Personal Data 

Qualifications and Career

Activities in the Research System

Organisation of Scientific Meetings

Co-organiser and co-chair of several Dagstuhl and other seminars.

Memberships in Scientific Groups and Societies

Supervision of Researchers in Early Career Phases

Current:

Completed:

Among the alumni, there are now 8 full professors / 2 FH professors, 7 associate professors / readers, and 4 assistant professors.

Academic Distinctions

Awards

Scientific Results

Anja Feldmann’s research work has contributed significantly to understanding and improving the Internet’s reliability and performance. Her contributions include work on understanding the dynamics of traffic in the Internet at large and subsequently developing the insights and tools for traffic engineering at Internet scale. A full publication list is available via dblp.

Application performance: Feldmann’s work has pointed out how to infer application performance from the network. Recently, she has studied the impact of COVID-19 on Internet traffic [A1] – highlighting the importance of well-managed networks that can react quickly to demand changes. Earlier she highlighted traffic characteristics of broadband Internet access users [A2]. Previously, she pointed out that, e.g., compression and delta-encoding (sending differences between a previously accessed Web object and the current one) are very beneficial [A3], and these are now part of almost all browser and server implementations (RFC 3229).

[A1] A. Feldmann, O. Gasser, F. Lichtblau, E. Pujol, I. Poese, C. Dietzel, D. Wagner, M. Wichtlhuber, J. Tapiador, N. Vallina-Rodriguez, O. Hohlfeld, and G. Smaragdakis. A year in lockdown: How the waves of COVID-19 impact internet traffic. Communication of the ACM, 64(7):101–108, 2021. DOI: 10.1145/3465212.  

[A2] G. Maier, A. Feldmann, V. Paxson, and M. Allman. On dominant characteristics of residential broad- band internet traffic. In Proc. of the ACM Internet Measurement Conf. (IMC), 90–102, 2009. DOI:

10.1145/1644893.1644904.

[A3] J. C. Mogul, F. Douglis, A. Feldmann, and B. Krishnamurthy. Potential benefits of delta encoding and data compression for HTTP. In ACM Conf. on Applications, Technologies, Architectures, and Protocols for Computer Communication (ACM SIGCOMM), 181–194, 1997. DOI: 10.1145/263105.263162.  

Traffic dynamics: A major insight about the dynamics of traffic is that self-similarity in Internet traffic is due to factors such as the traffic being generated by the superposition of many ON/OFF-sources or being constituted of sessions whose lengths have heavy-tailed distributions. Based on these insights she, together with colleagues at AT&T, has shown that Internet traffic has a multifractal structure and that this is a result of the dynamics of TCP/IP protocols. Further, these can be modelled by cascades to generate traffic with similar dynamics [B1]. A consequence of Internet traffic having heavy-tailed distributions is that models tend to be less tractable, and in Feldmann’s work with Ward Whitt it was shown how to approximate such distributions using hyperexpoential distributions [B2]. This not only simplifies many mathematical models, but also explains why ad-hoc approaches often lead to reasonable results.

[B1] A. Feldmann, A. C. Gilbert, and W. Willinger. Data networks as cascades: Investigating the multifractal nature of internet WAN traffic. In ACM Conf. on Applications, Technologies, Architectures, and Protocols for Computer Communication (ACM SIGCOMM), 42–55, 1998. DOI: 10.1145/285237.285256.  

[B2] A. Feldmann and W. Whitt. Fitting mixtures of exponentials to long-tail distributions to analyze network per- formance models. Performance Evaluation, 31(3-4):245–279, 1998. DOI: 10.1016/S0166-5316(97)00003-5.

Relevance of internet exchange points: With the advent of large content providers such as Akamai, YouTube, etc. the hierarchical AS structure flattened. However, the number of interconnection links between ASes for exchanging traffic on a “no-cost” basis was considered to be limited. Feldmann, together with colleagues, showed that this is no longer the case due to the increasing importance of Internet Exchange Points (IXPs) which offer new interconnection opportunities and carry traffic volumes that are of the same order of magnitude as major Internet Service Providers [C1]. This work established IXPs as major components of today’s Internet within the research community.

[C1] B. Ager, N. Chatzis, A. Feldmann, N. Sarrar, S. Uhlig, and W. Willinger. Anatomy of a large European IXP. In ACM Conf. on Applications, Technologies, Architectures, and Protocols for Computer Communication (ACM SIGCOMM), 163–174, 2012. DOI: 10.1145/2342356.2342393.  

Software defined networking: Recently, a new concept has emerged, with the potential for becoming a fundamental game-changer in networking: Software Defined Networking (SDN). SDN decouples control logic from the actual network elements and allows for more central views with more control and flexibility. Feldmann’s work has shown how to take advantage of SDN, e.g., to enable network-wide debugging [D1], but also that SDN comes with new dangers [D2].

[D1] A. Wundsam, D. Levin, S. Seetharaman, and A. Feldmann. OFRewind: Enabling record and re- play troubleshooting for networks. In USENIX Annual Technical Conf. (ATC), 327–340, 2011. DOI: 10.5555/2002181.2002210.

[D2] K. Thimmaraju, B. Shastry, T. Fiebig, F. Hetzelt, J.-P. Seifert, A. Feldmann, and S. Schmid. Taking control of SDN-based cloud systems via the data plane. In Proc. of the Symp. on SDN Research, 1–15, 2018.

DOI: 10.1145/3185467.3185468. Best Paper Award.

Network traffic engineering: Here, Feldmann’s work was influential in adding another degree of freedom to traditional traffic engineering. Her work proposed that with the advent of larger con- tent providers, traffic engineering is not only about adjusting routing but also about picking the source/destination of the traffic since there are often multiple choices, e.g., content that is duplicated at multiple locations. Thus, by changing the traffic matrix, it is possible to optimise traffic flow in the Internet. The benefits of this approach are highlighted in [E1]; it needs cooperation between content providers and Internet service providers, which is indeed possible in today’s Internet. Feldmann’s work together with colleagues also showed for the first time how to derive traffic matrices from an operational network at scale and how to use it to optimise the traffic within the ISP [E2]. This motivated many other researchers to then develop further methodologies for estimating traffic flows.

[E1] E. Pujol, I. Poese, J. Zerwas, G. Smaragdakis, and A. Feldmann. Steering hyper-giants’ traffic at scale. In ACM CoNEXT, 82–95, 2019. DOI: 10.1145/3359989.3365430. Best Paper Award.

[E2] A. Feldmann, A. Greenberg, C. Lund, N. Reingold, J. Rexford, and F. True. Deriving traffic demands for operational IP networks: Methodology and experience. IEEE/ACM Trans. on Networking (ToN), 9(3):265–280, 2001. DOI: 10.1109/90.929850.