In modern network structures that become more complex and emphasize flexibility, the demand for the automation of network management and Intent Driven Network (IDN) continues to increase. In response, technology utilizing virtualization and control plane separation has developed, and research on network automation based on this is being actively conducted. However, limited studies focus on building an automated network in environments comprised of traditional switches that lack support for these advanced functionalities. Consequently, this study presents a technology proposal that creates the CLI command for existing commercial switches by incorporating user requests conveyed through natural language. For this purpose, we applied Large Language Model (LLM) for generating and Network Digital Twin for verification environment.

Networks are becoming more complex, and there’s a growing focus on making them more flexible. To meet this need, Network Functions Virtualization (NFV) has been introduced. Accordingly, efficient scheduling for Virtual Network Functions (VNFs) and Service Function Chains (SFCs) is highlighted. Many research about SFC scheduling focus on deploying new SFC. However, during off-peak times, like early morning hours, it’s essential to reorganize existing SFCs to cut down on power usage. This study propose reorganization method that not only reduce power consumption, but also maintain or improve service performance (especially, we try to reduce service latency). To achieve this, we integrated VM consolidation, a strategy from Cloud Data Centers, with SFC scheduling. We also utilized a Deep Reinforcement Learning technique called Rainbow, and Self-Attention Mechanism. Consequently, our approach resulted in improved performance compared to other rule-based methods.

Network orchestration is pivotal in automating device, equipment, and service management within the network system. Currently, Network Function Virtualization (NFV) offers immense potential, but the challenge still lies in designing a capable orchestrator for dynamic networks. To address this challenge, we propose an AI-based NFV Orchestration Framework that leverages self-learning capabilities to detect dynamic network changes and make optimal decisions. This framework covers a range of essential functionalities, including NFV Orchestration, VNF Deployment, Service Function Chaining (SFC), Auto-Scaling, Migration, Anomaly Detection, Power Management, and Attack & Intrusion Detection. These functions collectively form a comprehensive ML-driven orchestration framework that offers adaptability, intelligence, and efficiency across the entire NFV environment. Our proposed structure aims for zero-touch automation, contributing to the efficient management of dynamic NFV network environments, and making it a compelling solution for the future of networking.

In this study, we focused on anomaly detection of network switches using log analysis. In previous research, we designed a log parser to find common patterns within logs, which replaced logs with ’pattern’ and classified similar patterns into the same ’event’. In this study, we utilize the log parser and calculate abnormality scores for each log pattern. In log analysis, analyzing the occurrence patterns is more important than the meaning of individual logs. With this perspective, we design the following three items to analyze the tendency of occurrence of each log pattern to better suit network switch log analysis. 1) Log-TF-IDF: the computed rarity of each log pattern, 2) Log-Prob: the probability of occurrence of each event predicted by machine learning, 3) Log-Freq: the numerical value for the degree of repetition of each log event. We calculate the ’Abnormal Score’ by considering all three metrics. By calculating Abnormal Scores from both normal and abnormal log data, we demonstrate the effectiveness of our approach. We implemented a simple anomaly detection system utilizing our proposed Abnormal Scores and achieved an F1 score of 88.2% in detecting anomalies within the log data collected from L2 and L3 switches.

Network Function Virtualization (NFV) enables the execution of Virtual Network Functions (VNFs) on standard commodity servers. This brings flexibility, allowing for the rapid deployment of various network services while reducing costs. However, NFV configurations are becoming increasingly complex, necessitating experts for the setup. Intent-based network configuration has emerged as a solution to simplify NFV configuration and management. Nonetheless, it presents challenges, such as translating high-level natural language intents into low-level network configurations. In this work, we propose NFV-Intent - a system that leverages in-context learning in Large Language Models to perform the intent translation task. In-context learning enables NFV-Intent to work without retraining the Large Language Models, which is a difficult and expensive task. NFV-Intent uses a JSON template as the desired output, allowing Large Language Models to learn with a small number of examples and enabling easy verification of the configuration. Our evaluation showed that the intent can be translated into JSON configuration with high accuracy. To demonstrate the feasibility of NFV-Intent, we implemented and integrated it into the NI-testbed, our previously developed system for AI-based NFV life-cycle management.

This study introduces Virtual Machine (VM) Consolidation using a Transformer-based Deep Reinforcement Learning (DRL) method, to address the complexity and inefficiency in operating Software Defined Networks-enabled Network Function Virtualization (SDN-enabled NFV). The distribution of Virtual Network Functions (VNFs) as VMs across servers often leads to energy loss due to irregular deployment. The proposed approach enhances energy efficiency while maintaining the performance of Service Function Chains (SFCs). By refining the VM consolidation process and leveraging a more sophisticated DRL method, this approach promises a more efficient solution to VM consolidation in SDN-enabled NFV environments.

In this study, we propose a three-step anomaly detection system for network switches. The proposed system consists of the following steps: 1) Log parsing, where log messages from switches are analyzed to identify patterns and events, 2) Analysis of the identified event flow to distinguish normal and abnormal event sequences, and 3) Prediction of the next log message, with detection of anomalies if the predicted log message differs from the normal log messages. For event classification, a log parser is proposed by modifying existing algorithms, and experimental results confirm that similar log patterns are correctly classified into the same event. To learn normal event sequences, both FSM and LSTM models are trained. Lastly, we proposed a BERT-LSTM model to predict the next log message and detect unexpected log messages. The proposed system is validated using data collected from a constructed testbed and achieves a high-performance level with an F1 score of 83.72%. Notably, our system achieved a recall of 94.74%. Our system has an advantage in that if misclassified cases occur, network administrators can retrain each model to improve precision during system operation.

Nowadays, Machine Learning (ML) approaches gain a lot of intention for automating and managing Software-Defined Networking, and Network Function Virtualization (SDNNFV) enabled networks. These networks are highly dynamic and flexible due to centralized control and scaleable Virtual Network Functions (VNFs). But the automatic management of the VNF lifecycle in a data center is still challenging; it includes several tasks such as VNF resource usage prediction, placement, consolidation, autoscaling, live migration, etc. So, accurately predicting VNF resource usage can be used for the several tasks mentioned above for performing VNF lifecycle management. It can also help Mobile Network Operators (MNOs) to ensure QoS by reducing Service-Level-Agreement (SLAs) violations. This article introduces an efficient mechanism that uses Adaptive Ensemble Learning to predict resource usage of virtual network functions. This mechanism has three modules: Machine-Learning Predictors (MLPs), Predictor Selector (PS), and Predictor Combiner (PC). The MLPs module contains several ML models for performing prediction. The PS module has a pretrained Random Forest model that is used to choose the best predictors from the MLPs. The PC module combines the selected predictors using an ensemble learning mechanism to generate the final prediction. In tests on three datasets, our method achieved a high${R}^{2}=0.96$for predicting CPU utilization and$R^{2}=0.97$for predicting memory utilization.

Softwarization of the internet network is promising for network service providers (NSPs) to satisfy various types of user requests while dynamically operating the networking system. However, it is challenging to provide optimal service quality as the complexity of the network increases. In particular, deployment of the VNF instances is a critical issue in providing better QoS while maintaining resources optimally. For the management of the VNF deployment task, dynamic programming algorithms are only feasible in small networks or rely on heuristics. In this paper, we propose a VNF deployment method based on reinforcement learning (RL) that can effectively satisfy the QoS with minimized resource consumption in the network. Our approach makes adjustment decisions, which are scale-in/keep/out for target nodes and target VNF types given ILP-based deployments. We formulate this VNF deployment task as RL by setting the reward as QoS and resources. Moreover, we propose a model architecture for our RL agent's policy, based on Graph Neural Network. In the experiment, our approach optimizes VNF deployment with improved QoS while keeping the similar or slightly less amount of resources, compared to ILP-based deployment.

In this study, we present a failure prediction study of VMs and VNFs in an NFV environment. For the proof of concept, we designed a machine learning model to predict the failure with log analysis and observed the cases where the failure-related logs do not exist in the failed VM, but in the server, or in other VMs operating on the same server. Therefore, in this paper, we propose a model which analyzes the logs of all the related VMs and the server and predicts the possibility that any of the VMs operating on the server will fail. To reduce the huge size of the logs collected from the server and VMs, we propose a pre-processing and tagging method that can improve the performance of our model. In addition, we designed a machine learning model using CNN with BERT, which has performed SOTA in various fields of NLP, to receive logs as input and calculate failure probabilities for the next 30 minutes. To validate the proposed model, we collected failure-related logs and normal logs from an OpenStack testbed, and the experimental result shows that the proposed model can predict the failure of VMs operating in the server with an F1 score of 0.74.

In this study, we propose a machine learning model that predicts failures by analyzing logs before failures occur in virtual machines (VMs) used in network function virtualization (NFV) environments. The proposed model utilizes convolutional neural network (CNN) and includes pre-processing and pre-failure tagging techniques. We collected log data from VMs built on OpenStack to validate the proposed model. We classified failures based on early fault messages and built a CNN model to predict VM failures. The experimental results showed that the proposed model can predict failures before 5 minutes with the F1 score of 0.67. The proposed model will be used for VM proactive live migration to avoid service degradation and interruptions caused by failures.

VM (Virtual Machine) live migration is a server virtualization technique for deploying a running VM to another server node while minimizing downtime of service the VM provides. Currently, in cloud data centers, VM live migration is widely used to apply load balancing on CPU workload and network traffic, to reduce electricity consumption, and to provide uninterrupted service during the maintenance of hardware and software updates on servers. It is critical to use VM live migration as a prevention or mitigation measure for possible failure when its indications are detected or predicted. Especially in NFV (Network Function Virtualization) environment, timely use of VNF (Virtual Network Function) live migration can maintain system availability and reduce operator's loss due to service failure. In this paper, we propose a proactive live migration method for vEPC (Virtual Evolved Packet Core) based on failure prediction. A machine learning model learns periodic monitoring data of resource usage and logs from servers and VMs/VNFs to predict future vEPC paging failure probability. We implemented the proposed method in OpenStack-based NFV environment to evaluate the real service performance gains for open source vEPC implementations.

Network Function Virtualization (NFV) and service orchestration has simplified the Service Function Chain management (SFC) tasks, while the edge cloud infrastructure has reduced the latency. Due to these existing technological advantages, there is an urgent need for a dynamic and flexible service chain placement model that performs resource allocation of substrate network in a delay-sensitive and resource-efficient manner. We propose an off-policy Deep Reinforcement Learning algorithm EdgeDQN for efficient SFC placement in the edge cloud environment. The problem of edge resource scarcity is handled by designing a network model that allows worst-case resource renting from neighbors and data centers. This network model is integrated with EdgeDQN using several constraints. This paper aims to find the optimal placement by minimizing the underlying resource utilization and SFC end-to-end delay for multiple SFCs at the same time. To achieve that, an intuitive reward model is proposed. We compare the proposed EdgeDQN algorithm with DQN, Q-learning, and EdgeQL algorithms in terms of performance parameters such as cumulative reward, cumulative standard deviation, latency, and learning convergence time for 420 different test cases. Extensive test results on a simulated and physical (OpenStack) testbed demonstrate the effectiveness of the proposed EdgeDQN algorithm.

Data center networks are designed with multi-rooted topologies to provide the large bisection bandwidth. These topologies provide high path diversity and need a load balancing algorithm to utilize bisection bandwidth efficiently. To distributes traffic efficiently, congestion-aware and fine-grained load balancing algorithms are suggested. However, these algorithms need to set parameters depending on the network states. This paper presents a reinforcement weight-cost multipath (RWCMP) load-balancing method on a data center network to balance the traffic. It uses reinforcement learning to learn the optimal traffic split ratio of egress ports for each node and determines the routes of multiple flows simultaneously. We compared the network utilization of our RWCMP load balancing algorithm with that of the LetFlow algorithm. In a fat tree topology, the RWCMP load balancing algorithm outperforms the traditional load balancing algorithm with 24-36% lower network utilization in three different traffic patterns. In the same topology with a link failure, the RWCMP load balancing algorithm outperforms the traditional load balancing algorithm with 12-27% lower network utilization. These results demonstrate that RWCMP learns traffic split ratio depending on underlying traffic patterns and topology of the data center.

Services in today’s network produce a heavy amount of traffic, and all this traffic might have to pass through a Service Function Chain (SFC). The service provider must place these SFCs in the appropriate locations, and scale them as the request to the service increases. Appropriate placement and scaling of these SFCs will ensure Service Level Agreement (SLA) in terms of throughput, end-to-end delay, or successful serving of the requests. SFC placement and scaling are correlated tasks, however need to be modeled separately. For placing SFCs we require to consider server resources such as memory, CPU, bandwidth, and server locations. On the other hand for scaling, we need to consider resource utilization of VNFs as extra matrixes. Scaling in and out would require an additional cost and time to deploy new VNFs, hence it is important to predict VNF resource requirements in advance and prepare the resources for incoming traffic needs. In this paper we predicted an incoming resource demand using Recurrent Neural Network (RNN) based Gated Recurrent Unit (GRU) model. We predict the resource demands 2 min in advance and deploy the VNFs using the EdgeQ-Leaning model. We call this integrated algorithm as GRU-EdgeQL. To validate our approach we compared it with threshold and random algorithms. We implemented our algorithm in the OpenStack testbed. Our validation reveals that GRU based prediction and EdgeQ-Learning based placement does not only meet the Service Level Agreement (SLA) requirement by reducing the overall latency and SLA violation but also reduces the number of scaling operations thereby reducing the overall scaling cost.

Network softwarization paradigms such as SDN and NFV provide network operators with advantages in terms of scalability, cost and resource efficiency, as well as flexibility. However, in order to fully reap these benefits and cope with new challenges regarding the heterogeneity of user demands and an ever-growing service landscape, management and operation of such networks requires a high degree of automation that ensures fast and proactive decision making. With the recent success of machine learning (ML) across numerous domains, a shift from traditional rule-based policies towards ML-based approaches in the context of network management is taking place. Although many individual contributions cover use cases such as predicting various network characteristics or optimizing the configuration of components, a fully integrated architecture for achievingNetwork Intelligenceis still missing. Hence, in this work, we propose such an architecture that combines the capabilities of softwarized networks with ML-based management. The contribution of this article is threefold: first, we present the proposed architecture alongside its components. Second, we implement a proof-of-concept version of all components in our OpenStack-based testbed. Finally, we demonstrate in a case study regarding VNF resource prediction how the proposed architecture can be used to generate realistic data sets to train and evaluate ML-based models for this task.

By softwarizing the legacy network functions, Network Function Virtualization (NFV) allows rapid development and deployment of network services as well as simplicity and flexibility in network operations and management. Monitoring the performance characteristics of Virtual Network Functions (VNFs), particularly packet processing time, is important to ensure that VNFs are operating correctly with desired performance. This is especially crucial for low-latency network services. In this paper, we present Packet Processing Time Monitoring (PPTMon), a real-time, fine-grained, and end-to-end solution for VNF packet processing time monitoring. PPTMon can provide per-hop monitoring for a single VNF as well as end-to-end monitoring for multiple VNFs in a service function chain. PPTMon allows monitoring in both sampling and continuous fashions. Continuously monitoring every packet may greatly degrade the performance of the VNFs and generate a huge amount of monitoring data. PPTMon’s event-filtering algorithm effectively filters out non-important data and reduces the performance overhead. PPTMon processes packets in-stack by embedding timestamp information directly into the packets, thus further reducing the effect on the VNF performance. PPTMon is implemented on top of extended Berkeley Packet Filter (eBPF) – a Linux framework that allows high-speed packet processing. Our experiment results shows that PPTMon can monitor VNF packet processing time with high accuracy and low impact on performance.

Software-Defined Networking (SDN) and Network Function Virtualization (NFV) help reduce OPEX and CAPEX as well as increase network flexibility and agility. But at the same time, operators have to cope with the increased complexity of managing virtual networks and machines, which are more dynamic and heterogeneous than before. Since this complexity is paired with strict time requirements for making management decisions, traditional mechanisms that rely on, e.g., Integer Linear Programming (ILP) models are no longer feasible. Machine learning has emerged as a possible solution to address network management problems to get near-optimal solutions in a short time. In this paper, we propose a Graph Neural Network (GNN) based algorithm to manage VNFs. The proposed model solves the complex VNF management problem in a short time and gets near-optimal solutions.

With the advent of the 5G network era, it is required to flexibly build and manage networks to meet rapidly changing service requirements. Software-defined networking (SDN) and network function virtualization (NFV) are key technologies that enable flexible network management by transforming networks into software-based networks. Besides, NFV has the advantage of virtualizing network functions, operating those on commercial (COTS) servers, and managing the network functions dynamically. However, the numerous virtual networks and resources created by NFV can cause problems complicating network management. To solve the problems, research on managing complex NFV environments using artificial intelligence (AI) has recently attracted attention. In particular, service function chaining (SFC) is one of the essential NFV technologies, and it is required to create an efficient SFC path in dynamic networks. In this paper, we propose a method of finding optimal SFC path considering the resource utilization of virtual network function (VNF) and VNF placement by using Q-learning, one of the reinforcement learning (RL) algorithms.

Data center network traffic fluctuates over time while applications in the data center require high throughput and low latency. To satisfy these requirements, many load balancing algorithms have been proposed. But existing load balancing algorithms show difficulties in processing short flows and have limited scalability. This paper proposes a load balancing algorithm using probe packets. Probe packet collects detailed network information such as link utilization, hop latency at packet units periodically. The algorithm uses link utilization, hop latency, and queue occupancy to calculate the degree of congestion. The algorithm uses the network status information for load balancing by designating best hops for each hop. Metrics that measure congestion was collected and tested through probe packets in the virtual programmable network environment, and the result shows 24.18% higher performance compared with ECMP.

Reinforcement learning (RL) has been used in various path finding applications including games, robotics and autonomous systems. Deploying Service Function Chain (SFC) with optimal path and resource utilization in edge computing environment is an important and challenging problem to solve in Software Defined Network (SDN) paradigm. In this paper we used RL based Q-Learning algorithm to find an optimal SFC deployment path in edge computing environment with limited computing and storage resources. To achieve this, our deployment scenario uses a hierarchical network structure with local, neighbor and datacenter servers. Our Q-Learning algorithm uses an intuitive reward function which does not only depend on the optimal path but also considers edge computing resource utilization and SFC length. We defined regret and empirical standard deviation as evaluation parameters. We evaluated our results by making 1200 test cases with varying SFC-length, edge resources and Virtual Network Function's (VNF) resource demand. The computation time of our algorithm varies between 0.03~0.6 seconds depending on the SFC length and resource requirement.

Network Function Virtualization (NFV) environment can deal with dynamic changes in traffic status with appropriate deployment and scaling of Virtualized Network Function (VNF). However, determining and applying the optimal VNF deployment in consideration of the cost and Quality of Service (QoS) is a complicated and difficult task. In particular, it is necessary to predict the situation at a future point when the deployment decision is applied because it takes processing time to apply the deployment decision to the actual NFV environment. In this paper, we randomly generate service requests in Multiaccess Edge Computing (MEC) topology, then obtain optimal VNF deployment and Service Function Chaining (SFC) result from an Integer Linear Programming (ILP) solution. We use the simulation data to train a machine learning model which predicts the optimal VNF deployment at a predefined future point. The prediction model shows the accuracy over 90% compared to the ILP solution for the 5-minute future time point.

By enabling the deployment of softwarelized network functions on commodity servers, Network Function Virtualization (NFV) brings many benefits such as rapid development and deployment, simplicity and flexibility in network operations and management. Monitoring the performance characteristics of Virtual Network Functions (VNFs), such as packet processing time, is crucial to achieving maximum benefit from NFV. In this paper, we present Packer Processing Time Monitoring (PPTMon) - a solution for real-time and lightweight VNF packet processing time monitoring. PPTMon embeds timestamp information directly into the packets. PPTMon is implemented using extended Berkeley Packet Filter (eBPF) - a new Linux framework that allows high-speed packet processing. Our experiments showed that PPTMon can monitor VNFs with high accuracy and low performance overhead.

Since the concept of Software-Defined Networking (SDN) and Network Function Virtualization (NFV) has been proposed, telcos and service providers have leveraged these concepts to provide their services more efficiently. However, as the virtual network in the data centers becomes more complex, a variety of new network management problems arise. To deal with these management problems, it is necessary to monitor and analyze resource usage and traffic load of Virtual Network Functions (VNFs) operating on the virtual network. Recently, there have been many attempts to develop technologies that enable network management without human intervention. In this paper, we specify our anomaly detection problem with scenarios involving SLA violations to satisfy the practical needs of network management. Also, we set the real-world NFV environment to generate anomalous data corresponding to each scenario and extend our approach to implementing the system for root-cause localization which identifies the exact VNF instance causing the SLA-related anomalies. We use the datasets collected from the VNFs' service function chain scenarios implemented on OpenStack environment, and compare the accuracy of the anomaly detection models generated by various machine learning algorithms. Our experimental results show the best model has F1-measure over 95% for anomaly detection and 93% for root-cause localization.

Software-Defined Networking (SDN) and Network Function Virtualization (NFV) help reduce OPEX and CAPEX as well as increase network flexibility and agility. But at the same time, operators have to cope with the increased complexity of managing virtual networks and machines, which are more dynamic and heterogeneous than before. Since this complexity is paired with strict time requirements for making management decisions, traditional mechanisms that rely on, e.g., Integer Linear Programming (ILP) models are no longer feasible. Machine learning has emerged as a possible solution to address network management problems to get near-optimal solutions in a short time. In this paper, we propose a Graph Neural Network (GNN) based algorithm to manage Virtual Network Functions (VNFs). The proposed model solves the complex VNF management prob-lem in a short time and gets near-optimal solutions.

Network function virtualization (NFV) is a key technology of the 5G network era. NFV decouples a network function from proprietary hardware so that the network function can operate on commercial off-the-shelf (COTS) servers as a form of virtual network functions (VNFs). Owing to the advantage of NFV, network functions can be applied dynamically to the networks. However, NFV complicates network management because this technology creates numerous virtual resources that should be managed. To solve the problem of complicated network management, studies on applying artificial intelligence (AI) to the NFV-enabled networks, i.e., VNF life cycle management, have attracted attention. In particular, autoscaling, which is one of the essential functions of VNF life cycle management, adds or removes VNF instances to meet service requirements. It is a challenging task to determine the optimal number of VNF instances in dynamic networks, satisfying service requirements. In this paper, we propose a novel auto-scaling method using reinforcement learning (RL) for scale-in/out of multi-tier VNF instances, i.e., service function chaining (SFC) in NFV environments. The proposed approach defines RL's states using a status of SFC composed of multi-tier VNF instances and uses service level objectives (SLO) to make a reward model. We validate the proposed approach in an OpenStack environment, and it shows that our proposed auto-scaling method provides the optimal number of VNF instances in each tier while minimizing SLO violation.

We proposed a best nexthop load-balancing algorithm that helps avoid the congestion path by using global network information, then compared the best nexthop algorithm with the ECMP algorithm. The best nexthop algorithm collects the network information in real time by using In-band network telemetry. The collected network information is hop latency, queue depth, and link utilization. Each switch stores the network information and calculates the degree of congestion as the sum of metric values for each path. When the traffic arrives, each switch divides the traffic into flowlet units and forwards each flowlet to the least-congested path. We compared the throughput of the best nexthop algorithm to that of the ECMP algorithm in three scenarios. In the first scenario, we send 10 flows sequentially and measured the throughput per the number of flows. The best nexthop algorithm shows stable throughput, but the ECMP algorithm shows significant descent throughput after the number of flows exists three. In the second scenario, we send traffic from two different sources to two destinations. The best nexthop algorithm showed 27% higher throughput than the ECMP algorithm in this scenario. In the third scenario, we send a large burst of traffic from a single source to a single destination. The best nexthop algorithm showed 81% higher throughput than the ECMP algorithm in this scenario. These results show that the best nexthop algorithm performs better than the ECMP algorithm in congested status.

Biggest challenge in deploying Service Function Chain (SFC) in the Edge Computing environment is the lack of resources at the edge. Hence while finding the optimum path for SFC deployment, the resource constraint environment should be observed and incorporated well in deployment scenarios. In this paper, we developed an environment aware adaptive Q-Learning algorithm to find an optimal SFC deployment path in edge computing environment. The available servers are divided into hierarchical network structure with local, neighbor, and datacenter servers to model an edge computing environment. The resource dynamics in the environment is modeled as a state transition probability. We compared the new algorithm with our base case algorithm that solely depends on Q-Learning and doesn't incorporate the state transition probabilities. An intuitive reward function is designed to give maximum reward to complex deployment with minimum delays. We integrated our algorithm with physical testbeds using OpenStack and open source REST APIs. We evaluated SFC deployment on physical testbed using 42 different scenarios by measuring RTT.

Network Function Virtualization (NFV) is the key to enable rapid development and deployment of network services as well as simplicity and flexibility in network operations and management. To achieve the maximum benefit of NFV, monitoring the performance characteristics of Virtual Network Functions (VNFs) is crucial. Packet processing time is one of the most important performance metrics when it comes to VNF monitoring. In this paper, we present Packet Processing Time Monitoring (PPTMon) - a real-time, end-to-end solution for VNF packet processing time monitoring. PPTMon can provide per-hop monitoring for a single VNF as well as end-to-end monitoring for multiple VNFs in service function chains. PPTMon works by embedding timestamp information directly into the packets. PPTMon is implemented on top of extended Berkeley Packet Filter (eBPF) - a new Linux framework that allows high-speed packet processing. Our experiment results showed that PPTMon can monitor VNF packet processing time with high accuracy and negligible performance impact.

By decoupling network functions from dedicated, proprietary hardware network devices, Network Function Virtualization (NFV) allows building Virtual Network Functions (VNFs) that can run on standard, commodity servers to reduce cost and gain flexibility in network deployment, operation, and management. However, building VNFs with high-throughput and low-latency is a big challenge. In this paper, we propose eVNF - a hybrid fast-slow path architecture to build and accelerate VNFs with eXpress Data Path (XDP), which is a Linux kernel framework that enables high performance and programmable network processing. The programmability of XDP is limited to ensure kernel safety, thus causing difficulties when using XDP to accelerate VNFs. eVNF solves this problem by taking a hybrid approach: leave the simple but critical tasks inside the kernel with XDP, and let complex tasks be processed outside XDP, e.g., in user-space. With the hybrid architecture, eVNF allows building fast and flexible VNFs. We applied eVNF to build four prototype VNFs: Flow Monitoring (eFM), Firewall (eFW), Deep Packet Inspection (eDPI), and Load Balancer (eLB). These VNFs are evaluated individually and in service function chains (SFCs) using OpenStack. Our experiments showed that eVNF can significantly improve service throughput as well as reduce latency and CPU usage. eVNF-based VNFs also can scale out with the number of CPU cores and can combine with Open vSwitch - Data Plane Development Kit (OvS-DPDK) for better performance.

Since the concept of Software-Defined Networking (SDN) and Network Function Virtualization (NFV) has been proposed, service providers leverage the concepts to provide their services efficiently. Mobile CORD (M-CORD) is a platform to provide 5G network services using containerized VNFs in Multi-access Edge Clouds (MECs) at the central offices of telcos. However, current M-CORD does not include network and resource monitoring functions. In this paper, we present the design and implementation of a monitoring system to help efficient management and operation of the M-CORD platform configured in multi-site. Through this monitoring system, we can monitor the computing resource usages and data traffic load of each VNF container in M-CORD. We conducted experiments on our multi-cluster testbed to evaluate our monitoring system in terms of CPU and memory usage. As a result, our proposed monitoring system has a slight increase in CPU and memory usage, but it was able to perform M-CORD monitoring functions well.

Service Function Chaining (SFC) can be a basic deployment unit that composes a chaining order of required network functions to provide a network service. With the proliferation of Software-Defined Networking (SDN) and cloud computing, such virtualized network functions can be dynamically deployed in different sites, depending on SFC requests. While offering the advantages of flexibility and efficiency, this also leaves management complexity and room for optimization where Machine Learning (ML) can be applicable to solve the problems based on monitoring data. In this paper, we treat SFC as a problem of finding a best source-to-destination routing path from multiple candidates with different link costs and a required traversal order of network functions. There are many mathematical approaches that ensure best optimum but not scalable to the problem size, whereas our approach hides underlying considerations by applying ML technique on measured SFC data to quickly find suboptimal routing paths on a new SFC request, based on their predicted network performance such as the number of successful requests or end-to-end delay. So, we developed a measurement system that records the performance and path costs of SFCs in emulated networks with different per link costs and chain lengths. Then, we evaluate four different ML models for the approach described above.

In addition to providing network operators with benefits in terms of flexibility and cost efficiency, softwarization paradigms like SDN and NFV are key enablers for the concept of Service Function Chaining (SFC). The corresponding networks need to support a wide range of services and applications with highly dynamic temporal profiles and heterogeneous demands. Hence, efficient management and operation of such networks requires a high degree of automation that is paired with fast and proactive decisions in order to cope with these phenomena. In particular, determining the optimal number of VNF instances that is required for accommodating current and upcoming demands is a crucial task that also affects subsequent management decisions. To enable fast and proactive decisions in this context, we propose a machine learning-based approach that uses recent monitoring data to predict whether to adapt the current number of VNF instances of a given type. Furthermore, we present a work flow for generating labeled training data that reflects temporal dynamics and heterogeneous demands of real world networks. In addition to demonstrating the feasibility of the approach in a case study, we provide guidelines regarding the choice of monitoring data that should be collected for reliable prediction as well as the amount of data that is required to train such a predictor.

One challenge of Network Function Virtualization (NFV) is to provide high throughput and low-latency services. In this paper, we propose eVNF - a hybrid architecture to build and accelerate Virtual Network Functions (VNFs) with eXpress Data Path (XDP). XDP is a framework in Linux kernel that enables high-performance and programmable network processing. However, the programmability of XDP is limited to ensure kernel safety, thus causing difficulties in applying XDP to NFV. eVNF solves this problem by taking a hybrid approach: leave the simple but critical tasks inside the kernel with XDP, and let complex tasks be processed outside XDP (e.g., in user-space). eVNF allows building VNFs with both speed and flexibility. We used eVNF architecture to build three VNFs: Firewall (eFW), Deep Packet Inspection (eDPI), and Load Balancer (eLB); then tested them in service function chains. Our experiments showed that eVNF can significantly improve service throughput as well as reduce latency and CPU usage.

In addition to providing network operators with benefits in terms of flexibility and cost efficiency, softwarization paradigms like SDN and NFV are key enablers for the concept of Service Function Chaining (SFC). The corresponding networks need to support a wide range of services and applications with highly heterogeneous requirements that change dynamically during the network's lifetime. Hence, efficient management and operation of such networks requires a high degree of automation that is paired with fast and proactive decisions in order to cope with these phenomena. In particular, determining the optimal number of VNF instances that is required for accommodating current and upcoming demands is a crucial task that also affects subsequent management decisions. To enable fast and proactive decisions in this context, we propose a machine learning-based approach that uses recent monitoring data to predict whether to adapt the current number of VNF instances of a given type. Furthermore, we present a methodology for generating labeled training data that reflects temporal dynamics and heterogeneous demands of real world networks. We demonstrate the feasibility of the approach using two different network topologies that represent WAN and mobile edge computing use cases, respectively. Additionally, we investigate how well the models generalize among networks and provide guidelines regarding the prediction horizon, i.e., how far ahead predictions can be performed in a reliable manner.

Network Function Virtualization (NFV) decouples network functions from dedicated, proprietary hardware into software Virtual Network Functions (VNFs) that can run on standard, commodity servers. One challenge of NFV is to provide high-throughput and low-latency network services. In this paper, we propose eVNF - a hybrid architecture to build and accelerate VNFs with eXpress Data Path (XDP). XDP is a Linux kernel framework that enables high-performance and programmable network processing. However, the programmability of XDP is limited to ensure kernel safety, thus causing difficulties in applying XDP to NFV. eVNF solves this problem by taking a hybrid approach: leave the simple but critical tasks inside the kernel with XDP, and let complex tasks be processed outside XDP, e.g., in user-space. With the hybrid architecture, eVNF allows building fast and flexible VNFs. We used eVNF to build three prototype VNFs: Firewall (eFW), Deep Packet Inspection (eDPI), and Load Balancer (eLB). We evaluated these VNFs in two service function chains using OpenStack. Our experiments showed that eVNF can significantly improve service throughput as well as reduce latency and CPU usage.

Software-Defined Networking (SDN) and Network Function Virtualization (NFV) greatly facilitate network service management. Specifically, these new network paradigms help manage the network environment dynamically and cost-efficiently. Virtual Network Function (VNF) and Service Function Chaining (SFC) are important aspects of the NFV environment. In terms of NFV management, resource demand of VNFs can be predicted at a future time to handle Quality of Service (QoS) and resource allocation problems efficiently. Hence, researchers study and build a management system where machine-learning-based predictions of VNF information are used to handle auto-scaling, deployment and migration of VNFs. In addition, in recent studies, these systems have involved SFC to obtain useful information, not just a lone VNF. However, not many of studies explain clearly how chaining dependency among VNFs in a SFC can be used to predict future resource demand of a VNF. In this paper, we introduce VNF resource prediction machine learning model that maximizes the benefits of using SFC. Then, we compare several machine learning models and analyze how SFC data can help predict resource usage patterns of VNFs. We also show benefits of Attention model to improve prediction accuracy and convergence time through experiments.

Software-Defined Networking (SDN) and Network Function Virtualization (NFV) are paradigms that help administrators to manage dynamic networks. While SDN allows centralized network control, NFV provides flexible and scalable Virtual Network Functions (VNFs). These paradigms are also enablers for concepts such as Service Function Chaining (SFC) where chains are composed of several VNFs to provide a specific service. However, in order to maximize the benefits from the above-mentioned flexibility, new research questions need to be addressed, e.g., regarding effective management processes for dynamic networks. We proposed a novel learning model based on the flexibility of softwarization and abundant volume of monitoring data in NFV environments to predict VNF resource demands using SFC data. Our model is based on Context and Aspect Embedded Attentive Target Dependent Long Short Term Memory (CAT-LSTM) that consists of Target-Dependent LSTM (TD-LSTM), context embedding, aspect embedding, and attention. We developed this model to obtain high accuracy for the prediction of VNF resources such as the CPU. Our model uses two labeling systems: the qualitative resource state and the quantitative resource usage, both of which are used to evaluate its performance. This assists the administrator in understanding the network conditions, improves prediction performance, and provides practically useful information. Our learning model for predicting VNF resource demands can be utilized to solve essential SFC problems such as auto-scaling and optimal placement, which in turn prevent service interruption and provide high reliability.

With the advancement of the Internet technologies, cloud services and various open data, there have been many active attempts to develop mashup services. To develop a mashup service, many data sources and service platforms should be interlinked. Therefore, efficient ways of interconnecting various service platforms and managing data are essential to provide easy service development environment. However, current mashup service development environments do not provide any of them, which forces each developer to fully understand every platforms, interfaces and data format to develop a mahsup service. So, it gives a big burden to mashup service developers and also hinders growth of mashup service industry. In this paper, we propose ICBMS Smart Mediator which connects IoT, cloud, big data, mobile, and security platforms and helps developers to easily utilize them with less overhead and provides easy access to data. We have designed and implemented the proposed ICBMS Smart Mediator and created a new mahsup service with the ICBMS Smart Mediator.

Currently, the Software Defined Networking (SDN) paradigm has attracted significant interests from industry and academia as a future network architecture. SDN brings many benefits to network operations and management including programmability, agility, elasticity, and flexibility. With SDN and OpenFlow, one of the promising SDN protocols, software defined Network Virtualization (NV) techniques can be designed and implemented via flow table segmentation to provision independent virtual networks (VNs). In this paper, we propose an intent based virtual network management platform based on software defined NV. The objective of the proposed NV platform is to automate the management and configuration of virtual networks based on high level tenant requirement specifications, called intents. The design and implementation of the platform is based on ONOS, an open-source SDN controller, and OpenVirteX, a network hypervisor. The platform is designed to provide multiple VNs over the same physical infrastructure to multiple tenants.

LTE-Advanced (LTE-A) theoretically can provide better network performance than 4G LTE. Mobile carriers around the globe are eager to deploy LTE-A to attract more subscribers. However, is it really making difference to the user experience compared to the existing LTE service? To investigate this question, we collected the network performance log from 111 user smartphones on 3G, LTE, and LTE-A. For in-depth analysis, we also asked for privacy information, such as data plan, subscribed network, monthly payment details, and etc. By analyzing the collected log, we observed that (i) LTE-A was faster than LTE only for download bandwidth by 14.1%, yet the users pay on average 13.6% more for LTE-A; (ii) the mVoIP traffic was blocked by all carriers when the users exceed their mVoIP quota; (iii) the subscribed data plan showed no discrimination in network performance.

Mobile network operators are deploying IPv6 to their mobile networks in recent years, to support ever increasing mobile and IoT devices. They also adopt IPv4-IPv6 transition techniques to provide connectivity from their IPv6 networks to IPv4 networks. Since IPv6 deployment to mobile network is still in an early stage, the performance assessment on IPv6 based mobile network is needed, regarding both IPv6 and IPv4. In this paper, we focus on assessing the performance of IPv6 based mobile networks. We first analyze the detailed TCP connection procedures in accordance with the different IP versions that the end hosts use. Based on the analysis result, we suggest improvement points to speed up the connection establishment. We also compare the performance difference between IPv6 and IPv4 by measuring the connection establishment time to 40 popular dual-stack websites. We witnessed that IPv4 can be connected faster than IPv6 in most cases because of the immaturity of the IPv6 infrastructure.

Software-Defined Network (SDN) is an emerging network paradigm which enables flexible network management by separating control plane from data plane. Since SDN adopts a centralized management scheme, with large scale SDN network, it brings intensive overhead to control plane. One way of limiting this overhead is to either distribute the overhead to multiple controllers, or offload the overhead to switches. However, both of the approaches require modification on SDN de-facto standard, therefore, it is not viable for practical use. As an alternative solution, we propose a new control plane management method without changing the underlying SDN protocol. The key idea of the proposed method is to maximize the switch resource utilization by maintaining as much information as possible inside the switches, so that the switches may less frequently query the controller for new information. However, it is non-trivial problem to control the switch resource utilization under the capacity, due to the absence of a detailed correlation model between resource utilization and the affecting parameters. To resolve this issue, our management method adopts a lightweight feedback loop based control scheme - Proportional-Integral-Derivative (PID) to adaptively tune the affecting parameters to minimize the control plane overhead, while avoiding the switch's resource exhaustion. We design and implement the proposed method as an SDN application and evaluate its performance in an emulated network.

The number of data centers has been increased for various reasons such as cloud computing, big-data analysis, multimedia service, etc. With public interests on data center, many novel technologies for data center networks have been proposed and deployed to support data center operations more efficiently and effectively. However, the construction of data center network incurs significant costs. Moreover, various technologies interplay each other to achieve multiple objectives, and it makes difficult to validate and/or verify characteristics of data center network. In addition, it difficult to perform experiments with a number of hosts and switches. Therefore, it is necessary to observe the characteristics of target data center network before building it. A common approach to evaluate data center is to run simulations that should be similar with real-world data center environment. However, generating traffic with the characteristics of data center networks is not matured yet. People still employ a traffic generator based on the characteristics of Internet traffic. We design a traffic generator that shows more accurate characteristics of data center network traffic. Various traffic characteristics exploited explored by several studies are considered. The proposed method generates flow-level network traffic matrix based on Poisson Shot-Noise model. We implemented the traffic generator using Python programming language to create traffic matrix. To evaluate the proposed method, we compare the results with real data center network traffic. Our results show that the generated traffic owns similar characteristics with the real network traffic in terms of flow size, duration, and the mean and variance of total traffic rate.

Voice-over-LTE (VoLTE) is a VoIP-based multimedia service which is provided using All-IP based LTE networks. VoLTE service was first commercialized by Korean telcos in 2012, and now more and more telcos are trying to adopt this technology. With the increased VoLTE service popularity, it is inevitable to have large VoLTE traffic volume (possibly degrading the service quality) and the potential attacks (possibly degrading the service reliability and availability) in the near future. Therefore, in order to avoid such potential issues, we need to perform thorough analysis on VoLTE traffic. As a first step, we propose a VoLTE traffic classification method and its distributed architecture. As the proposed classification method relies on Deep Packet Inspection (DPI) technique, it severely suffers from the large processing time and scalability issues. To overcome these issues, we further propose a distributed architecture for VoLTE traffic classification by adopting a high throughput computing framework — HTCondor. We performed a set of experiments using real-world traces captured from a commercial LTE core network, and have shown that with the proposed architecture, we can achieve up to 23.869 Gbps classification throughput which was almost 35 times faster than the system without distributed processing.

Cloud computing is a popular computing paradigm which provides the virtual resource as a form of VM to customers in an on demand manner. Existing provisioning solutions were only limited to provision computing resource (e.g., CPU, RAM and etc.), therefore, it was difficult to provide more advanced cloud services. Virtual Data Center (VDC) embedding, which is known as mapping the VDC resources to their physical counterparts, was recently introduced to provide more advanced cloud services. However, existing VDC embedding solutions were mostly focus on consolidating VMs in single physical data center. Therefore, in this work, 1) we expand the consolidated targets from VMs to network fabrics (e.g., paths and switches); 2) we also consider the VDC embedding problem in multiple physical data centers. In former point, a Traffic Engineering (TE) technique is required to realize the network fabrics consolidation. While, in latter point, a multi-data center VM live migration technique is required to realize the VM migration across the different data centers. SDN, which is a new network paradigm, can fulfill the requirements raised in above two points by providing flow-level virtualization and network address virtualization. With the help of SDN, we propose a Sdn Assisted Vdc Embedding solution - SAVE by consolidating network fabrics along with physical hosts. SAVE includes two VDC embedding algorithms with a TE heuristic, and a holistic system architecture is provided to realize the proposed algorithms. Our experiment results show the feasibility of the SAVE system architecture, and the effectiveness of the proposed algorithms.

As the current Internet architecture is suffering from scalability issues, the network research community has proposed alternative designs for the Internet architecture. Among those solutions, the Locator/Identifier Separation Protocol (LISP) has been considered as the most promising solution due to its incrementally deployable feature. Despite of various advantages provided by LISP, many ISPs are still conservative to adopt LISP into their production network due to the fact that the standard LISP does not fully satisfy ISP's requirements on LISP-enabled services. In this paper, we propose LISP controller, a centralized LISP management system. By using the LISP controller, we evaluate ISP's three representative LISP use cases: traffic engineering, VM live migration and vertical handover. The results show that the proposed LISP controller allows an ISP to control and manage its LISP-enabled services while satisfying ISP's requirements.

Today's Data Center Networks (DCNs) contain tens of thousands of hosts with significant bandwidth requirements as the needs for cloud computing, multimedia contents, and big data analysis are increasing. However, the existing DCN technologies accompany the following two problems. First, power consumptions of a DCN is constant regardless of the utilization of network resources. Second, due to a static routing scheme, a few links in DCNs are experiencing congestions while other majority links are being underutilized. To overcome these limitations of the current DCNs, we propose a Software Defined Networking (SDN)-based Traffic Engineering (TE), which consists of optimal topology composition and traffic load balancing. We can reduce the power consumptions of the DCN by turning off links and switches that are not included in the optimal subset topology. To diminish network congestions, the traffic load balancing distributes ever-changing traffic demands over the found optimal subset topology. Simulation results revealed that the proposed SDN-based TE approach can reduce power consumptions of a DCN about 41% and Maximum Link Utilization (MLU) about 60% on average in comparison with a static routing scheme.

With the emergence of the Long Term Evolution (LTE) technology, LTE based VoIP services are gaining more and more popularity in these days. Voice over LTE (VoLTE) is a VoIP based multimedia service on All-IP based LTE network, which was initially introduced and commercialized by Korea telecommunication companies in 2012. As more and more subscribers use VoLTE service, the VoLTE traffic volume increases, and this may in turn lower down the VoLTE quality; therefore to preserve QoS of VoLTE service, we need to periodically monitor VoLTE traffic among LTE traffic, and perform appropriate actions to LTE network. To be the first step of realizing this goal, in this paper, we propose a VoLTE traffic classification method and its distributed processing architecture to achieve high throughput. For the classification, we analyze SIP/SDP packets for VoLTE service and generate VoLTE tailored SIP packet signature using SIP User-Agent header field. Since the proposed classification method is relying on Deep Packet Inspection (DPI) method, the classification accuracy can be guaranteed, yet it suffers from the poor processing performance. Therefore, we further propose a distributed and parallel traffic processing architecture with the help of HTCondor.

In this paper we propose a data center traffic engineering method based on multi-objective optimization. We define a problem of traffic engineering in data center networks as a multi-objective optimization problem with two contrary objectives: load balancing and energy saving. Traffic engineering models for load balancing and energy saving are represented as a linear programming equation. Our simulation results show that the proposed method enables data center operators to change the relative importance between load balancing and energy saving. We also show the traffic engineering results with the predefined upper bound of maximum link utilization and energy cost in response to service requirements and energy OPEX. The contributions of proposed traffic engineering method are (a) to minimize both maximum link utilization and energy cost simultaneously; (b) to minimize maximum link utilization with upper bound of energy cost; and (c) to minimize energy cost with upper bound of maximum link utilization.

The number of data centers deployed by governments, enterprises, and universities has been increased affected by the development of cloud computing technologies to reduce CAPAX and OPEX. Many architectures or topologies for data center networks have been proposed to address the diverse purposes and requirements. However, the construction of data centers incurs significant costs. Moreover, there are many technologies that can affect the structure of the data center. Before building a data center, it must be confirmed that it possesses the characteristics necessary to satisfy requirements. Efficient ways to find and confirm network characteristics include simulation and tests using a traffic generation method. Our proposed method is designed to generate network traffic that address many characteristics of data center networks explored by several studies. The proposed method generates network traffic utilizing flow-level traffic matrix, not directly generates packets. We used Python programming language to create traffic matrix and iPerf to generate network packets. To evaluate it, we compared the generation results to real network traffic collected from a data center network. The result shows that the generated traffic is similar with the real network traffic.

With the emergence of significant amounts of mobile and cloud services, the scale of Data Center Networks (DCNs) has been growing rapidly. A DCN has different network requirements compared to a traditional Internet Protocol (IP) network, and the existing Ethernet/IP style protocols constrain the DCNs scalability and its manageability. In this paper, we present a scalable failover method for large scale DCNs. Because most of the current DCNs are managed in a logically centralized manner with a specialized topology and growth model, we adopt Fat-Tree [1] as the reference DCN topology and design our failover method using an OpenFlow-based approach. Further, to provide scalability, we design our failover algorithm in a local optimal manner, with which only three switches must be modified for handling a single fault, regardless of the size of the target network. We evaluate our failover method in terms of failover time by varying the network size and load balancing capability during failover. The experiment results show that our method scales well, even for a large-scale DCN with more than ten thousands hosts.

It is very difficult to identify the location of bottlenecks when there is deterioration of Quality of Service (QoS) because current Internet service is the complicate structure which is provided through a number of ISP. This paper proposes bottleneck detection algorithm which can detect the bottleneck point easily and efficiently even though there is no prior knowledge of the network connection status. The execution procedure of the proposed algorithm can be largely divided into two phases. In the first phase, the proposed algorithm finds the choke point which is suspected to be the bottleneck point using Recursive Packet Train (RPT). In second phase, it identifies the bottleneck point by monitoring packet process time of the detected choke point. For this, this paper proposes two methods. The first method monitors packet process time of the choke point utilizing timestamp option. This method has the advantage which can monitor the bottleneck point without increasing network load because it uses passive monitoring approach using the IP header part of the data packet of the actual users. But agent should be run in all of servers and clients. The second method monitors packet process time of the choke point utilizing ICMP packet. This method is possible to detect the bottleneck point by running agent solely in client without running agent of servers because ICMP packet which is sent to the senders to report the problems found while processing IP packet in routers.

Parallel download protocols that establish multiple TCP connections to distributed replica servers have the potential to reduce file download time and to achieve a server-side load balancing. Unfortunately, parallel download protocols are also inherently unfair towards single-flow downloads and may even starve them. This paper presents TCP-ROME, a parallel download protocol that allows a dynamic mitigation of throughput and fairness. The key novelty is a receiver-based framework that allows a dynamic adjustment of the congestion and rate control of each subconnection. TCP-ROME offers two usage modes: a binary mode where the congestion control of each subconnection can be switched between a TCP-fair rate (high priority) and at a TCP-LP fair rate (low priority), and a more complex range mode where the aggregated throughput aims at meeting a specified target rate. Apart from the protocol specification, this paper develops novel analytical throughput models for TCP-LP and for TCP-ROME. The models are validated via simulations. Extensive simulation scenarios show the flexibility of TCP-ROME in mitigating performance for fairness.

The number of ADSL subscribers in Korea Telecom is remarkably increasing by as many as eight hundred new subscribers every day, with a total number now already at 3 million. Because of such a large number of customers, we need a systematic networking system that can easily accommodate the explosive growth in the number of ADSL subscribers, and efficiently and uniformly manage the various heterogeneous network equipments that are increasing in proportion to the growth of ADSL subscribers. This paper, thus, proposes the distributed networking system architecture for Internet-access service provision using ATM over ADSL. In this paper, we describe the hierarchical network model in deploying ADSL services across the ATM access networks, which can easily accommodate the explosive growth ADSL subscribers in the future. In addition, this paper describes the distributed networking system and its capability to provide a systemic network management using the principal networking concepts of service ordering, addressing, routing, adaptation and switching. All of the networking system components with CORBA objects in favor of the distribution and location transparency are defined and described using the CORBA interface description language (IDL) for commonality. Lastly, we present its implementation and operation in Korea Telecom.

This paper describes an integrated service and network management system for provisioning the point-to-multipoint reservation service (PMRS) in an ATM network. There are two major issues confronting the network service provider in relation to this service: one is to rapidly confirm the acceptability of the subscriber's reservation at subscription time, and the other is to punctually activate the reserved point-to-multipoint service. To meet these requirements, this paper proposes a service provision model (SPM) and a network resource model of bandwidth allocation timetable (BAT). This paper likewise proposes a point-to-multipoint routing algorithm composed of ordering and backtracking procedures, which can find an optimal branch point under the complex network topology and can add more destinations to the existing point-to-multipoint route. Lastly, this paper shows the feasibilities of the SPM, the BAT and the point-to-multipoint routing algorithm by means of implementation and performance analysis under the real ATM network of Korea Telecom.