Agenda - TORNOG 1

Modern networks change constantly — routing policies evolve, security controls tighten, data centres expand, and automation is used more than ever. Yet most operational teams still have no platform to test changes other than ad-hoc lab environments, or worse, production. The challenge for modern network operators: Network operators increasingly manage multi-vendor environments Change velocity is increasing (automation, segmentation, DC refresh, etc.) Routing and security changes carry real risk Most teams still lack structured pre-production validation We would like to show that Lab-as-Code is a practical, vendor-agnostic approach to building reproducible, automated network test environments using open tools and infrastructure-as-code practices. We will explore: Designing topology definitions as code using open emulation frameworks Validating routing, policy, and control-plane behaviour before deployment Automating configuration testing and verification Integrating network validation into CI pipelines This session focuses on repeatable workflows and guardrails that apply across platforms and architectures. Attendees will leave with a practical blueprint for building reproducible test environments. We will show how implementing reproducible test environments can reduce operational risk, increase deployment confidence, and improve the change process — without requiring large capital investments.

Network operators have been using MPLS-TE auto-bandwidth for traffic load balancing. Segment Routing doesn't has no native mechanism for auto-bandwidth, and the existing implementations are vendor-specific. I would like to introduce a simple, standard-based and multi-vendor auto-bandwidth implementation for SR-TE, which includes an open source SR-TE bandwidth sampler.

Since 2020, we have developed and refined a selection of scripts used to collect operational data from routers across the NTT Global IP Network, a "Tier 1" IP transit provider. This automation has gone through four major phases. We started with scripts screenscraping data from devices using textfsm and netmiko while being ran in a crontab. We then introduced data collection using Netconf. Third, we've introduced a Gitlab CI/CD pipeline to maintain them. Finally, we modularized the scripts to make common functions reusable. This presentation will detail what I learned throughout this journey. This is intended to demonstrate that successful network automation can be learned and implemented in small increments as opposed to an all-or-nothing approach.

Contemporary Network Automation is a complex process that not only orchestrate multiple systems and data sources but also disparate business units that both support and consume the automation. This talk will explore the difficulties of developing and maintaining organization-wide Network Automation by examining common automation failures that have been observed in real-world scenarios across multiple enterprise Network Automation Programs. Listeners will leave this talk with a deeper understanding of the root causes of many Network Automation failures, a vocabulary that can be used to discuss these issues, and tangible mitigation strategies that they can use to streamline their own organization's automation processes.

Route leaks and hijacks continue to affect networks of all sizes, including regional ISPs, enterprises, research networks, and public-sector organizations. Resource Public Key Infrastructure (RPKI) is one of the most practical tools available today for improving BGP routing security - but many operators still have questions about what it really does, how risky it is to deploy, and where to start. This session provides a clear, high-level introduction to RPKI, with a focus on real-world operations rather than theory or standards language. We’ll cover what RPKI is, why does it matter to your network, and how RPKI works in practice. Attendees will leave with a realistic understanding of how to begin adopting RPKI in a cautious, operator-safe way.

Free for attendees. Select as an add-on during ticket purchase.

On July 8, 2022, a single misconfigured BGP policy filter at Rogers caused full Internet routing tables to flood into OSPF, crashing every core router in the network. Twelve million Canadians lost Internet, wireless, and even 911 service for nearly 19 hours. The CRTC investigation revealed that basic safeguards — route limits, redistribution controls, overload protection — were missing. Two years earlier, Rogers prefixes had been hijacked by a small US-based AS that announced more-specific routes that the network had never originated. And it was not until August 2025 that Bell Canada, the country's largest ISP, began filtering RPKI-invalid routes. These are not distant, theoretical problems. They are Canadian incidents, affecting Canadian operators and Canadian users — and they expose gaps that many networks in the Toronto area and across Ontario still have today. This talk examines what has changed since the Rogers outage and what hasn't. We start with the incident itself: what went wrong, what the CRTC found, and what it tells us about the state of BGP operational hygiene in Canadian networks. We then look at the broader routing security landscape through three defensive layers that every operator — from regional ISPs to enterprises — should consider. Layer one is ROV: validating that route origins match RPKI records, and why Bell's late adoption left a gap in the Canadian routing ecosystem for years. Layer two is ASPA: the emerging path validation standard that all five Regional Internet Registries have committed to supporting by the end of 2026, which would catch the route leaks that ROV cannot. Layer three is BMP: using BGP Monitoring Protocol data to detect routing anomalies that neither ROV nor ASPA can surface — including stealthy hijacks where your control plane looks clean but your traffic is being diverted. Attendees will leave with a concrete, prioritized checklist for hardening their BGP security — starting with the quick wins they can deploy this quarter.

The optical transport world tends to be viewed as "that other team's domain" or "let the carrier deal with it", but there's a lot of interesting stuff contained within, and even if you don't want to be an Optical Person it's great to be able to talk to providers in their own layer 1 language. We'll talk a bit about "the world beyond 1310nm", how the network world meets the physical (and physics) world, and get an OTDR and a few kilometres of fibre out and take a hands-on look at some troubleshooting. (* fun may not actually be present during outages)

According to a number of cloud vendor usage studies, cross-availability zone (AZ) data transfers regularly account for at least 25% of the public cloud users’ production cost. Cutting down these costs can affect your bottom line and your application affordability. All major cloud vendors provide daily aggregated cost metrics for the cross-zone network traffic, however oftentimes, these reports lack the granularity of information to tell which pods or workloads are responsible for the elevated cross-zone traffic. I will walk through our experiences with slashing network costs at Grafana by using eBPF network monitoring. We demonstrate how we leveraged eBPF to perform packet level traffic analysis in a manner that’s performant, requires zero-code changes and has extremely low overhead. We show how by enriching the low level network data with Kubernetes pod and node information, we can effectively identify cross-AZ traffic by using only OSS tools, while gaining detailed insights into the traffic patterns of each pod. We will show you how you can build your own solution for monitoring cross-AZ network traffic and identify savings in your cloud bills, regardless of the Kubernetes CNI you are using.

As Canadian network operators navigate the first full year of the Critical Cyber Systems Protection Act (CCSPA), the complexity of maintaining both 99.999% availability and $15M/day regulatory compliance has reached a tipping point. Traditional, siloed approaches to NetOps and SecOps are no longer sustainable in a world of AI-weaponized threats and multi-cloud fragmentation. This session explores the technical architecture of AgenticOps—an AI-first operational model that leverages domain-specific reasoning to bridge the gap between network performance and continuous security.

At AutoCon4, John said something that stuck with people: Get on the train. AI isn’t optional. If 30% of engineers are building agents, the other 70% aren’t competing with peers anymore — they’re competing with automation that never sleeps. Watching from the platform while it accelerates isn’t a strategy. At the last event, the #1 question afterward wasn’t “Is this real?” It was: “Okay… but HOW?” This TORNOG talk is the answer. Not theory. Not hype. Not vendor slides. This is a chronological practitioner’s journal — from November 2022 to Munich 2026 — documenting: The first clumsy scripts The early GPT experiments The agent frameworks that worked (and the ones that didn’t) The security scares The real production use cases The honest failures The breakthroughs that changed everything This is what it actually looks like to evolve from network engineer → automation engineer → AI-enabled builder. And because this is TorNug Toronto, this isn’t abstract. It’s practical. It’s local. It’s “what can I build after work this week?” energy. You won’t just hear a story. You’ll walk out with: A clear mental model of where AI fits in network engineering A realistic adoption path (without quitting your job to become an ML PhD) A personal AI roadmap you can start tonight No hype. No fear tactics. Just signal. The train is still at the station. 🚆

Our jobs are about to change, and free and open software is going to change with them. That change has to happen, but how it happens isn't cast in stone, and if we want it, we can still have a say in the shape of the future.

Based on his hobby networking experience, Quantum will be talking about the logistics and costs of joining Internet Exchanges, what you can do with them, as well as sharing some fun and not-so-fun stories.

Based on decades of experience working with Canadian operators, this presentation explores evolving trends over past few years in fiber network architectures from both a networking technology perspective and across the broader network lifecycle of engineering , construction, and intelligent network management and automation. We will also highlight examples of practical AI applications currently being used in the field today.