“Dependable Networking in a Dynamic World”
The work at DACS focuses on the design and analysis of dependable networked systems. A networked system is called dependable, whenever reliance can justifiably be placed on the services it delivers. This means that in the various (and changing) fields in which communication systems are being deployed, these systems have to be designed such that they fulfil the possibly changing requirements of its end-users. To put it briefly, the DACS group focuses on dependable networking in a dynamic world.
Dependability of networked systems is of crucial importance to our society, where the possibility to immediately obtain and exchange the latest information, irrespective of place or time, has become a fundamental requirement. For some end-users the impact of failures within the (public) network may be limited to being unable to browse the web, send messages or make phone calls. For other end-users, e.g., like those in healthcare applications, utility companies or automotive applications, the impact of failures can be such that important societal services get disrupted, or even lives are put in danger.
The term dependable, as described in the literature, implies that the networks should be secure, maintainable, well performing, energy aware, reliable and available, and robust. DACS focuses on all these aspects, both in the design of new (often embedded) networks, as well as in the analysis of existing (operational) networks.
The research challenges in the design of new networks are often related to performance and scalability, energy awareness, reliability and availability, and robustness. For operational networks, i.e., the Internet and public wireless networks like UMTS and LTE, the research challenges are most often related to security and maintainability (manageability).
Within DACS, we distinguish two broad classes of network systems: generic networks and networks in context, as follows:
- Examples of generic networks include the Internet and wireless networks such as UMTS and LTE. Many generic networks have already been developed, which means that challenges are related to deploying such networks, and keeping them operational with acceptable performance and availability. For such networks DACS therefore focuses on network and service management issues. Key challenges include network security, as well as capacity provisioning.
- The research on embedded networks (also often referred to as cyber-physical networked systems) started more recently. For such systems, the operating context is crucial for the requirements, e.g., with respect to performance, energy-use or reliability, and therefore plays a key role in the design. Currently there is a focus on specific networks in the automotive, the utility (energy and water), and the healthcare domain. Within the automotive sector DACS performs research in the area of vehicular networking, to improve the effective capacity of highways, to improve traffic safety, and to make transportation more environment-friendly. Within the utility sector DACS performs research on SCADA networks, primarily for the supply of water and energy, as well as on communication systems to support smart grids. Healthcare oriented research has only recently started, with focus on the performance of medical equipment, such as X-ray systems.
To enable the design of dependable communication systems, DACS employs a set of tools and techniques, based on:
- Measurements. For operational systems, such as the Internet, taking measurements is key to understanding the system’s health (performance and security-wise). DACS has a strong focus on flow-based analysis of Internet traffic, to detect traffic anomalies and intrusions. To improve the manageability of networks, DACS also develops self-management techniques for automated configuration. Key challenge is to find the right balance between accuracy of, and induced overhead by the measurements, thereby avoiding privacy issues.
- Modelling, analysis and simulation. When no operational system is (yet) available, a model-based approach towards exploring the design space is an important alternative. Based on such models, either mathematical analysis, including queuing theory and model checking, as well as simulation techniques can be employed. New and specialised tools and techniques for model checking and (rare-event) simulation are being developed
- Prototyping. As a further step, prototyping of systems allows for real-life testing new designs and algorithms.
Newly Acquired Projects in 2012 and 2013
In 2012 DACS has been very successful in the acquisition of new EU projects. The Flamingo Network of Excellence started in November 2012 and is coordinated by DACS. The Mobility 2.0 STREP started in September 2012, and the Mobile Cloud Networking IP started in November 2012; this project is one of the core EU projects for the development of 5G (5th generation mobile communications).
In 2013 the EU STREP projects SALUS and E-Balance were accepted for funding; SALUS (on next generation networks for public protection and disaster recovery) started September 2013, and E-balance (on energy balancing in smart neighbourhoods) started October 2013. Work also started in a new STW project on Cyber Physical Systems, in particular on energy-aware data-centres.
Prizes and Awards in 2013
The 2013 IRTF Applied Networking Research Prize (ANRP) was awarded to Idilio Drago for the paper "Inside Dropbox: Understanding Personal Cloud Storage Services". This paper, which provides a deep characterization of the leading solution in cloud storage (Dropbox), was presented at the ACM Internet Measurement Conference in November 2012. The IRTF received 36 nominations for the 2013 prize.
The paper “Measurement Artifacts in NetFlow Data”, authored by Rick Hofstede, Idilio Drago, Anna Sperotto, Ramin Sadre, Aiko Pras and presented at the Passive-Active Monitoring Conference in Hong Kong, won the best paper award. The paper presents an in-depth analysis of artifacts found in flow exporter devices.
Giovane Moreira Moura, who graduated in March 2013 in the DACS group, received the second prize with the KIVI/NIRIA Telecommunications Award for his PhD thesis entitled “Internet Bad Neighbourhoods”. The award is presented annually for the best PhD-theses of the three technical universities.
PhD student Hamed Ghasemieh was awarded a second place in the CTIT poster competition that took place during the CTIT Symposium 2013 at the University of Twente with a poster entitled “Survivability evaluation of critical infrastructures using hybrid Petri nets”.
For generic wired networks like the Internet, DACS will continue to focus on network and service management aspects. Within the Internet speeds increase to tens of Gbps, which implies that traffic measurements should be scalable. Techniques that capture and analyse all individual network packets are getting harder to apply, but flow-based measurement techniques, such as NetFlow and IPFIX, provide scalable and cost-effective alternatives that can be applied to solve many security and provisioning related problems. To validate the results of our research, ground truth is of key importance. Such ground truth can be obtained from real network traces, which means that the collection and sharing of (anonymous traces) traces, hence, collaboration with operators such as SURFnet and Quarantainenet, will be further pursued. At European level we will collaborate with institutes like INRIA to create a common network security lab. The increased reliance on generic wireless networks makes dependability aspects also more important in that area. The DACS group will especially address the flexible deployment and management of wireless networks, e.g., using concepts from Cloud computing, and the robustness of such networks.
Increasingly networks are designed in the context of specific, extremely demanding applications. Here, the DACS group is extending its current research on vehicular networks to Wireless Networks for Moving Objects (WiNeMO) in general. These are large systems of moving objects, such as intelligent transportation system, robot swarms, etc., that use their networking capabilities to cooperatively govern their movements. Wireless Networks for Moving Objects are fundamentally different from wireless networks that have been studied up to now, such as wireless LANs, cellular networks and ad-hoc networks. The strict requirements regarding delay and reliability, combined with the high network load coming from frequent communication between a potentially high number of objects, makes the design and application of wireless networks for moving objects extremely challenging. The DACS group will focus its research on the modelling of the fundamental behaviour of such networks and on the design and prototyping of specific applications, e.g., vehicular networking for fully electrical vehicles.
As another class of networks in context, DACS will continue to expand its work on networks for SCADA and smart-grid applications in the utility sector (foremost water and electricity). As a general trend in both these contexts, the employed networks become more open (open standards) as opposed to the proprietary solutions used in the past. Although this opens opportunities for better performance and more efficient operations, it also opens up cyber threats for these networks; both aspects will be worked upon. Furthermore, in SCADA networks the effects of misbehaving or malfunctioning networks and the societal services of the utilities they control (water or power delivery) is enormous; this interplay will also explicitly be addressed.
As for techniques and tools, DACS will continue to use a variety of well-developed open source software packages for measurement (YAF, NFSen), and simulation (OMNET++). However, we will also further develop specific methods and techniques, in the area of simulation, and in the area of analysis. As for discrete-event simulation, we will continue our efforts to deal efficiently with so-called rare-events; these are system events that are of utmost importance to study (like buffer overflows or component failures), however, their rarity makes that traditional simulation methods will just take too long. We will specifically also work on embedding our new developments in “standard” tools, so that they can easily used by non-experts. As for analysis methods, we will continue our work on system of very many interacting but similar objects, like they appear in wireless sensor networks or the Internet. We will continue to develop efficient new analysis methods based on mean-field analysis, and combine this with our previous results on stochastic model checking. To study the effect of malfunctioning SCADA systems on the applications they support, also new hybrid (discrete-continuous) analysis methods will be developed.
Numerical overview of DACS results
UT funding (k€)
EU/NWO funding (k€)
(All text and data from UT/EWI/QAR 2013)