Network Function Virtualization Orchestration Research Articles

A cost and demand sensitive adjustment algorithm for service function chain in data center network

The introduction of Network Function Virtualization (NFV) and Software-Defined Network (SDN) architectures has significantly reduced the Operational Expenditure (OPEX) and Capital Expenditure (CAPEX) of network system. However, NFV orchestration management also brings about challenges. After the initial deployment of VNFs, due to the volatility of network requests, the original deployment may not be able to meet user resource demands. The key issue is how to readjust resources dynamically to accommodate more network requests without violating Quality of Service (QoS) for users. Several existing techniques can be used to achieve this goal, such as horizontal scaling, vertical scaling, and virtual network function (VNF) migration. However, these techniques inevitably incur some overhead, such as the cost of instantiating VNF and link rerouting. Additionally, resource adjustment may also result in unbalanced distribution of network resources. In this paper, an Intelligent Service Function Chain Dynamic Adjustment Algorithm (ISFCDAA) is proposed to address the above challenges. Firstly, an Integer Linear Programming (ILP) model is established with the objective of minimizing the long-term adjustment cost and reducing the imbalance of resource distribution. Then we transform the optimization process into a Markov Decision Process (MDP). Secondly, to solve the problems that the state and action space is too large and the state transition probability is uncertain in MDP, a SFC dynamic adjustment algorithm based on deep reinforcement learning is proposed. This algorithm can obtain an approximate optimal adjustment strategy for ILP model. The simulation results show that ISFCDAA can reduce the adjustment overhead and maintain a balanced distribution of network resources while ensuring the QoS. Compared with the existing algorithms, the average standard deviation of resource distribution of ISFCDAA is reduced by up to 9.90%, the average acceptance rate of ISFCDAA is improved by up to 39.57%, and the average long-term profit is improved by up to 42.92%. The incorporation of cost and demand-sensitive considerations into ISFCDAA enhances its responsiveness to fluctuating network demands, solidifying its effectiveness in dynamic resource management scenarios.

Comparative Study of Orchestration using gRPC API and REST API in Server Creation Time: An Openstack Case

Cloud computing is the quick, simple, and economical distribution of computing services. To offer quality services in cloud infrastructure, the NFV Management and Orchestration (MANO) function is a critical task for managing both infrastructure resources and network functions. Since the demand for cloud services is rapidly increasing, cloud service providers are constantly trying to reduce operational costs. Hence the elements of MANO functions have gained critical importance in cloud infrastructure. To use MANO efficiently, there are hypervisors (OpenStack) that act as a middle layer between hardware and software that are used to provide Infrastructure-as-a-Service solutions through a set of interrelated services. These services are managed by an application programming interface (API). The existing orchestration method uses Representational state transfer (REST) API for communicating with core components of a hypervisor. These API have been integral part of creating and deploying a server. To improve the performance, this paper proposes a novel approach for orchestration based on the opensource remote procedure call framework, known as gRPC. To analyse the proposed solution, both REST and gRPC-based orchestration methods are implemented for creating and launching servers using OpenStack cloud computing platform. The server creation time is obtained for different scenarios and SFC use cases. The results show that by using gRPC, the performance in terms of server creation time of an orchestration is improved by up to 27% compared to the traditional REST-based orchestration.

Performance Evaluation of a Novel Intrusion Detection System in Next Generation Networks

The integration of Internet of Things (IoT) with 5G simply creates additional threat landscape and any network infrastructure is more vulnerable. Severe attacks on networks potentially damage organization reputation, customers or tenants lose confidence, and impacts operational and maintenance cost. Intrusion detection systems (IDSs) are an effective approach to mitigate threats. We present a novel IDS mechanism in which the unique Radio Frequency (RF) features of IoT devices are used to create a learning model which is later used to identify the illegitimate devices in the network. Leveraging the Deep Autoencoder (DAE), the existing steady-state feature extraction is generalized. The performance evaluation is conducted using a real data set from different aspects including the mobility of the nodes. The proposed IDS is broken down into pluggable virtual network function (VNF) components and its evaluation is presented for its integration into the 5G network slicing ecosystem from the perspective of the European Telecommunications Standards Institute (ETSI) standards. A Proof of Concept (PoC) is presented using ETSI Open Source NFV Management and Orchestration (OSM-MANO) test bed, deployed on AWS cloud systems, to show how the proposed approach would fit in with a real-life MANO.

Toward IoT Node Authentication Mechanism in Next Generation Networks

Although the next generation networks (5G-NGNs) provide a flexible infrastructure to support latency-sensitive and bandwidth-hungry mission-critical Internet of Things (IoT) applications, however, the 5G-IoT integration in NGNs has increased the threat surface. Unfortunately, IoT devices are resource constrained, and the traditional Intrusion Detection Systems (IDS) approaches based on cryptography are not effective on 5G-IoT ecosystems. In this paper, we propose an effective 5GIoT node authentication approach that leverages unique RF (Radio Frequency) fingerprinting data to train the Deep learning model to detect legitimate and non-legitimate IoT nodes. Our approach is based on Mahalanobis Distance theory and Chisquare distribution theories. The proposed approach achieves a higher detection accuracy (99.35%) as well as lower training time compared to other existing approaches which is a key benefit of our approach in NGNs. The experiments are conducted using ETSI-Open Source NFV Management and Orchestration (OSM-MANO) platform on AWS (Amazon Web Services) cloud platform to verify how the proposed approach would fit in reallife scenarios. The method can be used as a standalone security system or as a part of multi-factor authentication.

SAP: Subchain-Aware NFV Service Placement in Mobile Edge Cloud

Existing Network Function Virtualization (NFV) service placements that reuse already deployed network functions either reuse an entire Service Function Chain (SFC) or only individual network functions while ignoring the chain configuration cost for configuring the SFC traffic steering and ignoring the reliability of the network functions. Also, the Mobile Edge Cloud (MEC) frameworks that are required to implement an NFV service placement should ideally seamlessly cooperate with the various existing NFV Management and Orchestration (MANO) frameworks. However, the existing MEC frameworks lack multi-MANO support. We formulate the novel Subchain-Aware NFV service Placement (SAP) optimization model that accounts for the configuration cost for stitching together reused network functions to an SFC and strives to reuse existing subchains of consecutive network functions (with already deployed SFC traffic steering), while accounting for the recovery cost of network functions with limited reliability. We develop Tabu-SAP, a Tabu search approach to solve the SAP optimization problem. Furthermore, we introduce the novel Automated Provisioning framework for MEC (APMEC) with open-source OpenStack implementation to enable the deployment of Tabu-SAP in real networks; APMEC supports multiple MANOs through a loose coupling MANO-MEC design. Our Tabu-SAP evaluations indicate an around eightfold increase of the number of supported SFCs compared to the state-of-the-art reuse of individual network functions, while substantially reducing the total cost, which includes the chain configuration cost. Also, for long SFCs of seven or more network functions, the Tabu-SAP total cost is less than 10% higher than the optimal solution (which requires over ten times longer execution time).

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

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

Intent-based network slicing for SDN vertical services with assurance: Context, design and preliminary experiments

Network slicing is announced to be one of the key features for 5G infrastructures enabling network operators to provide network services with the flexibility and dynamicity necessary for the vertical services, while relying on Network Function Virtualization (NFV) and Software-defined Networking (SDN). On the other hand, vertical industries are attracted by flexibility and customization offered by operators through network slicing, especially if slices come with in-built SDN capabilities to programmatically connect their application components and if they are relieved of dealing with detailed technicalities of the underlying (virtual) infrastructure.In this paper, we present an Intent-based deployment of a NFV orchestration stack that allows for the setup of Qos-aware and SDN-enabled network slices toward effective service chaining in the vertical domain. The main aim of the work is to simplify and automate the deployment of tenant-managed SDN-enabled network slices through a declarative approach while abstracting the underlying implementation details and unburdening verticals to deal with technology-specific low-level networking directives. In our approach, the intent-based framework we propose is based on an ETSI NFV MANO platform and is assessed through a set of experimental results demonstrating its feasibility and effectiveness.

Scalable and Flexible Traffic Steering for Service Function Chains

Network Function Virtualization (NFV) has inspired numerous orchestration algorithms to decide Virtualized Network Function (VNF) placement and routing paths for service requests with Service Function Chain (SFC) demands. With different optimization goals, these algorithms may select various routing paths for request flows. Nevertheless, existing traffic steering solutions either fail to fully support NFV Orchestrator (NFVO) in path selection, or result in low scalability in the underlay Software Defined Network (SDN). In this paper, we propose STAR to tackle both problems simultaneously. STAR divides the entire routing path for SFC request into several path segments, then performs Output-Port-based, Default-Path-based or RSP-ID-based traffic steering on each path segment. With the idea of path division and these traffic steering mechanisms, STAR achieves flexible traffic steering along any paths selected by NFVO and enables different Rendered Service Paths (RSPs) with the same path segments to share the same forwarding rules. In this way, STAR ensures the correct enforcement of path selection decisions from NFVO and significantly reduces forwarding rule consumption and control overhead. We evaluate STAR with the experiments on a real testbed and large-scale simulations. The results show that our framework is scalable in the SDN data plane and control plane (e.g., reducing rules in hardware switches by more than 70% compared with NSH-based solutions) while retaining the flexibility of steering traffic along any SFC routing paths to provide full support for path decision enforcement (e.g., achieving the highest path decision enforcement ratio in 11 of the 12 cases) with acceptable overhead.

Joint optimization of Service Chain Graph Design and Mapping in NFV-enabled networks

Network Function Virtualization (NFV) is an emerging approach to serve diverse demands of network services by decoupling network functions and dedicated network devices. Traffic needs to traverse through a sequence of software-based Virtual Network Functions (VNFs) in a preset order, which is named as Service Function Chain (SFC). Since network operators usually deploy the same type of VNFs in different locations in NFV-enabled networks. How to steer a SFC request to an appropriate path in substrate networks to meet service demands becomes an important issue, which is typically known as SFC mapping. However, the existing research works on SFC mapping often assume that service chain graphs are given in advance. They do not consider VNF interdependency and traffic volume change, which are both theoretically challenging for NFV Management and Orchestration (MANO) framework. To this end, we study the joint optimization of Service Chain Graph Design and Mapping (SCGDM) in NFV-enabled networks. Our objective is to minimize the maximum link load factor to improve the performance of network system. We first formulate the SCGDM problem as an Integer Linear Programming (ILP) model, and prove that it is an NP-hard problem by reduction from a classical Virtual Network Embedding (VNE) problem. Further, we develop an approximation algorithm using randomized rounding method and analyze the approximation performance. Extensive simulation results show that the proposed algorithm effectively reduce the maximum link load factor.

Model-Driven Availability Assessment of the NFV-MANO With Software Rejuvenation

Network Function Virtualization enables network operators to modernize their networks with greater elasticity, network programmability, and scalability. Exploiting these advantages requires new and specialized designs for management, automation, and orchestration systems which are capable of reliably operating and handling new elements such as virtual functions, virtualized infrastructures, and a whole new set of relationships among them. Operations such as resource allocation, instantiation, monitoring, scaling, or termination of virtual functions are key lifecycle operations that NFV management and orchestration (NFV-MANO) frameworks need to correctly perform. Failures of the NFV-MANO prevent the network ability to respond to new service requests or events related to the normal lifecycle operation of network services. Thus, it is important to ensure robustness and high availability of the MANO framework. This paper adopts a model-driven approach to predict the availability of the NFV-MANO and assess the impact that different failure modes have. We propose different models, based on Stochastic Activity Networks (SANs), which abstract various MANO deployment configurations, inspired by current containerized open-source MANO implementations. Moreover, we integrate software rejuvenation and investigate the trade-off between its associated overhead and system availability increase. An extensive experimental campaign with fault-injection techniques on a real-life MANO implementation allows to derive a number of realistic recovery parameters. The case studies are used to quantitatively evaluate the steady-state availability and identify the most important parameters influencing system availability for the different deployment configurations.

OpenPATH: Application aware high-performance software-defined switching framework

Currently, core networking architecture is facing disruptive developments, due to the emergence of SDN for control, NFV for services and so on. SDN promises more versatility in routing and managing traffic flows, while NFV represents a large shift in how network functions and services are built, deployed, and managed. We present OpenPATH (aPplication Aware software-defined swiTcHing framework)—A software-defined switching framework for NFV processing and orchestration of Network Functions (NFs) and steering the flows through service chains. Inspired by the potential benefits of encapsulating the application logic into the SDN dataplane, OpenPATH is built on the concept of a modular dataplane, which consists of two layers - switching fabric layer to control packet forwarding; and switch management layer, which inspects the incoming packets, steers the flows through a sequence of NFs and determines the next forward/drop action. The application logic of the NFs can be introduced and pushed to the dataplane at runtime and the framework offers fast packet processing and I/O functionalities to support NF parallelism in the Service Function Chaining (SFC) scenarios. OpenPATH is a modular framework for software switches and offers flexibility for programming run time functions depending on the dynamic behavior of the network traffic and cyberattacks. The architecture components are not hard-coded or rigidly implementations in conventional switches/bridges and standard OpenFlow based SDN stacks. The design allows the vendors, operators, or developers to configure policies at run time and deploy custom logic and NF (also series of NFs) through software programs embedded in the switching fabric. While the basic concept is similar to some pioneering works in this area, OpenPATH does not sacrifice portability, performance, or security for programmability. The OpenPATH as a programmable switching platform takes a different approach to meet most of the requirements of application-aware and intent-based networking. OpenPATH helps administrators to quickly configure network security services using a rich set of standard APIs, with simplified flow tables. The evaluation shows that our design can leverage complex states in the data plane without overloading the SDN controller. Compared to conventional SDN methods, this provides much greater versatility and precision. The key findings indicate that OpenPATH achieves lower cost for scaling, higher overall throughput, and reductions in latency for real-world service chains.

NFV orchestration over disaggregated metro optical networks with end-to-end multi-layer slicing enabling crowdsourced live video streaming

Network infrastructure must support emerging applications, fulfill 5G requirements, and respond to the sudden increase of societal need for remote communications. Remarkably, crowdsourced live video streaming (CLVS) challenges operators’ infrastructure with tides of users attending major sport or public events that demand high bandwidth and low latency jointly with computing capabilities at the networks’ edge. The Metro-Haul project entered the scene proposing a cost-effective, agile, and disaggregated infrastructure for the metro segment encompassing optical and packet resources jointly with computing capabilities. Recently, a major Metro-Haul outcome took the form of a field trial of network function virtualization (NFV) orchestration over the multi-layer packet and disaggregated optical network testbed that demonstrated a CLVS use case. We showcased the average service creation time below 5 min, which met the key performance indicator as defined by the 5G infrastructure public private partnership. In this paper, we expand our field trial demonstration with a detailed view of the Metro-Haul testbed for the CLVS use case, the employed components, and their performance. The throughput of the service is increased from approximately 9.6 Gbps up to 35 Gbps per virtual local area network with high-performance VNFs based on single-root input/output virtualization technology.

On the Fly Orchestration of Unikernels: Tuning and Performance Evaluation of Virtual Infrastructure Managers

Network operators are facing significant challenges meeting the demand for more bandwidth, agile infrastructures, innovative services, while keeping costs low. Network Functions Virtualization (NFV) and Cloud Computing are emerging as key trends of 5G network architectures, providing flexibility, fast instantiation times, support of Commercial Off The Shelf hardware and significant cost savings. NFV leverages Cloud Computing principles to move the data-plane network functions from expensive, closed and proprietary hardware to the so-called Virtual Network Functions (VNFs). In this paper we deal with the management of virtual computing resources (Unikernels) for the execution of VNFs. This functionality is performed by the Virtual Infrastructure Manager (VIM) in the NFV MANagement and Orchestration (MANO) reference architecture. We discuss the instantiation process of virtual resources and propose a generic reference model, starting from the analysis of three open source VIMs, namely OpenStack, Nomad and OpenVIM. We improve the aforementioned VIMs introducing the support for special-purpose Unikernels and aiming at reducing the duration of the instantiation process. We evaluate some performance aspects of the VIMs, considering both stock and tuned versions. The VIM extensions and performance evaluation tools are available under a liberal open source licence.

Octans: Optimal Placement of Service Function Chains in Many-Core Systems

Network Function Virtualization (NFV) offers service delivery flexibility and reduces overall costs by running service function chains (SFCs) on commodity servers with many cores. Existing solutions for placing SFCs in one server treat all CPU cores as equal and allocate isolated CPU cores to network functions (NFs). However, advanced servers often adopt Non-Uniform Memory Access (NUMA) architecture to improve the scalability of many-core systems. CPU cores are grouped into nodes, incurring performance degradation due to cross-node memory access and intra-node resource contention. Our evaluation shows that randomly selecting cores to place NFs in an SFC could suffer from 39.2 percent lower throughput comparing to an optimal placement solution. In this article, we propose Octans, an NFV orchestrator to achieve maximum aggregate throughput of all SFCs in many-core systems. Octans first formulates the optimization problem as a Non-Linear Integer Programming (NLIP) Model. Then we identify the key factor for problem solving as evaluating the throughput drop of an NF caused by other NFs in the same SFC or different SFCs, i.e., performance drop index, and propose a formal and accurate prediction model based on system level performance metrics. Finally, we propose two online algorithms to quickly find near-optimal placement solutions for one-time and incremental deployment. Extensive evaluation on a prototype implementation shows that Octans significantly improves the aggregate throughput comparing to two state-of-the-art placement solutions by 27.1 ~ 45.2 percent for one-time deployment and by 20.9 ~ 38.1 percent for incremental deployment, with very low prediction errors. Moreover, Octans could quickly find a near-optimal placement solution with tiny optimality gap.

Energy and Cost Efficient Resource Allocation for Blockchain-Enabled NFV

Network function virtualization (NFV) is a promising technology to make 5G networks flexible and agile. NFV decreases operators' OPEX and CAPEX by decoupling the physical hardware from the functions they perform. In NFV, users' service request can be viewed as a service function chain (SFC) consisting of several virtual network functions (VNFs) which are connected through virtual links. Resource allocation in NFV is done through a centralized authority called NFV Orchestrator (NFVO). This centralized authority suffers from some drawbacks such as single point of failure and security. Blockchain (BC) technology is able to address these problems by decentralizing resource allocation. The drawbacks of NFVO in NFV architecture and the exceptional BC characteristics to address these problems motivate us to focus on NFV resource allocation to users' SFCs without the need for an NFVO based on BC technology. To this end, we assume there are two types of users: users who send SFC requests (SFC requesting users) and users who perform mining process (miner users). For SFC requesting users, we formulate NFV resource allocation (NFV-RA) problem as a multi-objective problem to minimize the energy consumption and utilized resource cost, simultaneously.

Toward a Real Deployment of Network Services Orchestration and Configuration Convergence Framework for 5G Network Slices

A seamless interworking between network function virtualization (NFV) and software defined networking (SDN) to orchestrate network services for the 5G systems is very fundamental for network slice creation. The orchestration of large scale network slices across multiple administrative as well as technological domains with heterogeneous resources and a distributed form of slice management can benefit from harnessing existing NFV orchestration (NFVO) solutions. In this regard, this article presents a network service orchestration and configuration convergence framework that is capable of providing a large scale network slicing solution for 5G network operators. Using this framework, 5G network operators can orchestrate and configure network slices directly from their infrastructure and that of credible registered slice providers who have resources for the orchestration of only a subset of the overall network slice. The framework is equipped with mechanisms that allow a distributed form of slice configuration and management.

A Novel Autonomous Profiling Method for the Next-Generation NFV Orchestrators

Currently, telecommunication research communities are striving towards the adoption of Zero-touch network and Service Management (ZSM) in Network Function Virtualisation (NFV) orchestration. Contemporary efforts on adopting Machine Learning (ML) and Artificial Intelligence (AI) have caused an upsurge of ZSM application in the VNF space. While ML and AI complement the ZSM goals for building the intelligent NFV orchestration, a deep knowledge about the resource consumption by Network Services (NSs) and its constituent Virtual Network Functions (VNFs) is required, which would enable AI and ML models to manage the available resources better and enhance user experience. In this article, we propose a Novel Autonomous Profiling ( NAP ) method that not only predicts the optimum network load a VNF can support but also estimates the required resources in terms of CPU, Memory, and Network, to meet the performance targets and workload by utilising ML techniques. Our performance evaluation results on real datasets show that the output of NAP can be used in the next generation of NFV orchestration.

The Need for Advanced Intelligence in NFV Management and Orchestration

With the constant demand for connectivity at an all-time high, Network Service Providers (NSPs) are required to optimize their networks to cope with rising capital and operational expenditures required to meet the growing connectivity demand. A solution to this challenge was presented through Network Function Virtualization (NFV). As network complexity increases and futuristic networks take shape, NSPs are required to incorporate an increasing amount of operational efficiency into their NFV-enabled networks. One such technique is Machine Learning (ML), which has been applied to various entities in NFV-enabled networks, most notably in the NFV Orchestrator. While traditional ML provides tremendous operational efficiencies, including realtime and high-volume data processing, challenges such as privacy, security, scalability, transferability, and concept drift hinder its widespread implementation. Through the adoption of Advanced Intelligence techniques such as Reinforcement Learning and Federated Learning, NSPs can leverage the benefits of traditional ML while simultaneously addressing the major challenges traditionally associated with it. This work presents the benefits of adopting these advanced techniques, provides a list of potential use cases and research topics, and proposes a bottom-up micro-functionality approach to applying these methods of Advanced Intelligence to NFV Management and Orchestration.

Improved QoS at the Edge Using Serverless Computing to Deploy Virtual Network Functions

Multiaccess edge computing (MEC) will strengthen forthcoming 5G networks by improving the Quality of Service (QoS), in particular, reducing latency, increasing data processing rates, and providing real-time information to develop high-value Internet-of-Things (IoT) services. To enable data-intensive network services and support advanced analytics, many network operators have proposed to integrate MEC systems with network function virtualization (NFV) consolidating virtual network functions (VNFs) and edge capabilities on a shared infrastructure. As of yet, this integration is not fully established, with various architectural issues currently open, even at standardization level. For instance, any update to VNFs deployed in a MEC system requires a time-consuming manual effort, which affects the overall infrastructure operations. To address these pitfalls, VNFs can be decomposed into microservices, which maintain their own states and exhibit different resource consumption requirements. This article presents an approach to integration that leverages serverless computing to merge MEC and NFV at the system level and to deploy VNFs on demand, by combining MEC functional blocks with an NFV orchestrator using a Kubernetes cluster. We further investigate whether the resource utilization of a MEC system can be improved by leveraging networked FPGA-enabled MEC servers, through an extension of the edge layer that takes advantage of available programmable hardware. We quantitatively evaluate and demonstrate the improvement of 75% end-to-end latency, 99.96% VNF execution time, 26.9% resource utilization, and 15.8% energy consumption in comparison with traditional baselines of cloud, edge, and serverless-edge test cases for a high-definition real-time video streaming application.

Delay-constrained NFV orchestration for heterogeneous cloud networks

Recently, network services are increasingly connecting computational elements within and across datacenters. In the Network Functions Virtualization (NFV) environment, to successfully orchestrate a network service, first a VNF-Forwarding Graph (VNF-FG) must be composed that realizes the required functionality. Second, this VNF-FG must be embedded onto the infrastructure, that is increasingly becoming heterogeneous. To efficiently allocate service demands, intelligent mechanisms and algorithms are needed to effectively tailor the VNF-FG to the cloud network onto which the service will be deployed. This paper introduces the first service model supporting on-demand orchestration of services with bidirectional chaining and delay constraints, optional VNFs and traffic aggregation. Building on this service model, we propose algorithms looking for the optimization of the order and number of VNF instances, to adapt the VNF-FG to the available resources in the substrate network. Numerical experiments show that, through coordination of composition and embedding tasks, our proposed algorithms can significantly improve the acceptance ratio, compared to algorithms that perform these tasks in two separate stages.