and QoS policies into a business process at the design plane, the IoT scenario can run automatically at the execution plane. Experiment results on the scenario show that the numbers of packets per second of different flows are changing gradually depend on QoS policies.