Skip to main content Link Menu Expand (external link) Document Search Copy Copied

ACM SIGMETRICS Workshop on Measurements for Self-Driving Networks, 2023

Call For Short Papers

The design and implementation of autonomous or self-driving networks, where network management and control decisions are increasingly made in real-time and without humans in the loop, are among the grand challenges of networking research today. Recent technological (e.g., SDN, 5G networks, etc.) and scientific innovations (e.g., XAI, transformers, etc.) have identified exciting new opportunities for realizing this vision. These advances and innovations include (1) fully programmable, protocol-independent data planes and languages for programming them; and (2) the emergence of scalable platforms for processing distributed streaming data while leveraging the latest (big) data analysis and machine learning (ML) tools and software. In particular, the coupling of programmable control capabilities in the data plane with advanced ML-based inference capabilities promises unprecedented opportunities for querying the state of the network at scale to obtain the necessary input data to the many different network management and control tasks that self-driving networks are required to perform automatically and autonomously.

However, progress towards developing such self-driving networks has been impeded by a paucity of system designs that are deployable in practice (i.e., are scalable and robust) and an excess of ML-based inference tools that are not production-ready (i.e., do not generalize, cannot be trusted, do not ensure system safety). Specifically, realizing the vision of self-driving networks that work in practice will require scalable system designs that leverage closed-loop feedback at different levels to ensure strong robustness properties with respect to the dynamic uncertainties in their environments. It will also demand a paradigm shift with respect to developing ML-based inference solutions whereby a learning model’s success will be measured in terms of properties such as explainability, trustworthiness, and safety rather than traditional metrics such as accuracy.

This workshop will provide a forum for researchers to present and share their latest research on new technologies that can help realize practical, deployable self-driving networks. This workshop seeks contributions from experts in networking, machine learning (applied and theoretical), network security, control theory, distributed systems, computer architecture, data science, etc., who share in the excitement of realizing the vision of self-driving networks. Strong preference will be given to research papers that describe original ideas related to the design of scalable and robust systems for monitoring and measuring network state and the development of production-ready ML-based tools that focus on explainability, trustworthiness, safety, etc., and enable the type of high-stakes decision making that self-driving networks demand.

Submissions related to all aspects of designing and building self-driving networks are welcome. Topics of special interest include

  • Flexible collection of diverse and high-quality data from realistic network environments
  • Design of in-network measurement architectures
  • Design, implementation, and deployment of testbeds for self-driving networks
  • Automated closed-loop traffic-engineering systems (e.g., routing, machine-learned TCP, or hypervisor rate controllers)
  • Development of trustworthy ML models for inferring (and subsequently mitigating) network attacks or network performance issues
  • New machine learning problems and questions that arise from network operations tasks that are related to performance or security
  • Network measurement techniques that adapt collection or measurement based on changing network conditions
  • New algorithms for performing approximate queries (with accuracy guarantees) and dynamic queries
  • Robust architectures for fine-grained and programmable network monitoring
  • Design and implementation of closed-loop feedback controls for ensuring system robustness to uncertainties in the environment
  • Examples of design choices informed by control-theoretic findings (e.g., hard limits, unavoidable tradeoffs)
  • AI/ML-based production-ready (i.e., explainable, trustworthy, safe) solutions for different management and control tasks (e.g., congestion control, QoS/QoE optimization, network security, packet scheduling, traffic classification, device fingerprinting, etc.)

In addition to presenting short papers, the workshop will include presentations from invited speakers, a panel discussion, and technical discussions between speakers and workshop participants.

Travel Support: We will provide travel support (including conference registration fee) to all workshop speakers and panelists.

Workshop Organizers

Important Dates

  • Paper submission: ~April 29, 2023~ May 26, 2023
  • Author notification: June 2, 2023
  • Final version submission: June 7, 2023
  • Workshop: June 19, 2023

NOTE: Authors will need to register upon being notified of acceptance.

Room

Magnolia 17

Submission Instructions

The forum encourages the submission of short papers describing early research on measurements for self-driving networks. Authors are asked to submit 3-page papers using the standard PER format (http://www.sigmetrics.org/sig-alternate-per.cls). Papers should be submitted electronically, as an e-mail attachment, to measureselfdn.sigmetrics23@gmail.com, with the email subject being: “Submission to Workshop on Measurements for Self-Driving Networks at SIGMETRICS.” Submission needs to be in PDF format. If there is any question, please contact Arpit Gupta (arpitgupta@ucsb.edu).

Accepted papers will be part of the workshop program and available during the workshop. Revised versions of these papers will be published in a special issue of ACM Performance Evaluation Review (PER) https://www.sigmetrics.org/per.shtml, and authors will have the opportunity to incorporate the results of or simply mention any relevant discussions at the workshop. Authors of accepted papers grant ACM permission to publish them in print and digital formats. Note that there are NO COPYRIGHT issues with PER, and thus authors retain the copyright of their work with complete freedom to submit their work elsewhere. The accepted workshop papers will appear in the ACM PER (https://www.sigmetrics.org/per.shtml).