Section: Application Domains
The next-generation network must overcome the limitations of existing networks and allow adding new capabilities and services. Future networks should be available anytime and anywhere, be accessible from any communication device, require little or no management overhead, be resilient to failures, malicious attacks and natural disasters, and be trustworthy for all types of communication traffic. Studies should therefore address a balance of theoretical and experimental research that expand the understanding of large, complex, heterogeneous networks, design of access and core networks based on emerging wireless and optical technologies, and continue the evolution of Internet. On the other hand, it is also highly important to design a next-generation Internet which we will call the "Future Internet" from core functionalities in order to ensure security and robustness, manageability, utility and social need, new computing paradigms, integration of new network technologies and higher-level service architectures.
To meet emerging requirements for the Internet's technical architecture, the protocols and structures that guide its operation require coordinated, coherent redesign. A new approach will require rethinking of the network functions and addressing a range of challenges. These challenges include, but are not limited to, the following examples:
New models for efficient data dissemination;
Coping with intermittent connectivity;
The design of secured, privacy protecting, and robust networked systems;
Understanding the Internet behavior;
Building network evaluation platforms.
The following research directions are essential building blocks we contribut to the future internet architecture.
Data centric Networking
From the Internet design, back to 1970, the resources to be addressed and localized are computers. Indeed, at that time there were few machines interconnected, and nobody believed this number the ever be larger that a few tens of thousand of machines. Moreover, those machines where static machines with well identified resources (e.g., a given hierarchy of files) that were explicitly requested by the users. Today, the legacy of this architecture is the notion of URLs that explicitly address specific resources on a specific machine. Even if modern architectures use caches to replicate contents with DNS redirection to make those caches transparent to the end-users, this solution is only an hack that do not solve the today real problem: Users are only interested in data and do not want anymore to explicitly address where those data are. Finding data should be a service offered by the network. In this context of data-centric network, which means that the network architecture is explicitly built to transparently support the notion of content, a data can be much more than a simple content. In such a network you can, of course, request a specific file without specifying explicitly its location, the network will transparently return with closest instance of the content. You can also request a specific service to a person without knowing its explicit network location. This is in particular the case of a VoIP or an instant messaging conversation. A data-centric architecture is much more than a simple modification in the naming scheme currently used in the Internet. It requires a major rethinking a many fundamental building blocks of the current Internet. Such networking architecture will however allow seamless handling of the tricky problematic of episodic connectivity . It also shifts the focus from transmitting data by geographic location, to disseminating it via named content. In the Planète project-team, we start to work on such data centric architectures as a follow-up and federating axe for three of our current activities (adaptive multimedia transmission protocols for heterogeneous networks, data dissemination paradigms and peer-to-peer systems). It is important to study such data centric architectures considering in particular the corresponding naming problem, routing and resource allocation, reliable transport, data security and authentication, content storage.
Today's Internet is characterized by high node and link heterogeneity. Nodes may vary substantially in terms of their processing, storage, communication, and energy capabilities. They may also exhibit very different mobility characteristics, from static nodes to nodes that are considerably mobile (e.g., vehicles). Links may be wired or wireless and thus operate at widely varying rates and exhibit quite different reliability characteristics. One of the challenges of data-centric architecture is to provide access to data anytime anywhere in the presence of high degree of heterogeneity. This means that due to a number of factors such as node mobility, link instability, power-aware protocols that, for example, turn nodes off periodically, etc., the network will not be connected all the time. Additionally, disconnections may last longer than what “traditional” routing protocols (e.g., MANET routing) can handle. These types of network, a.k.a, intermittently connected networks, or even episodically connected networks have recently received considerable attention from the networking research community. Several new routing paradigms have been proposed to handle possibly frequent, long-lived disconnections. However, a number of challenges remain, including: (1) The support of scalable and transparent integration with “traditional” routing mechanisms including wired infrastructure, infrastructure-based wireless and MANET routing. (2) The study of heuristics for selecting forwarding nodes (e.g., based on node's characteristics such as node's speed, node's resources, sociability level, node's historic, etc. (3) The design of unicast and multicast transmission algorithms with congestion and error control algorithms tailored for episodically connected networks and taking into account the intrinsic characteristics of flows. (4) The design of incentive-based mechanisms to ensure that nodes forward packets while preventing or limiting impact of possible misbehaving nodes. The solutions proposed, which are likely to extensively use cross-layer mechanisms, will be evaluated using the methodology and the tools elaborated in our new Experimental Platform research direction.
On the other hand, multicast/broadcast content delivery systems are playing an increasingly important role in data-centric networking. Indeed, this is an optimal dissemination technology, that enables the creation of new commercial services, like IPTV over the Internet, satellite-based digital radio and multimedia transmission to vehicles, electronic service guide (ESG) and multimedia content distribution on DVB-H/SH networks. This is also an efficient way to share information in WiFi, WiMax, sensor networks, or mobile ad hoc infrastructures. Our goal here is to take advantage of our strong background in the domain to design an efficient, robust (in particular in case of tough environments) and secure (since we believe that security considerations will play an increasing importance) broadcasting system . We address this problem focusing on the following activities: (1) The protocols and applications that enable the high level control of broadcasting sessions (like the FLUTE/ALC sessions) are currently missing. The goal is to enable the content provider to securely control the underlying broadcasting sessions, to be able to launch new sessions if need be, or prematurely stop an existing session and to have feedback and statistics on the past/current deliveries. (2) The AL-FEC building block remains the cornerstone on which the whole broadcasting system relies. The goal is to design and evaluate new codes, capable of producing a large amount of redundancy (thereby approaching rateless codes), over very large objects, while requiring a small amount of memory/processing in order to be used on lightweight embedded systems and terminals. (3) The security building blocks and protocols that aim at providing content level security, protocol level security, and network level security must be natively and seamlessly integrated. This is also true of the associated protocols that enable the initialization of the elementary building blocks (e.g. in order to exchange security parameters and keys). Many components already exist. The goal here is to identify them, know how to optimally use them, and to design/adapt the missing components, if any. (4) It is important that these broadcasting systems be seamlessly integrated to the Internet, so that users be able to benefit from the service, no matter where and how he is attached to the network. More precisely we will study the potential impacts of a merge of the broadcasting networks and the Internet, and how to address them. For instance there is a major discrepancy when considering flow control aspects, since broadcasting network are using a constant bit rate approach while the Internet is congestion controlled.
When a native broadcasting service is not enabled by the network, data should still be able to be disseminated to a large population in a scalable way. A peer-to-peer architecture support such an efficient data dissemination. We have gained a fundamental understanding of the key algorithms of BitTorrent on the Internet. We plan to continue this work in two directions. First, we want to study how a peer-to-peer architecture can be natively supported by the network. Indeed, the client-server architecture is not robust to increase in load. The consequence is that when a site becomes suddenly popular, it usually becomes unreachable. The peer-to-peer architecture is robust to increase in load. However, a native support in the network of this architecture is a hard problem as it has implications on many components of the network (naming, addressing, transport, localization, etc.). Second, we want to evaluate the impact of wireless and mobile infrastructures on peer-to-peer protocols. This work has started with the European project Expeshare. The wireless medium and the mobility of nodes completely change the properties of peer-to-peer protocols. The dynamics becomes even more complex as it is a function of the environment and of the relative position of peers.
The Internet was not designed to operate in an completely open and hostile environment. It was designed by researchers that trust each other and security was not an issue. The situation is quite different today and the Internet community has drastically expanded. The Internet is now composed of more than 300 millions computers worldwide and the trust relationship has disappeared. One of the reason of the Internet success is that it provides ubiquitous inter-connectivity. This is also one of the its main weakness since it allows to launch attacks and to exploit vulnerabilities in a large-scale basis. The Internet is vulnerable to many different attacks, for example, distributed Denial-of Service (DDoS) attacks, epidemic attacks (Virus/Worm), spam/phishing and intrusions attacks. The Internet is not only insecure but it also infringes users' privacy. Those breaches are due to the Internet protocols but also to new applications that are being deployed (VoIP, RFID,...). A lot of research is required to improve the Internet security and privacy. For example, more research work is required to understand, model, quantify and hopefully eliminate (or at least mitigate) existing attacks. Furthermore, more and more small devices (RFIDs or sensors) are being connected to the Internet. Current security/cryptographic solutions are too expensive and current trust models are not appropriate. New protocols and solutions are required : security and privacy must be considered in the Internet architecture as an essential component. The whole Internet architecture must be reconsidered with security and privacy in mind. Our current activities in this domain on security in wireless, ad hoc and sensor networks, mainly the design of new key exchange protocols and of secured routing protocols. We work also on location privacy techniques and authentication cryptographic protocols and opportunistic encryption. We plan to continue our research on wireless security, and more specifically on WSN and RFID security focusing on the design of real and deployable systems. We started a new research topic on the security of the Next-Generation Internet. The important goal of this new task is to rethink about the architecture of the Internet with security as a major design requirement, instead of an after-thought.
A lot of work has been done in the area of WSN security in the last years, but we believe that this is still the beginning and a lot of research challenges need to be solved. On the one hand it is widely believed that the sensor networks carry a great promise: Ubiquitous sensor networks will allow us to interface the physical environment with communication networks and the information infrastructure, and the potential benefits of such interfaces to society are enormous, possibly comparable in scale to the benefits created by the Internet. On the other hand, as with the advent of the Internet, there is an important associated risk and concern: How to make sensor network applications resilient and survivable under hostile attacks? We believe that the unique technical constraints and application scenarios of sensor networks call for new security techniques and protocols that operate above the link level and provide security for the sensor network application as a whole. Although this represents a huge challenge, addressing it successfully will result in a very high pay-off, since targeted security mechanisms can make sensor network operation far more reliable and thus more useful. This is the crux of our work. Our goal here is to design new security protocols and algorithms for constrained devices and to theoretically prove their soundness and security. Furthermore, to complement the fundamental exploration of cryptographic and security mechanisms, we will simulate and evaluate these mechanisms experimentally.
As already mentioned, the ubiquitous use of RFID tags and the development of what has become termed "the Internet of things" will lead to a variety of security threats, many of which are quite unique to RFID deployment. Already industry, government, and citizens are aware of some of the successes and some of the limitations or threats of RFID tags, and there is a great need for researchers and technology developers to take up some of daunting challenges that threaten to undermine the commercial viability of RFID tags on the one hand, or to the rights and expectations of users on the other. We will focus here on two important issues in the use of RFID tags: (1) Device Authentication : allows us to answer several questions such as: Is the tag legitimate? Is the reader a tag interacts with legitimate? (2) Privacy : is the feature through which information pertaining to a tag's identity and behavior is protected from disclosure by unauthorized parties or by unauthorized means by legitimate parties such as readers. In a public library, for example, the information openly communicated by a tagged book could include its title or author. This may be unacceptable to some readers. Alternatively, RFID- protected pharmaceutical products might reveal a person's pathology. Turning to authenticity, if the RFID tag on a batch of medicines is not legitimate, then the drugs could be counterfeit and dangerous. Authentication and privacy are concepts that are relevant to both suppliers and consumers. Indeed, it is arguable that an RFID deployment can only be successful if all parties are satisfied that the integrity between seller and buyer respects the twin demands of authentication and privacy. Our main goal here, therefore, is to propose and to prototype the design of cryptographic algorithms and secure protocols for RFID deployment. These algorithms and protocols may be used individually or in combination, and we anticipate that they will aid in providing authentication or privacy. One particular feature of the research in the RFID-AP project is that the work must be practical. Many academic proposals can be deeply flawed in practice since too little attention has been paid to the realities of implementation and deployment. This activity will therefore be notable for the way theoretical work will be closely intertwined with the task of development and deployment. The challenges to be addressed in the project are considerable. In particular there are demanding physical limits that apply to the algorithms and protocols that can be implemented on the cheapest RFID tags. While there often exist contemporary security solutions to issues such as authentication and privacy, in an RFID-based deployment they are not technically viable. And while one could consider increasing the technical capability of an RFID-tag to achieve a better range of solutions, the solution is not economically viable.
The current Internet has reached its limits; a number of research groups around the world are already working on future Internet architectures. The new Internet should have built-in security measures and support for wireless communication devices, among other things. A new network design is needed to overcome unwanted traffic, malware, viruses, identity theft and other threats plaguing today's Internet infrastructure and end hosts. This new design should also enforce a good balance between privacy and accountability. Several proposals in the area have been made so far, and we expect many more to appear in the near future. Some mechanisms to mitigate the effects of security attacks exist today. However, they are far from perfect and it is a very open question how they will behave on the future Internet. Cyber criminals are very creative and new attacks (e.g. VoIP spam, SPIT) appear regularly. Furthermore, the expectation is that cyber criminals will move into new technologies as they appear, since they offer new attack opportunities, where existing countermeasures may be rendered useless. The ultimate goal of this research activity is to contribute to the work on new Internet architecture that is more resistant to today's and future security attacks. This goal is very challenging, since some of future attacks are unpredictable. We are analyzing some of the established and some of the new architectural proposals, attempting to identify architectural elements and patterns that repeat from one architectural approach to another, leading to understanding how they impact the unwanted traffic issue and other security issues. Some of the more prominent elements are rather easy to identify and understand, such as routing, forwarding, end-to-end security, etc. Others may well be much harder to identify, such as those related to data-oriented networking, e.g., caching. The motivation for this work is that the clean slate architectures provide a unique opportunity to provide built in security capabilities that would enable the prevention of phenomenon like unwanted traffic. New architectures will most likely introduce additional name-spaces for the different fundamental objects in the network and in particular for routing objects. These names will be the fundamental elements that will be used by the new routing architectures and security must be a key consideration when evaluating the features offered by these new name-spaces.
The Planète project-team contributes to the area of network monitoring. In addition to the work on extensions for what we have already proposed, our focus is now on the monitoring of the Internet for the purpose of problem detection and troubleshooting. Indeed, in the absence of an advanced management and control plan in the Internet, and given the simplicity of the service provided by the core of the network and the increase in its heterogeneity, it is nowadays common that users experience a service degradation. This can be in the form of a pure disconnectivity or a decrease in the bandwidth or an increase in the delay or loss rate of packets. Service degradation can be caused by protocol anomalies, an attack, an increase in the load, or simply a problem at the source or destination machines. Actually, it is not easy to diagnose the reasons for service degradation. Basic tools exist as ping and trace-route, but they are unable to provide detailed answers on the source of the problem nor on its location. From operator point of view, the situation is not better since an operator has only access to its own network and can hardly translate local information into end-to-end measurements. The increase in the complexity of networks as is the case of wireless mesh networks will not ease the life of users and operators. The purpose of our work in this direction will be to study to which extent one can troubleshoot the current Internet either with end-to-end solutions or core network solutions. Our aim is to propose an architecture that allows end-users by collaborating together to infer the reasons for service degradation. This architecture can be purely end-to-end or can rely on some information from the core of the network as BGP routing information. We will build on this study to understand the limitations in the current Internet architecture and propose modifications that will ease the troubleshooting and make it more efficient in future network architectures. We are investigating a solution based on a two-layer signaling protocol a la ICMP in which edge routers are probed on end-to-end basis to collect local information on what is going on inside each network along the path. The proposed architecture will be the subject of validation over large scale experimental platforms as PlanetLab and OneLab.
Network Evaluation Platforms
The Internet is relatively resistant to fundamental change
(differentiated services, IP multicast, and secure routing protocols have not seen wide-scale deployment).
A major impediment to deploying these services is the need for coordination: an Internet service provider (ISP) that deploys the service garners little benefit until other domains follow suit. Researchers are also under pressure to justify their work in the context of a federated network by explaining how new protocols could be deployed one network at a time, but emphasizing incremental deployability does not necessarily lead to the best architecture. In fact, focusing on incremental deployment may lead to solutions where each step along the path makes sense, but the end state is wrong. The substantive improvements to the Internet architecture may require fundamental change that is not incrementally deployable.
Network virtualisation has been proposed to support realistic large scale shared experimental facilities such as PlanetLab and GENI. We are working on this topic in the context of the European OneLab project.
Testing on PlanetLab has become a nearly obligatory step for an empirical research paper on a new network application or protocol to be accepted into a major networking conference or by the most prestigious networking journals. If one wishes to test a new video streaming application, or a new peer-to-peer routing overlay, or a new active measurement system for geo-location of internet hosts, hundreds of PlanetLab nodes are available for this purpose. PlanetLab gives the researcher login access to systems scattered throughout the world, with a Linux environment that is consistent across all of them.
However, network environments are becoming ever more heterogeneous. Third generation telephony is bringing large numbers of handheld wireless devices into the Internet. Wireless mesh and ad-hoc networks may soon make it common for data to cross multiple wireless hops while being routed in unconventional ways. For these new environments, new networking applications will arise. For their development and evaluation, researchers and developers will need the ability to launch applications on endhosts located in these different environments.
It is sometimes unrealistic to implement new network technology, for reasons that can be either technological - the technology is not yet available -, economical - the technology is too expensive -, or simply pragmatical - e.g. when actual mobility is key. For these kinds of situations, we believe it can be very convenient and powerful to resort to emulation techniques, in which real packets can be managed as if they had crossed, e.g., an ad hoc network.
In our project-team, we work to provide a unified environment for the next generation of network experiments. Such a large scale, open, heterogeneous testbed should be beneficial to the whole networking academic and industrial community. It is important to have an experimental environment that increase the quality and quantity of experimental research outcomes in networking, and to accelerate the transition of these outcomes into products and services. These experimental platforms should be designed to support both research and deployment, effectively filling the gap between small-scale experiments in the lab, and mature technology that is ready for commercial deployment. As said above, in terms of experimental platforms, the well-known PlanetLab testbed is gaining ground as a secure, highly manageable, cost-effective world-wide platform, especially well fitted for experiments around New Generation Internet paradigms like overlay networks. The current trends in this field, as illustrated by the germinal successor known as GENI, are to address the following new challenges. Firstly, a more modular design will allow to achieve federation, i.e. a model where reasonably independent Management Authorities can handle their respective subpart of the platform, while preserving the integrity of the whole. Secondly, there is a consensus on the necessity to support various access and physical technologies, such as the whole range of wireless or optical links. It is also important to develop realistic simulators taking into account the tremendous growth in wireless networking, so to include the many variants of IEEE 802.11 networking, emerging IEEE standards such as WiMax (802.16), and cellular data services (GPRS, CDMA). While simulation is not the only tool used for data networking research, it is extremely useful because it often allows research questions and prototypes to be explored at many orders-of-magnitude less cost and time than that required to experiment with real implementations and networks.
The evaluation of new network protocols and architectures is at the core of networking research. This evaluation is usually performed using simulations (e.g., NS), emulations (e.g., Emulab), or in the wild experimental platforms (e.g., PlanetLab). Simulations allow a fast evaluation process, fully controlled scenarios, and reproducibility. However, they lack realism and the accuracy of the models implemented in the simulators is hard to assess. Emulation allows controlled environment and reproducibility, but it also suffers from a lack of realism. Experiments allow more realistic environment and implementations, but they lack reproducibility and ease of use. Therefore, each evaluation technique has strengths and weaknesses. However, there is currently no way to combine them in a scientific experimental workflow. Typical evaluation workflows are split into four steps: topology description and construction, traffic pattern description and injection, trace instrumentation description and configuration, and, analysis based on the result of the trace events and the status of the environment during the experimentation. To achieve the integration of experimental workflows among the various evaluation platforms, the two following requirements must be verified:
Reproducibility: A common interface for each platform must be defined so that a same script can be run transparently on different platforms. This also implies a standard way to describe scenarios, which includes the research objective of the scenario, topology description and construction, the description of the traffic pattern and how it is injected into the scenario, the description and configuration of the instrumentation, and the evolution of the environment during the experimentation
Comparability: As each platform has different limitations, a way to compare the conclusions extracted from experiments run on different platforms, or on the same platform but with different conditions (this is in particular the case for in the wild experimental platforms) must be provided.
Benchmarking is the function that provides a method of comparing the performance of various subsystems across different environments. Both reproducibility and comparability are essential to benchmarking. In order to facilitate the design of a general benchmarking methodology, we plan to integrate and automate a networking experiments workflow within the OneLab platform. This requires that we:
Automate the definition of proper scenario definition taking in consideration available infra-structure to the experiment.
automate the task of mapping the experimentation topology on top of the available OneLab topology. We propose to first focus on a simple one-to-one node and link mapping the beginning.
define and provide extensive instrumentation sources within the OneLab system to allow users to gather all interesting trace events for offline analysis
measure and provide access to "environment variables" which measure the state of the OneLab system during an experimentation
define an offline analysis library which can infer experimentation results and comparisons based on traces and "environment variables".
To make the use of these components transparent, we plan to implement them within a simulation-like system which should allow experiments to be conducted within a simulator and within the OneLab testbed through the same programming interface. The initial version will be based on the ns-3 programming interface.