 Hi everyone, my name is Muriel Franck, I am a General Research at the University of Zurich, and I have dedicated my time in research about NFV issues and problem identification in the past three years, part in Brazil, part in the University of Zurich, in Switzerland, and the main idea of this talk is to investigate literature and try to highlight some research trends in the academy about NFV issues, and also try to identify some future directions for NFV at all. So here we have a brief agenda about this presentation. First I will introduce some basic concepts about computer networks and NFV. I will introduce some general challenges since NFV was released that is well explored during the last years, and highlight the research trends by using an approach to literature investigation, also provide some market opportunities regarding the market of NFV in the future, and conclusions try to highlight and forecast some things about how NFV can evolve and be largely adopted in the next years. So first we need to define a basic concept from computer networks, if you notice what is network function, I can define that network function is a network node or a physical hardware, a middle box, if I want to define a behavior, for example, it can be a simple IP table based firewall or also a complex network accelerator or compressor, for example. When I use it, for example, connect in a network service, Google, Netflix, and so on, the user connection is for audit between different middle box, that can be a DPI, a firewall, a quite off-serve accelerator, a little banser, and so on. And what is the problem with the middle box? Most of them is very expensive, it's not easy to manage and make some configuration operations. When we have a lot of middle box to manage, then we have to expend a lot of money with the team to go to make some, to correct some problems in the middle box or to manage or to deploy. And this middle box, in the most of the time, is very expensive. So for this, NFV has immersion as a technology that not only reduces the cost and simplifies the management, but also allows us to make some evolution in the network core. The basic NFV allows us to virtualize session work functions, move them from middle box, dedicated middle box, to run any kind of virtualization solution. So it is a very good opportunity about cost reduction and provides some agility and flexibility regarding NFV, regarding networks at all. For example, one of the main benefits from NFV is deliver agility and flexibility. We can, for example, make available a service, a complete service or a network function to supply some specific demands in a few seconds. And also we have flexibility to research a location and to configure a network service, a network function, to supply specific demands for each user. We have also a time to market. For example, we can develop a network function and make it available for a customer or a end user, a self-provider in a few seconds also. And of course we can reduce the cost, not only the capital cost, because you don't need to buy middle box more. We can use commodity off-the-shelf servers. And we have also reducing the operation expensive, because we can use, for example, autonomous systems, autonomous platforms or solutions to deal with a lot of network functions at the same time and to deal with a lot of information that will be available for management orchestrate such network functions. So NFV is very nice. Okay, we can reduce costs, accelerate time to market, but it's not simple to do. It's not only we visualize a network function and it works fine to compound a service or to provide a complete service in an easy way. So, in force I required many to provide an efficient deploy of such VNFs to orchestrate and also manage such VNFs in an easy way. So for this NFV initial idea was released by ETSI, that is European Telecommunications Standards Institute. It was released as a white paper in 2012 and says that a lot of standards and reports to define which we need to implant NFV environment was released. I can highlight one of the most important, in my opinion, that was the first use case for NFV that was released in 2013. It was basically to us five use cases, which includes the visualization of mobile core networks and content delivery networks. So we can make evolution in the core of network and to the edge of the computer networks. In the same year, the concept architect of NFV was released. It is the standard that we need to follow to implant or to implement solutions that will deal with NFV and return to work functions at all. In 2014, the management of the Cossetia framework was proposed by ETSI also. I can say that this framework is the base for any solution regarding the management of VNFs, both to Academy and industry. So we need to use this framework as base to build some solution to manage the Cossetia. Here we have the concept architect of NFV. So we have basically an NFV shooter that is compounded by a substrate, an arbitralization layer. Here we have the arbitralization resource. And the VNF is hosting up to this NFV shooter. We have an element management to provide some APIs to communicate other solutions with the VNFs. And the most important, I can say, that is these three components that would compound the management of the Cossetia framework. This is the orchestrator, the VNF manager, and the virtualized infrastructure manager. The orchestrator is in charge of a global management, mainly focused on resource management and service management. The VNF manager controls all of the life cycle of a VNF, and the virtualized infrastructure manager is in charge of counter the resource allocation and all of the NFV infrastructure. So this architecture, for example, can be constructed using different open source frameworks and platforms available. For example, you can construct the NFV infrastructure by using OpenStack or OpenDelight or others. We have a lot of well-defined frameworks to manage orchestration. For example, I can say the OpenStack attacker, that is a good solution because we have a good integration with OpenStack utilization. So we can use OpenStack attacker, for example, to orchestrate and manage the NFV that is running in an OpenStack-based arbitralization system. I have an open source manual that was released in 2014. It's a very nice project about management orchestration of VNFs. I have OpenBattle, and of course, I have all the relevant solutions for NFV already available. For example, the Linux foundation maintain the OP NFV project, which is a collaborative platform for developing and evolution of an NFV, is a consortium of a lot of different open source companies that is trying to provide and incentive the evolution and adoption of NFV in large scale. So, okay, now there are some challenges for NFV that have been, since it was released. But take this challenge placement, management orchestration, and security. In mind, we can analyze and have some ideas about the current state of the art about NFV. First of them, there is a classical problem for NFV that is the placement. Basically, the question is, where is the best place to deploy a VNF in an efficient way? Of course. For example, we can chain one or more VNFs to build a complete service, and it's not a trivial issue, because we need to know where is the best, the best infrastructure or the best place or best technology to use, for example, to compound a security service. You can deploy a file, a DPI, and an IPS, for example. Okay, how can you deploy such network functions, such VNFs, to construct a service? So we need to take into account some important issues about latency, quality of service, resources available, and of course, monetary cost, that is one of the main issues about use in NFV. Another important challenge, general challenge, is the management orchestration of written work features. For example, how to orchestrate and bond such VNFs, we need to build some monetary solutions, APIs, we have a lot of challenges and space for works in such area. And when we can migrate it, for example, how to identify some misbehaviors or some problems in VNF or in the SFC, the SFC from Tuncini, to know I need to migrate such VNF because this is a bottleneck in the service at all. So it's a very important question. And a lot of techniques to identify misbehaviors, bottlenecks and other VNF related problems have been exploited by academy. Most of them, I try to use artificial intelligence and information visualization to detect and identify such problems. And other important question, of course, as computer networks at all, is security. We need to provide trust communication among NFV components. I mean, orchestrate and VNF management, for example, can be placed in an infrastructure and needed to manage and orchestrate VNFs or a composition of VNFs that can be deployed around the globe in different facilities. So we need to provide trust communication between such components. I issue that, I think, requires attention in terms of security. And the interoperability among different vendors is very important because, for example, if I develop a VNF and had had to develop other VNF and I want to use both VNFs to compound the services and I want to use OpenStack attacker, for example, orchestrate such VNFs. I need to ensure that different VNFs from different vendors will be integrated and have interoperability with OpenStack attacker, for example. So it's an important question also. We need to guarantee also interoperability between, for example, legacy network functions. I can use, for example, a middle box to put, in the middle of ISFC, a VNF, a middle box, a VNF, for example, how it can work. So it is relevant to questions, not only about security, but about other issues. And of course, we have treats and countermeasures. I think as NFV arise as a solution, as a real solution, I think a lot of new treats will be arise. And countermeasures to deal with such new treats is a challenge to the next few years for deal with some security aspects about NFV and NFV environments at all. So now we will highlight, identify some research trends, by analyzing a lot of important congress and conference in the computer networks and computer management. I can identify three main research trends in computer networks. The three is management orchestration-related issues. That is a vast scenario of problems we have. NFV, as I enable for 5G networks, is a really important aspect also there are a lot of resources about that. And business models to build a lucrative and profitable market landscape to NFV in the future. So, okay, one of current resources talk related to management orchestration framework is how to detect VNF performance and identify bottlenecks in a SFC, in a service from Cruchini. For this, we have a lot of works in the Academy exploring real monetary solutions to take on this problem. For example, we have an FV Perf that was proposed in NFV CDN 2016, three years ago. Basically, it is a solution that makes some performance monetary and bottlenecks detection tool in NFV and SFCs at all. It basically needs packets and our VM to VM communication path computes the delays and throughputs and uses these measurements to identify bottlenecks in real time. The solution, for example, is the Perf Shacker that was proposing a work published in the network and management conference in the past year in 2018. Such a work monitors the elements of the software data plane and provides performance reports, for example, graphs and visualizations to highlight bottlenecks into the NFV service. Other candidates to solve this kind of problem that is identify performance bottlenecks, I can see that is machine learning based solutions. We have a lot of works in this direction. For example, in this work that was published in the communication magazine in January in the past year, the authors discuss the whole that machine learning can play regarding NFV issues. The authors highlight, for example, research opportunities by using network machine learning to failure prevention detection of degradation patterns in NFV enabling networks. And I can say that there are a lot of opportunities that you can explore by using machine learning to identify patterns and misbehaviors in an SFC or in a unique NFV, in a unique VNF, for example. In the other work published in the network and managed in 2018. Our work to identify quiz guilty in a SFC was proposed. This work proposes an adaptive algorithm based on linear regression and neural networks to adjust the model parameters according to environmental particulates. For example, you can use the type of VNF, the number of VNFs in a SFC to train, to teach the algorithm how it can detect which is the guilty for a bottleneck or misbehavior or a problem of performance in a SFC. So there are a lot of opportunities in such a direction. Another important point that we need to consider is management orchestration of VNFs or SFCs and the composition of SFCs. How can I create a service, a composition of VNFs to provide a service in a secure, adaptive way? So there are some works in academy that are trying to solve this problem. For example, we have in addition to the telecommunications magazine in 2017, a framework to facilitate the development of nephreatic components. If you focus on SFC, it was proposed as a demonstration. Also, we have a scenario, a user case in this journal. A different algorithm for VNF placement can be compared by using such a solution and also adapted to meet some SFC high-dynamic demands. So it is a very nice work to know about dynamic security composition. Other work, for example, is open-net VM. That was published in SIGCOM 2018. I can say that SIGCOM is one of the most important congress in computer networks around the world. The open-net VM, which is an open source in FV platform to accelerate research and development of software-based networks at all, was presented as a demo also. So it was implemented and the demonstration was provided. And basically open-net VM creates an abstraction layer over DPDK to allow the chain of VNFs running in the same host. I can say that the unique limitation, the unique dependence of this work is the work proposed that it's obligatory to use Docker. But its own dependence, it's not a limitation, of course not. It's only a dependence you need to use Docker. You can use Docker and create high-performance substance chains in FV. So other relevant issues regarding SFC construction is the validation of such SFCs that was created. For example, I mean, when I talk about validation, it's how can I validate that SFC is running as proposed or as promised and how to guarantee the integrity of the packets that are being forwarded between the SFC. For this, we have different kind of works. For example, we have SFC Path Tracer that enables the identification of problems in SFC and simplify the configuration by using visualizations to show the whole path traversed by packets in a given SFC. Such work was developed in the HP Enterprise in the south of Brazil. So I can say that it's a kind of tool from industry to validate the solutions that have been developed in HP, for example. Other works that helps during the validation configuration of SFCs was presented in Netsoft 2018. It is container net. The container net was proposed for support in the creation and localization to validate hybrid SFCs composed of both container-based and virtual machine-based VNF. So we can create, for example, a service, but IPA tables-based fire running in Docker and DPI running Ubuntu VM machine, for example. So it provides some intelligibility to create SFCs using different technologies and solutions. And other important thing about, that's a very general problem, is placement, migration, and resource allocation by considering SFCs or VNFs in dependence. There are a lot of problems trying to solve the problem of performance utilization of resource management. For example, we have Hyper-NFV that was proposed to minimize the lattice when running large numbers of VNFs in eight days at NFV infrastructure. It was published in 2017. And we have, for example, in October 2018, in the ETPOE communications ladders, we have our work that provides algorithm for the cloud and bandwidth for such a location in multi-providing environments that was proposed taking into account different infrastructure costs. These algorithms in this paper, in this work, provide some results that demonstrate that this work can save 65% more in terms of money to deploy a VNF in the infrastructure regarding the when compared with solutions that doesn't take account important aspects of NFV. So we can use, for example, an algorithm for cloud computing, of course, to research a location for VNFs. But it's not so efficient when we can use a lot of information provided by the environment to try to identify and to provide which is the best way to put a VNF, for example. So it's a very important question to do about resource management for specific demands and particular types of end users and VNF demands. And NFV has a strong economic appeal, of course. Cost and energy savings are one of the main goals to the next years. So we have different works trying to address energy savings and cost efficiencies for NFV. In this work, for example, the authors investigated how to add a VNF into an existing service chain. You wait to not impact negatively in the infrastructure costs. So if you have ISFC, for example, running the same host or infrastructure provider, and you need to put other VNF in this ISFC, how you can put this to not impact much in the costs or in the economic aspects of these infrastructure providers. Other works are focused mostly on the issues regarding energy efficiency. For example, in the International Conference on Communications in 2018, a work that joint optimization of VNF placement and traffic routing for energy efficiency in telecom networks were proposed. I can say that currently telecom networks is one of the most interested in energy savings by using NFV, because, of course, there are a lot of opportunities to make innovation and evolution in the network core that is directly related to telecom networks providers. Also, there are several works focusing NFV potential as enabled to 5G networks or 5G core. Therefore, academy and industry are interested in how to use NFV for supporting NFV, IoT, and edge computing as well. So in this way, there are efforts mainly to build fully-framed works for management and deployment of VNFs to sustain 5G networks. In general, I can say that these works are splitting networks elements into VNFs to address the demands of 5G era. For example, to provide ultra-low latency applications and dynamic resource allocation for 5G and IoT devices or applications. In the talk, for example, published in January past, the IEEE communications address, researchers discussed a realistic 5G scenario by considering NFV as a player that involves resource allocation and orchestration in a cloud-enabled radio access network, taking into account the dynamics of 100,000 of persons' movement in a cloud event. So, CCNFV and 5G networks are becoming a reality. I think a lot of new opportunities and challenges are imagined together. And one of these challenges is how to provide performance to sustain or to address the new applications to 5G here. In this year, for example, in a conference, our work to provide ultra-high-quality multimedia by creating NFV-based dynamic adaptive streaming system were proposed. We have other works trying to take some issues about edge computing. For example, these work researchers adapted NFV technology for mobile edge computing and allocated cloud computing features near the base stage of radio access networks, which resulted in an extremely fast service access to user equipment. And we have other opportunities about, opportunities no, challenges about business perspectives. I think business models should be improved in the next few years. There are a lot of resource in such direction. The current works are trying to propose business models based on options, fixed price and pay-as-you-go. Such models are inspired in well-known cloud models. However, I think that a lot of new information from VNFs and users' demands must be considered to explore an effective way to create business models and put different kind of players to explore the market of NFV. NFV as a service, for example, is a research topic right now. There are a lot of works trying to provide not only VNF as a service but also infrastructure as a service to support such VNFs. So infrastructure supply is a research topic right now. And distribution instantiation of VNFs as a service also is an emerging topic. We have, for example, two projects, Tinova project and Fendt project was released to basically to provide a place where end users can purchase a VNF available in a catalog, for example, and such a VNF can be instantiated in a few minutes remotely or if it's to be provided by the end user. I mean, the idea is that end users can buy VNFs in a similar way that currently smartphone users can buy applications in App Store or Google Play. I'm part of Fendt project. It's a project that is supported by Brazilian government. Our main intentions in Fendt project is to propose an ecosystem that simplifies not only the process of acquisition and instantiation but also the management of VNFs. For this we introduced a marketplace for VNFs where end users can, for example, buy a VNF and we can instantiate this VNF by using OpenStack attacker in an infrastructure provider by us as an infrastructure provider or directly in the infrastructure that client can provide the credentials for us, for example. So, okay, so taking account these research challenges discussed and research trends and assuming that most of them will be solved in the next few years, I can say that the market of VNFs is very optimistic for the future. Today we have up to five billion of dollars being spent in NFV at all for hardware, software, and so on. And based on different reports for NFV, in 2024 we will have up to 71 billion of dollars in NFV market at all. For example, the acquisition of VNFs, hardware to VNFs enables infrastructures and operations related to managed registration of such NFV environments will be one of the most part of such amount of money. So a consumer capital should be also invested also for staff training, for example, to integrate a model as today to network for NFV models. Today we have, for example, in this table we can see which is the sector that is most interested in NFV and invest more in NFV solutions. First of course we have IT sectors and telecom providers, telecom networks followed by bank and financial systems and healthcare. I can consider also that in the next years the government can be a good player for NFV because of the IoT evolution in smart series and a lot of opportunities to use NFV as enabled to sustainability communication between different sectors in smart series, for example. Today most of such values is spent with hardware, of course, to create NFV environments with software to management orchestration and not much is spent with services but I think in the future this amount of services can be improved because as NFV becoming a reality I think a lot of innovative services or VNFs will arise and a lot of money can be spent for developers that create innovative services can be paid to one operator to buy such VNF for a specific demand or for example rent this VNF pay as you go I don't know but I think there are a lot of opportunities to developers to create innovative services in future and be paid for that. So now we'll talk about some market opportunities regarding VNF as a service, if it's to supply as a service, of course and smart series and IoT applications. So first I believe that NFVs as a service or VNFs as a service is the feature. I can say for example a model where developers can develop innovative services for example I as a research can develop innovative services to deal with DDoS attack. I construct a service by using IP tables firewall, a DPI by using neural network models, I don't know and I can create and I can put that in a marketplace for example a catalog of VNFs and one operator that has a specific demand can contract I don't know who can buy or rent such VNF so there is a lot of opportunities to developers being incentivated to develop a new innovation services, innovative services but that's available to any user and user interested. And users of course can be one of the most benefited by that because can contract service on demand a network is under attack right now and I know that the guys from the research from the MAT develop a new solution that can be deployed as a VNF to solve my problem. Ok, I will contract a service for example for 30 minutes only to mitigate the attack so there are a lot of opportunities and when we talk about NFV and VNFs, we are talking about a few minutes, a few seconds to have such network function deployed and available for your network and steps of providers can explore such market by building profitable business models to offer, distribute and also manage such VNFs for example to orchestrate to detect some problems, behaviors and to ensure that which user will contract will be delivered to such a user so there are a lot of opportunities also and if a service supply is other interesting opportunity because the provision of service to host VNFs is not a trivial issue so as there will be a lot of VNFs available, eventually for example a tree cooperator will not have an NFV environment configured to deploy such VNFs thus an NFV infrastructure must be available as a service so we can have an infrastructure provider that make a complete NFV environment available to sustain or to host a VNF contract by the user it is very similar to platform as Amazon AWS and Microsoft Azure for example however by considering an NFV, an infrastructure should be able to deal with others metrics and aspects to meet particular demands of each user and for example an infrastructure provider can make available some specific monitoring solutions and no platforms to problem notification, specifically for each case or each demand of end users and so that last but not less important scenario that can be explored in a lucrative way is the area of smart cities I think an opportunity described before is valid for this scenario as well developer and service provider for example can provide customized solutions to supply different areas in the city and provide ultra-high ultra-low latency for example applications and innovative services for example developers can do solutions to ensure low latency and reliable infrastructures for vehicle networks healthcare for example we can make available a solution or infrastructure to sustain video surgeries with reduced costs so there are a lot of opportunities mainly for developers and of course for smart cities digital sites at all so for the future I can imagine that NFV will be fully operational in the next few years by integrating NFV and software defined networks SDN, we can provide for example ultra-low latency and dynamic application fastly to end user and service provider for the network core to create different kind of communication and innovative applications in the future hopefully of course cyber attacks will become more complex and sophisticated when we have a lot of network functions being virtualized and a lot of issues related to security will be required attention I think that NFV should be combined with other technologies to improve these capacities I can say for example the hype of blockchain is a good opportunity to take some NFV challenges regarding accounting and billing for example currently I'm working at UZ8 in an auction-based solution for NFV infrastructure supplying using blockchain smart contracts for example if a end user needs to contract a VNF and don't have to host VNF we can use a kind of auction-based solution by using blockchain to not require a centralized solution I think there are a lot of opportunities not only by using blockchain but other kind of technologies to be merged with NFV to solve some problems but it's still open and I can say that more solutions to simplify to simplify the management and orchestration of VNFs, NFV or SFC will arise because we will have a lot of number of VNFs and we need to manage and orchestrate such VNFs so I think artificial intelligence is a good opportunity and information visualization as we have network and computer management to provide highlights for end user and other cooperators identify easily and fast way some problems can be a good friend for NFV in future so that's it, if you need some see some of the reference I will make available the conference how the conference is here and if you have some questions we will have a few minutes we will see requirements for getting a job if you are talking about bandwidth intensive applications in that case what is it that is scaling those VNFs out I am sorry I am sorry let me see if you can understand the question when a developer presents a project that he has what are some examples of the NFV service that might be in practice right now coming up very soon like are those health care related or the health care related for example for example we have a customer-premise equipment for example a halter we can put the Halter the DNS server the network translator involved from the home environment and put for example this is a provider in the service provider so we can put it because a halter with other fun examples that have some implementation for example the same project it is one of the first the DNS server to provide I will say any questions regarding the NFV and the service I would probably ask some little questions I have a service provider who is providing me NFV a service I am kind of a so what can be the common use case for me that I have to be using that service for what purpose I have to be using their service but I would also like to ask someone else I will then really get on my own environment I can rely for example in the so this is a challenge we need to for example a head hat has a marketplace for the NFV and I can buy the NFV and send the source code of the NFV or the NFV as a service for example we need to provide some money for example this is a blockchain to rely on a contract to be had and the service provider will deliver me what contract and ensure it is a kind of we need to rely each other we know how it is a challenge it is difficult I know it is a question I would like to start from one point of view I am just a customer I go to the to create my VMs I have the VMs on there now when I talk about NFV as a service what do I have and how would I use it with my own environment my current existing VMs in my private and every cloud will it be able to connect with your NFV service I will need to provide the package from your NFV we can use VPNs for example but of course it can it is a challenge there are a lot of opportunities for example today we have but can definitely help people to offer them most of our scenarios is for example the same as FC in a unique infrastructure but in future for example we need to have a VMF being run in Brazil and we need to communicate how can I do it is a problem I can say for healthcare the main is to provide an infrastructure to supply healthcare communication or IoT at all so NFV will be used to virtualize some functions in the network core to provide for example I need to make a surgery right now it is a video surgery I don't know that will be live and I need to provide a kind of facility we can use NFV to supply surgery for healthcare specifically for example I can say that we can use NFV or as I am able to communicate or to create a more generic approach in which we can communicate for healthcare for example we need to utilize the healthcare is to provide an infrastructure healthcare can communicate between different devices NFV is for if you still supply it is not for healthcare but you can do some services specifically to help healthcare to provide some performance improvements for healthcare in the eyes it is a challenge to use the helpers if you take that at the time it is still free can you comment on the performance difference between the practice model and the NFV model what is the first model the first model when we use the hardware in this piece it is like hardware the performance specifically is not a problem because we can locate a lot of the source we can manage the source to add more source I think it is about the communication for example the most problem is about latency maybe for example if we have some NFV that is a compressor and we need to see the practice and deliver to a lot of NFV and use it because we have some problems with communication latency the performance will not be the problem between the future the problem is how to secure communication and not impact negative in the performance for example I can use a VPN yeah I can but it will be impact negative in the latency of the source to change if you have some other questions we can discuss thank you