Personalized and Seamless Access to Wi-Fi Services
In this contribution, we present a solution to provide a personalized and seamless access to venue-based services offered through public Wi-Fi networks. Our objective is to improve Wi-Fi user experience and deliver to him the services he desires, where and when he desires. To do so, we propose an extension to IEEE 802.11 management frames to enable venue service discovery prior to Wi-Fi association while avoiding channel overhead. We also define a set of extensible service labels to uniquely and globally identify the most known venue-based services. These labels can be used in the proposed extension to identify advertised services in the Wi-Fi networks. Through deep analysis and comparison to existing solutions, we show how these two proposals could be efficient and useful and help to have a personalized, transparent and automated access to Wi-Fi venue-based services based on user preferences and context. This contribution is the object of a publication  and a proposal for an IETF draft .
NFV- and Cloudlet-based Carrier Wi-Fi Architecture
In this contribution, we propose a novel architecture for carrier-managed Wi-Fi networks that lever-ages Network Function Virtualization and Edge Cloud Computing concepts. We aim through this architecture to bring more flexibility and adaptability and allow operators to easily implement new services while reducing CapEx and OpEx costs. We also aim to decrease access latency by placing network functions and certain services close to end-users. To achieve these objectives, we intro-duce a new architecture element, called WLAN Cloudlet located in the end-user premises, that offloads MAC layer processing from APs and consolidates network functions and value-added ser-vices. All these functions and services are based on software instances. To prove the feasibility and evaluate the performance of our proposal, we develop a proof-of-concept prototype and compare it with a reference architecture. Results show that the WLAN Cloudlet solution achieves good perfor-mances, while providing at the same time many advantages in terms of cost, flexibility, and agility. This contribution is the object of publication  and .
User Experience and Service Discovery in Wi-Fi Networks: State of the Art
In this section, we first describe the most recent standards which aim to improve the Wi-Fi user experience (section 2.2.1). Second, we outline the different service discovery approaches according to which we classify the existing protocols and solutions (section 2.2.2). Finally, we summarize the reviewed solutions (section 2.2.3).
Hereafter, we describe the most recent and relevant standards related to the user experience im-provement in Wi-Fi networks, namely i) IEEE 802.11u and Hotspot 2.0 and ii) Wi-Fi Aware.
IEEE 802.11u and Hotspot 2.0 (Network Discovery)
Recent standards and technologies, such as IEEE 802.11u  and Hotspot 2.0 (Wi-Fi Alliance) [3, 4], have been developed to deliver a seamless Wi-Fi access and to enable automated network discovery and selection without any active intervention from the user. A key innovation of these technologies is to provide information to a mobile device about a Wi-Fi network before deciding to join it. Particularly, this helps to assist the mobile device in selecting a network.
For this purpose, 802.11u focuses on enhancing network discovery by adding new beacon and probe response information elements (e.g., Extended capabilities, Interworking, and Roaming Con-sortium). In addition, it introduces a new pre-association protocol, called Access Network Query Protocol (ANQP), used by Wi-Fi client devices to query the hotspot for additional network in-formation (e.g., Venue Name and Network Authentication type) that is not advertised in beacon and probe response frames. These queries are transported using Generic Advertisement Service (GAS), an IEEE 802.11 service that provides over-the-air transportation for frames of higher layer advertisements. The Wi-Fi Alliance has extended ANQP protocol with its own Hotspot 2.0 ANQP elements (e.g., Operator Friendly Name, WAN Metrics and Connection Capability) to provide fur-ther querying functionality.
All this information carried by beacon frames and ANQP protocol is related to the hotspot’s capabilities but it does not include information about the services that are locally reachable via the hotspot. Indeed, these technologies enable only mechanisms of pre-association for network discovery and selection and do not support service discovery. Thus, the network selection could not be based on available services in the Wi-Fi network.
Wi-Fi Aware (Peer-to-Peer Service Discovery)
Wi-Fi Aware  is a new capability launched by the Wi-Fi Alliance in 2015 which enables proximity-based service discovery before making a connection. It is an always-on technology that helps to find nearby information and services without a connection to a wireless AP. Then, it ini-tiates interactions between devices and people. Wi-Fi Aware is a key enabler of an interactive and personalized mobile experience, enabling users to find near video gaming players, share media content, and get contextual notifications and offers according to their preferences.
This technology is based on the Wi-Fi Alliance Neighbor Awareness Networking (NAN) Tech-nical Specification . It enables a continuous device-to-device discovery using NAN Discovery Beacon, a modified version of the IEEE 802.11 beacon management frame. When a device discov-ers an interesting service, the device then initiates a Wi-Fi connection.
Wi-Fi Aware is a promising “neighbor awareness” technology. However, it is based on a peer-to-peer connectivity and does not provide discovery and access to local services offered through public Wi-Fi networks.
Service Discovery Approaches
Service Discovery (SD) is a process enabling dynamic discovery of available services in the net-work and automatic configuration of devices. It provides necessary information about available services and help users and applications to access to network resources such as devices, data and
services. Many Service Discovery Protocols (SDPs) were developed in research and industrial com-munities. These protocols are designed to minimize human intervention and administrative over-head and improve user experience. We classify these protocols into two major categories. The first category regroups protocols used after the device is associated and connected to the network, called post-association protocols. The second category is pre-association protocols that enable devices to discover services prior to associate to the network. These protocols are specifically used in wireless networks, particularly Wi-Fi.
We review, in the following, the most relevant existing solutions corresponding to each category.
Post-association Service Discovery
With the aim to enable devices to join the network and dynamically discover and access the needed services with zero-configuration, a number of service discovery protocols and architectures was developed. Examples include Jini , Universal Plug and Play (UPnP) , Service Location Protocol (SLP) , Salutation , and Bonjour . These protocols are generally used in wired networks but some of them can also be used in wireless networks (e.g., UPnP and Bonjour). Each one of these protocols is designed to address a specific set of issues.
Jini  has been developed by Sun Microsystems and is a Java based technology. It provides a software platform enabling service discovery and invocation among java enabled devices. In a Jini environment, users are able to share resources and services in a network and to easily access to available services. To ensure the advertisement and discovery of services, Jini uses a set of Discovery protocols, namely multicast request protocol, multicast announcement protocol, and unicast discovery protocol . Jini specification is independent of the network protocol, but the most current implementations are based on TCP and UDP.
UPnP  is a Microsoft’s peer-to-peer networking initiative. It enables service advertise-ment and discovery in small home and corporate environments. It provides peer-to-peer con-nectivity between appliances, services and wired/wireless devices. It was introduced as an extension to the plug-and-play peripheral model to support discovery and configuration of devices throughout the network. UPnP uses Simple Service Discovery Protocol (SSDP)  for service discovery. This protocol is used for discovering devices or services and announc-ing the presence of a device or availability of services in the network. To do so, it uses HTTP over unicast and multicast UDP packets. UPnP is independent of operating systems, programming languages and physical media but is designed for only TCP/IP networks.
Salutation  is another approach for service discovery. The Salutation architecture was developed by the Salutation Consortium. The aim of this architecture is to address the prob-lem of discovery and determining the capabilities of heterogeneous information appliances that can be encountered in a networked environment. It is independent of operating systems, physical platforms and communication protocols.
SLP  is an IETF standard that provides a scalable and flexible framework for service discovery on IP networks. SLP can be deployed in small networks, like home networks, without any specific configuration, as well as it can scale well in large networks with prede-fined policies. SLP advertises services through a service URL, which contains all information necessary to connect to a service. The protocol has the advantage of not depending on any programming language or communication protocol.
Bonjour is a technology developed by Apple to provide service and device discovery among computers, electronic appliances and other networked devices (e.g., printers, faxes, etc.) over IP networks. It is based on a service discovery protocol, called DNS-based Service Discovery (DNS-SD) .
Centralized WLAN Architecture
As the need for centralized monitoring and dynamic configurability grew, vendors introduced controller-based systems with “thin” low-cost APs. In this architecture, the controller is respon-sible for controlling, configuring, and managing the entire WLAN access network. Furthermore, all the traffic is routed from the APs to the controller . According to RFC 4118 , central-ized WLAN architectures are categorized into three main variants: i) the Local MAC in which the MAC functions stay intact and local to APs, ii) the Remote MAC in which the MAC has moved away from the AP to a remote Access Controller (AC) in the network, and iii) the Split MAC in which the MAC is split between the APs and the ACs. The centralized WLAN architecture is char-acterized by the ease of deployment especially for wide networks and it provides more security and control. However, it has two major drawbacks. Firstly, the WLAN controller represents a single point of failure. Secondly, since all transmissions require passing through the controller, the latter became a major bottleneck, thus eroding network performance.
Distributed WLAN Architecture
To resolve the issues mentioned above, the third generation of WLAN, called distributed WLAN architecture, introduced distributed data forwarding . A controller still provides a central point of control for APs, however, all traffic is no longer backhauled to the controller. This solution eliminates network bottleneck but it is much more expensive than the other solutions.
Virtualized WLAN Architecture
Recently, a new trend has emerged in the world of WLAN by introducing virtualized WLAN ar-chitecture which has rapidly been gaining the attention of industry and academia. Some vendors offer the controller as a virtual appliance and even as a cloud-based hosted service [35–37]. The cloud-based controller centrally manages and monitors APs and user data is not going through the controller. This solution has several advantages in terms of ease of deployment and availability but it exhibits a number of shortcomings. Note that even if this solution eliminates the need of deploy-ing on-premises hardware WLAN controllers, this does not imply cost reduction. In fact, controller functionality is distributed between the APs and the cloud. Thus, the cost of the controller is in-tegrated into the price of the APs and the Cloud controller subscription that must be continuously renewed. Over the long haul, this solution is much more expensive than simply purchasing a WLAN controller from the beginning. Moreover, for security reasons, many organizations still require on-premises WLAN controller to easily manage network traffic and there may be regulatory needs to tunnel traffic to the controller. Finally, this solution is only suitable for low- to medium-density locations as APs are not designed to support high-scale tunnel traffic. In contrast, centralized con-trollers have a dedicated hardware that makes them extremely efficient at moving traffic through the network.
Table of contents :
1 General Introduction
1.1 Evolution of Wi-Fi Networks towards 5G
1.1.1 Wi-Fi Offload and Roaming
1.1.2 Voice over Wi-Fi
1.1.3 5G-oriented Technologies
1.2 Carrier-grade Wi-Fi Networks
1.3 Problem Statement
1.4.1 Personalized and Seamless Access to Wi-Fi Services
1.4.2 NFV- and Cloudlet-based Carrier Wi-Fi Architecture
1.4.3 VNF Placement and Provisioning in Carrier Wi-Fi Network
1.5 Thesis Outline
2 Service Discovery and Access in CarrierWi-Fi Networks
2.2 User Experience and Service Discovery in Wi-Fi Networks: State of the Art
2.2.1 Existing Standards
184.108.40.206 IEEE 802.11u and Hotspot 2.0 (Network Discovery)
220.127.116.11 Wi-Fi Aware (Peer-to-Peer Service Discovery)
2.2.2 Service Discovery Approaches
18.104.22.168 Post-association Service Discovery
22.214.171.124 Pre-association Service Discovery
2.3 Personalized and Seamless Access to Wi-Fi Services
2.3.1 Pre-association Discovery of Local Services
2.3.2 Unique and Global Service Identifiers
2.4.1 The Client Perspective
126.96.36.199 Link Setup Time
188.8.131.52 Power Consumption
184.108.40.206 User Experience and Satisfaction
2.4.2 The Network Operator Perspective
220.127.116.11 Bandwidth Usage
18.104.22.168 Ease of Deployment
2.5 Use Case Examples
3 Next-Generation CarrierWi-Fi Architecture
3.2 Background and State of the Art
3.2.1 Evolution of WLAN Architectures
22.214.171.124 Autonomous WLAN Architecture
126.96.36.199 Centralized WLAN Architecture
188.8.131.52 Distributed WLAN Architecture
184.108.40.206 Virtualized WLAN Architecture
3.2.2 Emerging Concepts for Future Wireless Networks
220.127.116.11 Network Function Virtualization
18.104.22.168 Emerging Cloud Computing Models
3.3 NFV- and Cloudlet-based Carrier Wi-Fi Architecture
3.3.1 System Description
22.214.171.124 WLAN Cloudlet
126.96.36.199 Wireless Termination Points
188.8.131.52 Cloud-Based Platform
3.3.2 Benefits and Possible Applications
184.108.40.206 Possible Application Scenarios
3.4 Feasibility and Implementation
3.4.1 Case Study
3.4.2 Implementation Aspects
3.4.3 Performance Evaluation
4 Service Management in NFV-oriented CarrierWi-Fi architecture
4.2 Virtual Machine Placement in Virtualized Environments: State of the Art
4.2.1 Analysis of Existing VMP Approaches
220.127.116.11 Optimization Problem Formulation
18.104.22.168 Computing optimized VM placement
4.2.2 Related Literature Review
22.214.171.124 VM Placement across Geographically Distributed Clouds .
126.96.36.199 VM Placement in Hybrid Clouds
188.8.131.52 VNF Placement
4.3 QoS-driven VNF Placement and Provisioning in Edge-Central Carrier Cloud Architecture
4.3.1 System Modeling
184.108.40.206 Performance Model
220.127.116.11 QoS Model
4.3.2 Problem Description and Formulation
4.3.3 Solutions Description
18.104.22.168 Trade-off between Cloudlet Utilization and QoS Violation (To-CUQV)
22.214.171.124 Fixed QoS Violation Threshold (FQVT)
126.96.36.199 Fixed Maximum Cloudlet Utilization level (FMCU)
4.4 Performance Evaluation
4.4.1 Simulation Settings
4.4.2 The Baseline Approach
4.4.3 Performance Metrics
4.4.4 Simulation Results
5 General Conclusion
5.1 Summary of Contributions
5.2 Future Work
5.4 WBA Projects
List of Figures
List of Tables