Gábor Rétvári, PhD — Senior Research Fellow, BME

Gábor Rétvári received the M.Sc. and Ph.D. degrees in electrical engineering from the Budapest University of Technology and Economics (BME), Budapest, Hungary, in 1999 and 2007, respectively. He is now a Senior Research Fellow at the Department of Telecommunications and Media Informatics, BME. His research interests include routing and switching in packet networks, cloud-native networking and the service mesh, switch and router design, cloud computing and software-defined networks, and all applications of theoretical computer science to real-world problems. He is a Perl hacker, maintaining numerous open source scientific tools written in Perl, C and Haskell.

Department of Telecommunications and Media Informatics
Budapest University of Technology and Economics
Budapest, Hungary
H-1117 Budapest, Magyar Tudósok krt. 2.
Phone: +36-1-463-2403
Twitter: @littleredspam
Email: ---Photo---

News

  • L7mp, our new multiprotocol service mesh for legacy telco applications, gets its first public announcement at ServiceMeshCon NA 2020, colocated with KubeCon/CloudNativeCon NA 2020. Presentation is now online.
  • Our paper Full-stack SDN: The Next Big Challenge?, challenging the current practice of software-defined networking being restricted to the lower 3 layers of the protocol stack (Link, Network and partially Transport), has been accepted to ACM SOSR'20. The motivation for this work is the recent service-mesh hype in the cloud-native community and whether there is anything we could do about it in the SDN area. Turns our we can: we argue that it is time to extend SDN all the way up into the Application layer, which has all sorts of interesting consequences. This is joint work with Gianni Antichi from Queen Mary University, London. See the full paper, the talk and the slides online.
  • Amazing news, we got accepted to USENIX NSDI'20! The paper is titled Batchy: Batch-scheduling Data Flow Graphs with Service-level Objectives and it is just about what the title says: feeding packet processing functions in a software switch with as large packet-batches as possible, and thereby exploiting well-known batch-processing gains, subject to delay constraints. This is joint work with Tamás Lévai, Barath Raghavan and Felicián Németh. See the full paper and the obligatory tweet!
  • Two papers accepted to ACM CoNEXT'19. The first one is a short paper on removing inherent redundancy from match-action programs in a systematic way, very similar to how it's done in relational database normalization, and the other is a full paper on exploiting certain algorithmic corner cases in the venerable tuple-space-search packet classification algorithm for malicious purposes and whether there's anything one can do about it.
  • Take a look at the The Internet Routing Entropy Monitor for 6 years of scalability statistics and historical trends from the Internet data plane.

Recent publications

See selected publications and a full publication list (also with abstracts).

Courses

  • Applied Optimization and Game Theory (Alkalmazott optimalizálás és játékelmélet), VITMD097, (Hungarian/English)
  • The Internet Ecosystem and Evolution (Az internet ökoszisztémája és evolúciója, Inf MSc Internet architektúra és szolgáltatások főspecializáció, VITMMA00V)

Projects

  • The Internet Routing Entropy Monitor - statistics, charts and downloadable forwarding tables samples from the Internet data plane collected daily from 20 vantage points.
  • Math::GLPK - Perl extension for the GNU Linear Programming Kit (GLPK).
  • Math::MatrixSparse - a pure Perl module implementing sparse matrices.
  • Lemon::Graph - Perl extension for LEMON, a combinatorial optimization and graph library.
  • XML::XDV - the XML Data Visualization framework.

Created: 2020-12-29 Tue 14:05