Designed 40 years ago, DNS is still a core component of internet: billions of DNS queries are processed each day to resolve domain names to IP addresses. Originally designed for performances and scalability, its transport protocol is unen-crypted, leading to security flaws. Recently, secure protocols have emerged, but the question of their scalability and sustainability remains open. In this paper we study the cost of switching from the legacy DNS transport to the newer ones, by first characterising the shape of the traffic between clients and secured public resolvers. Then, we replicate said traffic, to measure the added cost of each protocol. We found that, while connections usually stayed open, many closures and openings were made in some cases. Comparing these profiles over different DNS transports, we observe that switching from the legacy protocol to a more secure one can lead to an important performance penalty.

The throughput supported by network interfaces can be hardly followed by the packet processing speed in software. The processing rate is bounded due to the overheads imposed by the architecture of the network stack. For this reason, multiple research proposals try to offload packet processing on different types of hardware like GPUs and FPGAs. In this paper we present an architecture which offloads the data plane packet processing on a programmable hardware with parallel processing capability. For this purpose, we use the MPPA (Massively Parallel Processor Array) smart NIC from Kalray which offers the ODP API that can be used for packet processing. Our goal is to build a full mesh non-blocking layer 2 network. We implemented a TRILL protocol on the MPPA processor, which can be used as a basis for fabric network, but some other protocol may be used as well. Our performance evaluation shows that we can process TRILL frames at full-duplex line-rate (up to 40Gbps) for different packet sizes while reducing latency.

TRILL is a Layer 2 protocol which allows the establishment of a layer 2 mesh network while improving infrastructure utilization in comparison with Ethernet. Indeed, TRILL does not use STP and keeps all links active. It always uses the shortest path between any two nodes and can dispatch the flows following ECMP. However, the problem is that throughput wise the performances of TRILL are not on par with Ethernet. This is the consequence of the fact that Linux does not have a fast path solution for TRILL and, moreover, the fact that CPU and memory resources have to be used in order to process the frames. The goal of this paper is to propose a fast path solution for TRILL that is based on a smart NIC. We experimented with our solution and the results show that we successfully increased the throughput of TRILL by 100%, thus reaching the network interface limit of 10Gbps while reducing latency.

Cloud computing is expected to keep growing and therefore more and more data centers are being built. In parallel of this increasing amount of data centers, there are also more cloud providers. On one side, this heterogeneity in the cloud market allows a user to choose and change its cloud provider. On the other side, thanks to this growing number of cloud providers, it is now possible to have small cloud providers forming a cloud federation. This way, they can mutualize their resources in order to accommodate their clients and compete with bigger cloud providers. However, solutions are needed to interconnect data centers. Those solutions must allow a cloud provider to interconnect its data centers to other cloud providers' one. This interconnection must be achieved even if the data centers do not use the same technology. This is the problem we focus on in this paper and we propose a solution to interconnect a TRILL data center with an OpenFlow one. We design and implement this solution in order to analyze its impact on both throughput and latency. Concerning the latency, we found that our solution increases it by 3% compared to the OpenFlow solution. Throughput-wise, we have an 8% lower throughput than TRILL.

Virtualization is a key technology to enable cloud computing. It enhances resource availability and offers high flexibility and cost effectiveness. However, the driver domain-based model for network I/O virtualization exhibit poor networking performance. In this paper, we evaluate the networking performance of VMs based on Xen. The I/O channel transferring packets between the driver domain and the VMs is shown to be the bottleneck. To overcome this limitation, we proposed a packet aggregation-based mechanism to transfer packets from the driver domain to the VM. Experimental performance evaluation showed that our proposal allows the VMs throughput to scale up at line rates. Furthermore, we studied the impact of the aggregation on the packets delay. For this purpose, we proposed a system modeling based on queuing theory. The proposed model allowed us to dynamically tune the aggregation mechanism to achieve the best tradeoff between the packets delay and throughput. The proposed I/O virtualization model henceforth satisfies the infrastructure providers to offer cloud computing services.

Nowadays, wireless networks with heterogeneous technologies are largely deployed and, mobile terminals have multiple interfaces to access these networks. This raises many research issues such as the simultaneous access to heterogeneous networks and the mobility management. This paper proposes to manage the mobility of a terminal in this context. In particular, we want to guarantee that, while a terminal is moving from a given network to a different one, the quality of service of its applications will be maintained. For this purpose, we combine a multi-homing protocol (SHIM6) with an IMS architecture. The protocol SHIM6 ensures a seamless change of networks without any application disruption. The IMS architecture allows the establishment of multimedia session with quality of service. Our solution can manage the inter-domain mobility of a terminal in a reactive and proactive mode. For this purpose, we implement a SIP proxy inside the mobile terminal. The experimental evaluation shows the feasibility of the SHIM6-IMS integration and a considerable reduction of the handover delay.

Cloud computing is an emergent paradigm that allows to customers to rent infrastructure, platforms and software as a service. With resource sharing and reuse through virtualization technology, cloud environments become even more effective and flexible. Nevertheless, networking within virtualized cloud still presents some challenges in performance and resource allocation. In this paper, we propose to integrate an SLA-based Dynamic Bandwidth Allocator (DBA) in a virtualized cloud environment. DBA manages bandwidth allocation efficiently through allocating bandwidth according to the application requirements and the established agreement. It also adjusts the allocated bandwidth dynamically upon change and reduces physical resources usage by dropping packets in the virtual machines rather than the driver domain. Through experimental evaluation we showed the efficacy of the proposed algorithm and the agreements respect.

The context of our work is mobile multi-homed terminals in a heterogeneous environment. We propose a solution for seamless terminal mobility in an intra-domain context. We combine a multi-homing protocol (SHIM6), which ensures seamless network change, with the IMS architecture, which allows the establishment of multimedia session with Quality of Service. Our proposal allows a mobile terminal to change its access network seamlessly, without any application disruption. To achieve this, we implement a Proxy-SIP inside the terminal to manage session establishment. The proposed scheme can manage access network change in two modes: reactive and proactive. A testbed allowed us to show the feasibility of our approach and the gain obtained by considerably reducing the handover delay and, thus, data loss.

Cloud computing is an emergent paradigm that allows customers to rent infrastructure, platforms and software as a service. With resource sharing and reuse through virtualization technology, cloud environments become even more cost effective and flexible. Nevertheless, networking within virtualized cloud still presents some challenges in performance and resource allocation. In this paper, we propose DBA-VM, a Dynamic Bandwidth Allocator for Virtual Machines with regard to the established SLAs. The proposed scheme enforces the isolation between the virtual machines through the transmission bandwidth adjustment at the network I/O channel. The experimental performance evaluation shows that DBA-VM allows to the virtualized system to respect each virtual machine SLA while reducing the global physical resources (CPU and memory) consumption.

Network virtualization is a promising technology that offers high levels of flexibility, isolation, extensibility and cost-effectiveness. In this paper, we focus on router virtualization. We evaluate the forwarding performance of virtual routers when the data plane runs in the guests. This scenario offers a high level of isolation and flexibility, however, it suffers from performance limitations due to the virtualization overhead. We show that the he I/O communication between the driver domain and the guests is the bottleneck. To overcome this limitation, we propose a new packets aggregation mechanism that transfers groups of packets between the driver domain and the guests. This enhancement makes the forwarding performance of one guest scale up to 1600 Kp/s. Furthermore, we propose a dimensioning tool in order to determine the maximum achievable throughput with regard to the container size.

Router virtualization seems as the obvious next step to system virtualization and the key to easily deploy and manage next generation overlay virtual networks. In this paper, we investigate the viability of virtual routers on a Xen-based system. We first evaluate the system throughput when achieving forwarding in the virtual routers. Then, we consider the context where virtual routers are dedicated to flows of different types and propose a mechanism to guarantee the required throughput and latency to real time applications while maintaining an optimal aggregated system throughput. We achieved this through both configuring the Xen Credit scheduler and establishing priorities between packets in the driver domain before switching them to the target virtual router.

Virtual routers on commodity hardware are an attractive solution for service providers that look for extensibility, flexibility, reuse and low deployment cost. However, these routers still suffer from performance limitations due to the virtualization overhead and the commodity hardware architecture itself. In this paper, we first evaluate the baseline forwarding performance of virtual routers based on a Xen environment. Then, we show that the memory latency is the bottleneck. Hence, we propose a Fast Bridge that demultiplexes incoming packets and then transfers them to the destined guest machines. It constructs packets containers in the driver domain, based on their destination and their delay constraints. Then it transfers them as a unit to the guests. This allows transferring more packets and fastest memory access and results in a much better throughput with an acceptable guaranteed delay.

The increasing popularity of WLANs-due to the use of license-free radio spectrum with low-cost, easily deployable and high-data-rate wireless services- has encouraged service providers to consider deploying them in high density usage areas, such as public hotspots, to provide complementary broadband access to their networks and services. This article proposes a Session Initiation Protocol based QoS management architecture that aims to provide consistent QoS control for multimedia applications (VoIP, VoD, ...) over wireless access networks. In particular, we will present a generic QoS based management architecture for decentralized management of multimedia applications over wireless access networks, such as WLAN. We will apply our architecture to the management of the WLAN in order to assure QoS (throughput and delay) for the delivered multimedia applications according to users' priorities and providers' objectives. Performance evaluations are discussed to illustrate the feasibility and the efficiency of the proposed architecture through the implementation of a real-world testbed.

In this application paper, we investigate the introduction of the autonomic technology into usual centralized management systems. First, we present the global architecture of the proposed hybrid management system. Then, we describe the core entity of this architecture which is the autonomic agent. Finally, we develop a demonstrator and we define a relevant testing scenario in order to assess the feasibility and the effectiveness of our proposals. This work achieves the first steps towards an autonomic management system.

In this paper we present a complete analysis of dynamic resource allocation in policy-enabled networks. This analysis is carried out throughout both practical and analytical evaluations. Firstly, we present the details of our dynamic resource allocation architecture based on the policy-based management framework. Then, we evaluate, through extensive experimentations on a developed test-bed, the scalability of such architecture in a real environment. Finally, we develop an analytical model for the proposed architecture. This analytical evaluation allows us to confirm our practical analysis and to identify the weakness of such architectures