Yesterday (Nov 29), I had the pleasure to present at the ACM Workshop on hot topics in networking (HotNets) how analog components, in particular memristors, can improve energie efficiency and expressiveness of current packet processing architectures. We propose a new programming abstraction named PCAM which accounts in contrast to TCAM memory for the analog nature memristive components. The work has been conducted in the context of the CogniGron research center. More details, on the paper “The Future is Analog: Energy-Efficient Cognitive Network Functions over Memristor-Based Analog Computations” you can find in the ACM digital library. Special congratulations to Saad Saleh for doing a great job in developing and evaluating PCAM! Finally, thanks to the HotNets organizers for a great event with many exciting propositions for future networking research.
After being officially welcomed at TU Ilmenau, Boris Koldehofe gave his inaugural lecture at TU Ilmenau on Nov 1st 2023. In his talk named “Accelerating the performance of data driven applications with in-network computing” he discussed how advances in programmable networks can aid to accelerate data-intensive applications in form of in-network computing. Besides discussing technical solutions and research challenges he highlighted also recent research towards energy-efficient in-network computing architectures.
- Bochra Boughzala, Christoph Gärtner, and Boris Koldehofe. Window-based Parallel Operator Execution with In-Network Computing. Proceedings of the 16th ACM International Conference on Distributed and Event-based Systems (DEBS ’22), pp. 91–96, ACM press.
- Ralf Kundel and Fridolin Siegmund and Rhaban Hark and Amr Rizk and Boris Koldehofe. Network Testing Utilizing Programmable Networking Hardware. IEEE Communications Magazine, 7 pages, IEEE 2022.
- Bowmik, Tariq, Koldehofe, Kohler, Dürr, Rothermel. High Performance Publish/Subscribe Middleware in Software-defined Networks. IEEE Transactions on Networking (ToN), 2016.
Interested to work towards a Ph.D. in distributed systems? The Distributed and Operating Systems Group has opened two research positions with a focus on distributed in-network computing. You can find more details about the position and how to apply at the official web pages of TU Ilmenau:
For further questions do not hesitate to contact directly Boris Koldehofe!
On February 1st, Boris Koldehofe was appointed by the president of TU Ilmenau, Prof. Dr. Kai Uwe Sattler, as Professor of the university. He will be heading the Distributed Systems and Operating Group at the Faculty of Computer Science and the Institute of Applied Computer Science. Boris Koldehofe will continue to collaborate with the University of Groningen and remains a TU Darmstadt fellow as part of the CRC MAKI.
Saad Saleh presented a novel memory architecture TCAmMCogniGron at the IEEE International Conference on Rebooting Computing, 2022. The proposed architecture TCAmMCogniGron is designed to replace a highly energy intensive memory, named Ternary content addressable memory (TCAM), used in almost all current network switches and routers to store processing rules and forward traffic through the Internet. TCAmMCogniGron enables the general purpose integration of special energy saving materials, named Memristors, which allow to keep the processing rules in form of a state, often represented in form of resistance, and this state is non-volatile. In collaboration with Anouk S. Goossens and Tamalika Banerjee from Zernike Institute for Advanced Materials characteristics of a physical memristor built at the CogniGron research center were studied. The findings demonstrate the high energy efficiency and extreme resiliency of the proposed design. The research concluded that in addition to supporting non-volatile state so-called match and mismatch operations require only about 1–16 femtojoules of energy.
Interested to learn more about the potential of memristive devices to a more energy-friendly Internet, please check out the following research works including a recent survey studying the use of memristive devices for performing energy efficient operations using novel computing architectures in the Internet.
This research is supported by the Groningen Cognitive Systems and Materials Research Center (CogniGron), University of Groningen.
- Saad Saleh, Anouk S. Goossens, Tamalika Banerjee, and Boris Koldehofe. TCAmMCogniGron: Energy Efficient Memristor-Based TCAM for Match-Action Processing. In Proceedings of the 7th International Conference on Rebooting Computing (ICRC 2022). IEEE, 2022. Available at: https://research.rug.nl/en/publications/tcaimimsupcognigronsup-energy-efficient-memristor-based-tcam-for-
- Saad Saleh and Boris Koldehofe. On Memristors for Enabling Energy Efficient and Enhanced Cognitive Network Functions. IEEE Access, 34 pages, 2022. DOI: https://doi.org/10.1109/ACCESS.2022.3226447
Congratulations to Manisha Luthra! Manisha was awarded with a dissertation award for her PhD thesis “Network-centric Complex Event Processing” by the German Special Interest group on Communications and Distributed Systems (KuVs) of “Gesellschaft für Informatik” (GI).
Manisha Luthra was co-advised by Ralf Steinmetz und Boris Koldehofe in the context of the DFG Collaborative Research Center MAKI. Manisha is now Postdoctoral researcher at the Systems Group of TU Darmstadt. Furthermore, she started recently a new position as senior researcher at the SAIDE Lab of German Research Centre for Artificial Intelligence (DFKI) led by Carsten Binnig.
More information and links:
The Rosalind Franklin Fellow program of the University of Groningen has opened several positions as Tenure Track Assistant/Associate Professor for female researchers with three positions at the Bernoulli Institute. One position is dedicated to distributed or network-centric systems research. Candidates with a strong background in this area can apply via the RUG online application portal until Oct 27th 2022.
Saad Saleh presented his research on Memristor-based Cognitive and Energy Efficient In-network computing at the Brainspiration 2022 conference, University of Twente, Netherlands. The conference focused on brain-inspired concepts and materials for information processing.
The motivation for brain-inspired computing emerges from the current generation of computers which huge amounts of energy resources and lack cognitive functionality which is critical for many applications like self-driving cars. In this regard, the CogniGron research center at the University of Groningen has developed materials, called memristors, which consume less energy resources and provide cognitive functionality similar to the human brain. Saad presented the use of these cognitive materials for the Internet components and showed various design models and techniques for incorporating cognitive functionality inside the Internet. Preliminary results have shown promising power consumption statistics of these materials for Internet components requiring little (16 uW during operational mode) to no energy consumption (during standby mode).
The speakers and attendees at the conference included renowned scientists and researchers from academia and industry including Beatriz Noheda (CogniGron), Julie Grollier (Thales), John Paul Strachan (Forschungszentrum Jülich), Abu-Sebastian (IBM), Andrea Liu (UPenn), Wilfred Gerard (UT), Wolfram Pernice (Heidelberg University), and researchers from HP labs (USA), Sandia National Labs (USA), AMOLF (NL).
This research is in collaboration with Anouk Goossens and Tamalika Banerjee from Zernike Institute of Advanced Materials, University of Groningen. For more information about this research, contact Saad Saleh or Boris Koldehofe (Adviser).
Young system researchers, e.g., Ph.D. students and PostDocs, can now apply for the EuroSys 2023 Shadow PC until October 12th.
A Shadow PC is meant to allow young researchers to obtain insight into the review process of major conferences. Typically, the review process of a big conference like EuroSys is rather intransperrant to young researchers. Therefore, it isn’t easy to understand what requires excellent research work to be accepted.
In a shadow PC (shadowing the programm commitee), young researchers actively review the research papers submitted to the conference. In EuroSys, papers submitted to the conference will also be assigned to members of the shadow PC for review with the permission of the authors. In addition, the shadow PC will determine and discuss its own ranking of submitted papers. Although the outcome will and must not impact the decision of the conference program committee, the discussion allows the researchers to obtain in-depth experience in the review process. Moreover, the reviews will also provide, in many cases, additional helpful
feedback to the authors.
Interested in joining? Please, check out the Call for Participitation of EuroSys 2023!
After a long time of virtual conference events, our distributed systems research team had finally the opportunity to participate in physical meetings.
Two highlight this summer where our team attended and contributed were
- The conference 16th ACM Conference on Distributed and Event-based Systems (DEBS) in Copenhagen
- ACM SIGCOMM, the flagship annual conference of the ACM Special Interest Group on Data Communication (SIGCOMM), this year located in Amsterdam.
Three contributions at the ACM DEBS conference
At DEBS our team was present with three contributions on improving the performance of stream and event processing systems.
|„Travel Light – State Shedding for Efficient Operator Migration” is a cooperation with the University of Oslo, with Espen Volnes , Thomas Plagemann and Vera Goebel. (University of Oslo). The paper looks into the combination of state shedding in combination with operator migration to improve the utility of detected events.|
|The work of “Window-based Parallel Operator Execution with In-Network Computing” was presented by Bochra Boughzala. It proposes a novel P4-based in-network computing model that can enhance the parallelization degree of event processing systems by utilizing the capability of programable network switches.|
|In the Poster and Demo track Pratyush Agnihotri presented with “PANDA: Performance Prediction for Parallel and Dynamic Stream Processing” his idea on better performance modelling over heterogenous compute resources by building on zero-shot learning models.|