🗓️ Event Agenda
Start | End | Presentation | Category | Speaker |
---|---|---|---|---|
8:00 AM | 9:00 AM | Registration / Breakfast | ||
9:00 AM | 9:15 AM |
Opening RemarksWelcome and opening remarks to the conference. |
Tom Kacprzynski
|
|
9:15 AM | 9:45 AM |
How networking changed with pre-training, post-training and inferenceHigh-performance cluster networking for GPU systems has been traditionally associated with large-scale pre-training. As time passed, post-training and distributed inference applications brought their unique requirements as well. This talk surveys the training/inference parallelism-induced traffic patterns from training/inference of LLMs and the impact on evolution of scale-out and scale-up network. |
AI Networking |
Petr Lapukhov
|
9:45 AM | 10:15 AM |
Networking at the Speed of Light – A peek at networking inside a high frequency trading firmThe high frequency trading industry is a highly competitive field that relies on secrecy to maintain relative advantages. These companies also have unique technology requirements that make it difficult to recruit capable engineers who do not already work within the industry. During this presentation I’ll describe three unique aspects of trading, and how they impact network design and operation: 1) The need for speed – Why it is important and how to optimize it using wired and wireless connections 2) Distributing market data – What is it, where is it, and how to transport it efficiently using IP Multicast 3) Timestamping – Why accurately timestamped packet captures are the lifeblood of the industry |
Low Latency Networks |
Jeremy Filliben
|
10:15 AM | 10:45 AM | Break 1 | ||
10:45 AM | 11:15 AM |
Feasibility of State Space Models for Network Traffic GenerationMany problems in computer networking rely on parsing collections of network traces (e.g., traffic prioritization, intrusion detection). Unfortunately, the availability and utility of these collections is limited due to privacy concerns, data staleness, and low representativeness. While methods for generating data to augment collections exist, they often fall short in replicating the quality of real-world traffic In this paper, we i) survey the evolution of traffic simulators and generators and ii) propose the use of state space models, specifically Mamba, for packet-level, synthetic network trace generation by modeling it as an unsupervised sequence generation problem. Preliminary evaluation shows that state space models can generate synthetic network traffic with higher statistical similarity to real traffic than the state-of-the-art. Our approach thus has the potential to reliably generate realistic and informative synthetic network traces for downstream computer networking tasks. |
Network Modeling |
Nick Feamster
|
11:15 AM | 11:45 AM |
What a Distributed Network Sees: Internet Observations from All OverInternet activity, connectivity, and performance can vary depending on from where you make your observations. A single network can make some inferences about what it sees, but only a widely distributed network with vantage points around the globe can distinguish isolated events from global phenomena. We demonstrate the insight of a distributed network of over 500 servers, with IP addresses originating from more than 150 ASNs, and physically located in 70 different countries. We first show how this network helped solve a source address spoofing mystery and confirm the true source through triangulation. Then, we illustrate how popular port scanning systems appear in the global address space. Lastly, we show how this distributed network is helping to facilitate reachability experiments for academic research projects. |
Network Observability |
John Kristoff
|
11:45 AM | 12:15 AM |
Building Trustworthy Network Automation, From Principles to Practiceabstract |
Network Automation |
Damien Garros
|
12:15 PM | 1:45 PM | Lunch (Onsite) | ||
Afternoon Track A | ||||
1:45 PM | 2:15 PM |
SRv6 Deployment StrategiesThe IETF has an Operations and Management (Ops) Area with many Working Groups. The newest WG is the SRv6 Ops WG and one of the topics is how to transition an existing network to one that supports SRv6. It is one of the first questions network operators will ask about SRv6. There are a couple of IETF drafts now addressing this topic. Mike McBride will provide an overview of this work in the IETF and provide a summary of how to upgrade an existing network to a SRv6 network. Should they run ships in the night, utilize various tunneling techniques or other deployment solution? If they are currently running an IP/MPLS network how should they transition to SRv6? This presentation summarizes various deployment alternatives to help provide guidance to those wanting to upgrade their network to SRv6. |
Routing |
Mike McBride
|
2:15 PM | 2:45 PM |
Comfortable Complexity of OverlaysThis presentation examines the intricate landscape of overlay technologies, tracing their historical development, dissecting current deployments, and prognosticating future directions. We will explore the evolution from basic VLANs to advanced solutions like VXLAN, GENEVE, and SRv6, analyzing the architectural nuances, protocol intricacies, and operational considerations that define each. The critical roles of microsegmentation and tunneling in overlay functionality will be examined, along with the trade-offs between hardware and software implementations. A deep dive into the driving forces behind technological trends, such as virtualization, will highlight the paradigm shifts in data center architecture and the resultant demands on network overlays. The intricacies of L2 and L3 VPN evolution, including EVPN standardization, will be reviewed, alongside real-world deployment scenarios illustrating multitenancy, L2 stretching, and enterprise Zero Trust architectures. Finally, the presentation will address the ongoing trends shaping the field, including the convergence of L2 and L3 services, cost optimization, and the imperative of multivendor interoperability. |
Datacenter Networking |
Sergey Kolobov
|
2:45 PM | 3:15 AM |
Fighting Route Leaks at CloudflareCloudflare runs a vast global network with complicated peering relationships that present a unique complexity for detecting route leaks. AS-to-AS relationships alone are insufficient to distinguish a legitimate BGP prefix announcement from a leaked announcement. Instead, each prefix needs to be precisely evaluated at the prefix level for path validity. In this presentation, we will show you how we built prefix-level route leak detection at Cloudflare with a highly scalable pipeline. We will share our methods that other networks might find useful in capturing their own route leaks, especially to those who operate BGP Anycast networks. In addition to this, we will discuss our progress toward BGP ASPA (Autonomous System Provider Authorization) compliance, Cloudflare’s strategy to leverage RFC9234 Only-to-Customer (OTC) attribute, and how they fit alongside our prefix-level BGP route leak detection as preventative measures. |
Routing Security |
Bryton Herdes, Mingwei Zhang
|
3:15 PM | 3:45 PM | Break 2a | ||
3:45 PM | 4:15 PM |
The Routing Security Crystal Ball: RPKI Yesterday, Today and TomorrowJoin us for a deep dive into the evolution of Routing Security and the key factors that led to the development of Resource Public Key Infrastructure (RPKI). We’ll explore the timeline of its initial deployments and the pivotal moments that brought RPKI to the forefront of the global Internet community. Recent U.S. government initiatives have prioritized the creation of RPKI Route Origin Authorizations (ROAs) to bolster network security. Finally, we’ll look ahead to the future of RPKI, examining upcoming developments and long-term plans being discussed within standards bodies and the broader Internet community. |
Routing Security |
Jon Worley
|
4:15 AM | 4:45 AM |
SD-WAN is Dead, Long Live SD-WAN!Since the emergence of SD-WAN in 2014, the WAN ecosystem has undergone a major transformation. Traditionally, businesses relied on carrier-managed private WAN solutions, but today, most have transitioned or are in the process of transitioning to SD-WAN leveraging Internet-based connectivity. Improvements such as service availability, bandwidth capacity, and link quality combined with lower costs and the need for more flexible connectivity options as applications migrate from on-premises to the cloud made this a relatively obvious business choice. However, for many businesses, the transition has not been as seamless as expected. In regions like the United States, Canada, and Europe, Internet-based WAN connectivity performs comparably to legacy private circuits, making it a viable replacement. Yet, in other parts of the world, inefficient routing presents significant challenges. For example, traffic between South American nations often routes through Florida or Houston, and traffic between African and Middle Eastern nations frequently routes through Frankfurt or London. Even in 2025, some networks—particularly “”eyeball networks””—remain poorly peered in certain regions, making Internet-based WAN transport suboptimal and often unfit for purpose. Conversely, other networks—””content networks””—are exceptionally well-peered in these regions. The key to overcoming the challenge of Internet-based WAN transport lies in leveraging these well-peered content networks to provide transit, creating regional network bridges, and optimize WAN routing. SD-WAN, with its intelligent traffic steering and dynamic path selection, is the solution to unlocking the full potential of Internet-based WAN transport.” |
SD-WAN |
Dane Jackson (Cato Networks)Dane Jackson is a seasoned technology leader with over two decades of expertise in network engineering, cyber security, and IT infrastructure. As the Director of Global Professional Services at Cato Networks, he spearheads global initiatives, driving innovation and excellence in the field of Security Access Service Edge. Before joining Cato Networks, Dane held a key management role at Hyatt Hotels Corporation, where he managed global network operations, ensuring seamless connectivity across corporate offices and hotel properties worldwide. His tenure at West Monroe Partners saw him leading complex enterprise networking and security projects, particularly in the M&A vertical. Additionally, his experience at Purdue Research Foundation and Purdue University laid a strong foundation in network architecture and systems administration. Along with a B.S. in Network Engineering Technology from Purdue University, Dane also holds multiple industry certifications from Cisco Systems, VMware, and Juniper along other vendors underscoring his deep technical expertise in the fields of cloud networking, cybersecurity, SD-WAN, and IT infrastructure modernization. |
Afternoon Track B | ||||
1:45 PM | 2:15 PM |
Vendor-agnostic automation with Ansible Network Resource ModulesNetwork resource modules are a newer addition to Ansible, providing a (semi)-consistent interface to device fact collection and configuration section management across vendors.However, examples of how to use them agnostically are difficult to find. This talk will walk through calling network resource modules to gather facts from network devices and manage configuration with several examples while highlighting pitfalls. |
Network Automation |
Brandon Ewing
|
2:15 PM | 2:45 PM |
Network Automation Pipelines – From Near-Zero to Hero in 30 MinutesIn this session, we’ll walk through the creation of a modern, fully integrated network automation pipeline that starts with a service request in Jira Service Management and ends with a verified configuration pushed to a Palo Alto firewall — all in under 30 minutes. Using GitHub Actions as the automation engine, we’ll demonstrate how to orchestrate workflows that interact with Ansible playbooks, Nautobot for source of truth and intent validation, and Palo Alto firewalls for real-world enforcement. We’ll also show how to automatically update the Jira ticket at each step for traceability and status awareness. Attendees will see a live demo of infrastructure-as-code principles applied to network operations, gaining practical insights into how to streamline request fulfillment, improve change reliability, and build a scalable automation framework that aligns with DevOps practices. Whether you’re just starting your automation journey or looking to enhance an existing toolchain, this talk will showcase how to deliver real business outcomes through automated, compliant, and repeatable pipelines. |
Network Automation |
Jon Howe
|
2:45 PM | 3:15 PM |
ContainerLab: Building Multi-vendor labs as code (Part 1)These immersive, two 30 minute hands-on sessions guide participants through a structured, incremental build-out of a modern data center fabric using a Lab-as-Code approach powered by Containerlab. The session begins with deploying a transparent IP CLOS topology, then layers in VXLAN/EVPN overlay to introduce network virtualization, and concludes with the integration of streaming telemetry for real-time observability. Each lab builds seamlessly upon the previous, demonstrating how to rapidly prototype and validate architectures using open standards, freely available tools, and reproducible infrastructure definitions. Designed for engineers and architects, this session highlights the power and simplicity of using Containerlab to model, test, and monitor next-gen network designs with agility and precision. All attendees are requested to bring a laptop to follow along the labs and get most out of the sessions. |
Networ Lab Tooling |
Amer Fahar and Mo Zaman
|
3:15 PM | 3:45 PM | Break 2b | ||
3:45 PM | 4:15 PM |
ContainerLab: Building Multi-vendor labs as code (Part 2)These immersive, two 30 minute hands-on sessions guide participants through a structured, incremental build-out of a modern data center fabric using a Lab-as-Code approach powered by Containerlab. The session begins with deploying a transparent IP CLOS topology, then layers in VXLAN/EVPN overlay to introduce network virtualization, and concludes with the integration of streaming telemetry for real-time observability. Each lab builds seamlessly upon the previous, demonstrating how to rapidly prototype and validate architectures using open standards, freely available tools, and reproducible infrastructure definitions. Designed for engineers and architects, this session highlights the power and simplicity of using Containerlab to model, test, and monitor next-gen network designs with agility and precision. All attendees are requested to bring a laptop to follow along the labs and get most out of the sessions. |
Networ Lab Tooling |
Amer Fahar and Mo Zaman
|
4:15 PM | 4:45 PM |
Operation Technology Approach to NetDevOpsThe purpose of this presentation is to present the concept and use cases for NetDevOps Operational Technolgy. Section 1: Problem Statement Section 2: NetDevOps an ethical embellishment of DevOps Section 3: Use Cases Section 4: Partnership/collaboration Section 5: Onboarding/Level of effort/Learnings Section 6: Pragmatic NetDevOps OT Roadmap |
Network Automation |
Phillip McGough
|
4:45 PM | 5:00 PM | Closing Remark – Survey | ||
5:00 PM | 5:30 PM | Break | ||
5:30 PM | 7:30 PM | Social Event / Food & Drinks / Raffle |