You're reading the Wednesday, October 14, 2026 edition. Today's briefing →
Toronto
Wednesday, October 14, 2026No. 122

Digital Plumber

Plumbing the information age

AI-curated intelligence for people who run networks. Daily coverage of AIOps, network automation, agentic operations, AI infrastructure, security and the vendors shaping them.

Today's 3 things that matter

Picked by the AI editor
  1. Security·Industry news

    Ivanti Connect Secure zero-day exploited for unauthenticated root access

    CISA urged immediate patching of CVE-2026-104935, an Ivanti Connect Secure authentication bypass exploited in zero-day attacks since October 1st that lets remote threat actors execute arbitrary commands without credentials.

    Why it matters Network ops teams managing remote-access VPN infrastructure must immediately patch Ivanti Connect Secure versions 25.1.1.1 and prior; CVE-2026-104935 is on the CISA KEV catalog with a October 16 federal deadline and active exploitation in production environments.

  2. Routing·Analysis

    Strike and BGP collapse cascade through Russian networks

    A UAV strike on Yandex's Sasovo data center combined with routing collapse in Qrator Labs' BGP filtering network (AS197068 and AS200350) caused cascading outages across Russian infrastructure on October 8, 2026.

    Why it matters Demonstrates how concentrated infrastructure and BGP routing vulnerabilities compound during physical disruptions—critical for designing resilient multi-vendor, geographically diverse routing architectures.

  3. Automation·Primary source

    NetBox Labs adds NVIDIA and Grafana integrations to platform

    NetBox Labs announced partnerships with NVIDIA and Grafana Labs at NetBox Evolve conference, expanding the Infrastructure Intelligence Platform with new integrations and AI-agent capabilities.

    Why it matters NetBox now serves as the vendor-neutral context layer for AI agents in infrastructure operations, with integrations connecting design intent to configuration, monitoring and remediation across multi-vendor environments.

Today's briefing

What happened, and why it matters

14 stories · 8 topics · Updated 11:06 AM ET

NetBox Labs adds NVIDIA and Grafana integrations to platform

NetBox Labs · Oct 13, 2026 · Primary source

NetBox Labs announced partnerships with NVIDIA and Grafana Labs at NetBox Evolve conference, expanding the Infrastructure Intelligence Platform with new integrations and AI-agent capabilities. Arista CloudVision, AWS VPC IPAM and Cisco Nexus Dashboard integrations are generally available; Palo Alto Panorama, Tenable and Kubernetes previews come in October.

Why it matters NetBox now serves as the vendor-neutral context layer for AI agents in infrastructure operations, with integrations connecting design intent to configuration, monitoring and remediation across multi-vendor environments.

NetBox Evolve on October 13, 2026 at Kennedy Space Center marked the 10th anniversary of NetBox as open source. NetBox is used by more than 90% of top neoclouds including Arm, J.P. Morgan Chase and CoreWeave. The platform now connects infrastructure records to configuration tools (Arista, AWS, Cisco), monitoring systems (Grafana), and AI orchestration through NVIDIA partnerships. NetBox Analytics adds ready-made dashboards; NetBox Visual Explorer enables rapid navigation of large-scale datacenters. NetBox Asset Lifecycle carries planned objects through bill of materials, purchase order, shipment and installation workflows. For NetDevOps practitioners, the strategic value is bridging source-of-truth records to both human operators and autonomous agents—connecting what should exist (design) to systems that operate it. The NVIDIA and Grafana integrations signal a direct bet on AI datacenter infrastructure where records feed AI reasoning loops. NetBox Analytics and Visual Explorer target general availability in November 2026.

Read the original at netboxlabs.com ↗

Network Automation Forum opens 2026 automation survey

Network Automation Forum / ipSpace.net · Oct 8, 2026 · Primary source

The Network Automation Forum is running the 2026 State of Network Automation Survey to track practitioner experience with automation tools, frameworks, and infrastructure maturity. Results will inform AutoCon 6 and industry understanding of adoption patterns.

Why it matters Annual benchmark of NetDevOps adoption, source-of-truth platforms, automation frameworks, and AI-assisted operations maturity directly shapes peer-driven content and informs practitioner strategy on tooling and team structure.

The Network Automation Forum's annual survey captures real practitioner experience across NetDevOps, automation tooling, source-of-truth platforms, and infrastructure-as-code maturity. The 2025 survey results informed 2026 planning; this year's survey will benchmark adoption of sources of truth (NetBox, Nautobot, Infrahub), automation frameworks (Ansible, Nornir, intent-based platforms), and emerging patterns around AI-assisted operations and agent-driven workflows. Survey results shape AutoCon conference agendas and provide data on which practices drive real outcomes versus aspirational tooling. For busy practitioners, the lightweight 15-minute survey covers deployment models, team structure, tool adoption, perceived challenges, and infrastructure automation ROI. Results are published and inform both the broader automation community and the Network Automation Forum's advisory board planning for future events.

Read the original at blog.ipspace.net ↗

AI spots backbone warning signs before silent outages

IEEE Transactions on Network and Service Management / TechXplore · Oct 10, 2026 · Research

Researchers from Istanbul Technical University developed an AI method to identify warning signs of silent network failures by monitoring backbone traffic patterns. The approach detects subtle anomalies that precede packet loss, achieving up to 98% detection accuracy on test datasets.

Why it matters Shifts network outage detection from reactive (observing packet loss) to predictive (identifying precursor patterns), enabling NOCs to prevent degradation before users experience impact.

Published in IEEE Transactions on Network and Service Management, the study integrates multiple machine-learning techniques to analyze how backbone network traffic evolves over time. Rather than spotting obvious issues like congestion, it detects warning patterns that resemble normal network activity. The unsupervised model achieved up to 98% detection and approximately 90% F1 score on datasets without confirmed black-hole incidents. However, testing on unlabeled ISP traffic did not establish predictive accuracy for real black holes or prove outage prevention in production. The researchers note that effectiveness varies by network type and structure, and that model retraining may be needed as network infrastructure evolves. Future work should explore explainable AI techniques to help operators understand why specific traffic patterns are flagged as suspicious, broader network type coverage, and automated retraining mechanisms.

Read the original at techxplore.com ↗

LogicMonitor brief covers multi-vendor NOC monitoring

LogicMonitor · Oct 7, 2026 · Vendor release

LogicMonitor released a solution brief for unified network visibility in NOC/ITOps environments, covering hybrid and multi-vendor network monitoring with automated detection and alert correlation capabilities.

Why it matters Addresses core NOC challenges—fragmented dashboards, alert noise, slow incident response—with unified visibility across on-prem, cloud, and hybrid infrastructure; Edwin AI agent automates triage and correlation.

LogicMonitor's October 2026 solution brief highlights how unified network monitoring helps ITOps teams detect issues faster and reduce alert noise in complex hybrid environments. The platform supports 3,000+ out-of-the-box device integrations and provides role-based dashboards for different team stakeholders (NOC analysts, leadership, customers). The solution includes LM Envision for unified visibility and Edwin AI Agent for automated incident triage and runbook execution. Key capabilities cover bandwidth, latency, packet loss, jitter monitoring; hardware degradation detection; and configuration change tracking. The brief emphasizes that modern networks require single-pane-of-glass monitoring across on-premises, cloud, and edge infrastructure rather than isolated tools. For NOC practitioners, the Edwin AI Agent specifically addresses alert fatigue by automatically correlating events from multiple sources, surfacing resolutions from past incidents, and reducing unnecessary escalations. The platform enables baseline anomaly detection that flags degradation before users report issues, and integrates with existing ITSM workflows for faster response.

Read the original at logicmonitor.com ↗

Micas launches three 102.4T Tomahawk 6P Ethernet switches

Globe Newswire · Oct 12, 2026 · Vendor release

Micas Networks announced three 102.4 Tbps open Ethernet switch platforms built on Broadcom Tomahawk 6P, spanning 800G air-cooled, 1.6T liquid-cooled, and 1.6T co-packaged optics variants, with production scheduled for Q4 2026.

Why it matters Multiple vendor implementations of open Ethernet platforms eliminate lock-in and provide migration flexibility from 800G to 1.6T for AI cluster operators.

Micas Networks delivered a portfolio of three systems at OCP 2026: a 4RU air-cooled 800G platform with 128×800G OSFP ports on Tomahawk 6P for high-radix AI fabrics; a 1.6T liquid-cooled variant; and a 1.6T Davisson co-packaged optics system. The company positions these as a practical migration path from 800G to 1.6T while maintaining openness, contrasting with vertically integrated vendors. Systems will be demonstrated at OCP with production beginning Q4 2026. This reflects the broader ecosystem shift: open Ethernet platforms from merchant silicon vendors are now assembled into complete AI fabric solutions by multiple ODMs, creating real vendor choice. For operations teams, this means qualified interop paths and production timelines for 1.6T topologies—not vaporware, but systems entering volume production within weeks.

Read the original at globenewswire.com ↗

Fuji automates SMT placement of Broadcom co-packaged optics switch

Electronics Media · Oct 14, 2026 · Primary source

Fuji Corporation successfully demonstrated automated SMT placement of Broadcom's Tomahawk 6 – Davisson co-packaged optics switch using its NXTR pick-and-place machines, marking the first industry demonstration of production-scale manufacturing for CPO switching silicon.

Why it matters Successful automated assembly of CPO dies removes a critical manufacturing gating factor and signals transition from lab prototypes to volume production timelines.

Fuji's successful SMT placement evaluation of Tomahawk 6 – Davisson demonstrated component transport performance and placement accuracy using automated machinery. As of October 5, 2026, this was the first industry demonstration of Tomahawk 6 – Davisson SMT placement at production scale. This milestone is operationally significant because CPO switches have complex mechanical and thermal requirements; successful automated placement is a gating factor for volume production. The achievement removes a critical manufacturing constraint and validates that Tomahawk 6 – Davisson platforms can transition from limited early access to broad availability. For network operators planning 1.6T deployments, this validates equipment suppliers have solved the hard part of integrating CPO dies into production lines. Expect this to accelerate OEM system qualifications and customer trials through Q4 2026 and into 2027.

Read the original at electronicsmedia.info ↗

AMD AI NIC supports Ultra Ethernet and MetaRoCE transports

Blockchain.News · Oct 11, 2026 · Vendor release

AMD published details on its programmable AI NIC architecture supporting Ultra Ethernet Consortium, Multipath Reliable Connection, and MetaRoCE transports for AI workloads, with Gartner projecting 80% of hyperscalers will choose Ethernet for AI by 2028.

Why it matters A single NIC capable of supporting multiple open Ethernet transports eliminates upfront vendor lock-in and allows operators to trial different cluster approaches without hardware replacement.

AMD describes how large-scale AI training requires fast congestion recovery, path failure handling, packet spray load balancing, and source routing—capabilities traditional RoCEv2 lacked. The Ultra Ethernet Consortium, comprising over 100 members, developed open Ethernet transport specifically for AI and HPC. AMD, as a founding member who authored Network-Signaled Congestion Control now in the UEC specification, delivered one of the first AI NICs supporting UEC. Meta also developed MetaRoCE in 2026 for its own AI training. The ecosystem is converging on open Ethernet but settling on three competing transport protocols (UEC, MetaRoCE, and legacy RoCEv2), each with different performance profiles and tuning requirements. A single programmable NIC supporting all three eliminates the need to choose upfront and allows operators to trial multiple approaches or support different cluster cohorts. This flexibility was unavailable 12 months ago and signals maturation of the open Ethernet ecosystem for production AI workloads.

Read the original at blockchain.news ↗

Strike and BGP collapse cascade through Russian networks

debuglies.com · Oct 9, 2026 · Analysis

A UAV strike on Yandex's Sasovo data center combined with routing collapse in Qrator Labs' BGP filtering network (AS197068 and AS200350) caused cascading outages across Russian infrastructure on October 8, 2026. BGP route announcement churn spiked from 12 updates/hour baseline to 4,820 updates/hour, exposing how domestic consolidation mandates eliminated redundancy.

Why it matters Demonstrates how concentrated infrastructure and BGP routing vulnerabilities compound during physical disruptions—critical for designing resilient multi-vendor, geographically diverse routing architectures.

The incident was driven by concurrent failures: kinetic physical disruption from a UAV strike on Yandex's flagship hyper-scale data center in Sasovo (Ryazan Oblast), combined with architectural routing collapse within Qrator Labs, the dominant BGP filtering and DDoS mitigation network in Russia. This cascading blackout paralyzed banking, media, and state administration, exposing how Russia's digital autarky doctrine—mandating domestic consolidation for surveillance and control—eliminated redundant architecture and turned localized strikes into nationwide systemic failure. BGP routing across Qrator Labs' autonomous systems destabilized, with route announcement churn spiking from below 12 updates per hour to 4,820 updates per hour beginning at 06:18 MSK on October 8. The analysis reveals that import substitution in cloud computing and cyber defense creates acute, unhedged operational fragility—a critical lesson for practitioners designing carrier-grade networks that depend on third-party routing services.

Read the original at debuglies.com ↗

Ivanti Connect Secure zero-day exploited for unauthenticated root access

BleepingComputer · Oct 14, 2026 · Industry news

CISA urged immediate patching of CVE-2026-104935, an Ivanti Connect Secure authentication bypass exploited in zero-day attacks since October 1st that lets remote threat actors execute arbitrary commands without credentials. Threat actor UTA0576 chains this flaw with two other vulnerabilities to achieve unauthenticated root-level code execution.

Why it matters Network ops teams managing remote-access VPN infrastructure must immediately patch Ivanti Connect Secure versions 25.1.1.1 and prior; CVE-2026-104935 is on the CISA KEV catalog with a October 16 federal deadline and active exploitation in production environments.

CISA disclosed CVE-2026-104935 on October 13, 2026, a high-severity authentication bypass in Ivanti Connect Secure actively exploited since at least October 1st, allowing remote attackers without credentials to execute arbitrary commands on targeted systems in low-complexity attacks requiring no user interaction. Volexity identified threat actor UTA0576 chaining this flaw with two other vulnerabilities to achieve root-level code execution, discovering post-exploitation actions targeting Active Directory credentials including LAPS passwords and BitLocker recovery keys. Critical for SOC teams: Ivanti's Integrity Checker Tool may not detect exploitation because threat actor commands execute in memory, requiring implementation of forensic procedures from Ivanti's recovery documentation and review of the Volexity attack chain blog. Federal agencies were required to remediate by October 16, establishing urgency for private sector network defenders.

Read the original at bleepingcomputer.com ↗

Cisco patches 35 flaws including critical NX-OS bugs

The420.in · Oct 9, 2026 · Industry news

Cisco released fixes for 35 vulnerabilities in October 2026, including critical flaws in NX-OS, Nexus switches, Meraki, APIC and License On-Prem enabling remote attacks. Some vulnerabilities were identified using frontier AI models alongside existing security-testing processes.

Why it matters Network engineers managing Cisco fabric infrastructure must track that 14 critical NX-OS vulnerabilities affect switches and data-centre networking products; no active exploitation reported but hardening patches require urgent deployment across Nexus platforms.

Cisco published security updates on October 7, 2026, addressing 35 vulnerabilities across widely used networking and enterprise products, with more than a dozen rated critical. A significant portion of critical vulnerabilities affect Cisco NX-OS, the operating system used across several Cisco switches and data-centre networking products. Cisco's October release reflects a methodological shift in vulnerability discovery: some flaws uncovered during internal security reviews were identified using frontier AI models alongside existing testing processes, marking early adoption of AI-assisted vulnerability detection in routine product hardening. Cisco stated it is not aware of malicious exploitation involving the October 7 advisories, distinguishing this as a proactive hardening release rather than an emergency response to confirmed zero-day attacks.

Read the original at the420.in ↗

Citrix discloses critical NetScaler SAML memory overflow flaw

CybersecurityDive · Oct 12, 2026 · Industry news

Citrix disclosed CVE-2026-107406, a critical memory overflow vulnerability (CVSS 9.5) affecting NetScaler ADC and Gateway appliances configured as SAML Service Provider or Identity Provider, enabling remote code execution or denial of service. Citrix was not aware of active exploitation but urged immediate patching.

Why it matters Network defenders managing NetScaler perimeter appliances must apply urgent patches, particularly for SAML-configured instances; this follows recent waves of critical Citrix flaws with over 20,000 instances estimated at risk across the attack surface.

Citrix disclosed CVE-2026-107406 on October 10, 2026 (published October 12), a critical memory overflow vulnerability scoring 9.5 CVSS affecting NetScaler ADC and Gateway appliances used for remote access and application delivery. The flaw requires NetScaler to be configured as a SAML Service Provider or SAML Identity Provider; Citrix was not aware of exploitation but strongly urged customers to immediately upgrade to recommended versions. Security researcher analysis from Rapid7 noted that while CVE-2026-107406 is concerning, broad successful exploitation across unpatched appliances is less likely with memory-corruption flaws compared to simpler authentication-bypass vulnerabilities. This disclosure follows recent waves of critical Citrix flaws; Shadowserver Foundation estimated over 20,000 NetScaler instances were visible and potentially at risk during prior vulnerability incidents, underscoring the perimeter risk for federated identity configurations.

Read the original at cybersecuritydive.com ↗

SolarWinds Avior pairs AI reasoning with deterministic workflows

Network World · Oct 14, 2026 · Vendor release

SolarWinds launched Avior, a standalone agentic AI system that pairs probabilistic AI reasoning with deterministic execution for enterprise IT operations. The system grounds model decisions in fixed workflows with predefined goals and approval controls, ensuring auditability and reliability in production environments.

Why it matters Avior brings MCP-integrated agents to network operations with out-of-the-box capabilities for alert correlation, configuration drift detection, compliance auditing, and incident remediation—practical agentic patterns that practitioners can adopt immediately across hybrid IT.

SolarWinds' Avior addresses a critical operational problem: AI agents producing variable outcomes on production systems. Rather than functioning as a chat interface, Avior operates as a standalone application pursuing defined operational objectives across networks, systems, databases, and cloud workloads while keeping risky changes behind human approval gates. The system reaches third-party tools through MCP, requiring no data migration. Initial capabilities include correlating alerts into actionable incidents, identifying configuration drift against compliance policies, managing configuration backups, tracking capacity, monitoring utilization, and detecting obsolete assets. The design encodes 25 years of SolarWinds operational experience into predefined goals and approval frameworks. Avior becomes generally available October 14, 2026, through guided onboarding.

Read the original at networkworld.com ↗

RFC 10065 standardises YANG model for network access control

RFC Editor / IETF · Oct 7, 2026 · Primary source

RFC 10065 published with YANG data model and RADIUS extension for policy-based network access control. Provides standardized configuration and authentication framework for network access policies, critical for ops automation.

Why it matters Operators can now leverage standardized YANG models for consistent policy configuration across heterogeneous infrastructure, reducing manual NAC configuration errors and enabling policy versioning and change tracking.

RFC 10065 was published on October 7, 2026, specifying a YANG data model and RADIUS extension for policy-based network access control. This standard moves network access control from proprietary vendor implementations to a unified data model, enabling operators to version, test, and audit access policies as code. The YANG model integrates directly with NETCONF/YANG-based configuration management tools already in use across enterprise and service provider networks, reducing operational friction in NAC deployments. RADIUS extension support ensures backward compatibility with existing AAA infrastructure while enabling modern, policy-driven access control. For NetDevOps teams, access policies can now flow through CI/CD pipelines, be validated against compliance requirements, and rolled back safely on failure.

Read the original at rfc-editor.org ↗

AI governance spending to reach $3.15 billion in 2026

EINPresswire · Oct 14, 2026 · Vendor release

The enterprise AI governance and compliance market is projected to grow from $2.44 billion in 2025 to $3.15 billion in 2026, reflecting 29% compound annual growth as organizations prioritize safety, transparency, and regulatory alignment in AI system management.

Why it matters Governance infrastructure is now a critical enterprise purchase category; practitioners must treat compliance tooling, observability, and risk assessment as table-stakes for scaling agentic AI deployments.

The 29% annual growth in enterprise AI governance significantly outpaces general IT spending, driven by two converging pressures: EU AI Act full enforcement as of August 2, 2026, and operational gaps in autonomous agent oversight. Market data showed 74% adoption but only 47% governance controls, with 86% of organizations experiencing at least one AI-related incident in the past year. This disconnect is driving a projected 25% average increase in AI governance technology budgets, as organizations recognize that deploying agents without governance visibility creates unacceptable operational and compliance risk. Vendors are building governance as a service layer, creating new integration points and dependencies in the AI operations stack that practitioners must evaluate.

Read the original at einpresswire.com ↗
Nothing in today's briefing matches that.

Listening

Podcasts and talks
  • Packet Pushers - Heavy Networking

    HN845: Real Life Use Cases for BGP Monitoring Protocol (BMP)

    Bart Dorlandt discusses practical BGP Monitoring Protocol implementations for service providers, covering data collection architecture, pre/post-policy filtering capture, and handling excessive prefix volumes in production environments.

  • Packet Pushers - Day Two DevOps

    D2DO315: Running Agents Securely at Scale

    Ned and Kyler explore challenges and best practices for running AI agents securely at scale with guest Michael Levan, discussing specialized AI gateways, semantic routing, sandboxing, and robust observability for agentic workloads.

  • NANOG

    NANOG 98 Conference: Global Connectivity Focus with Keynote on BGP Routing Geography and Geopolitical Challenges

    NANOG 98 in Miami (October 19-21, 2026) focuses on global connectivity with keynote by Paul Brodsky on internet routing geography shifts, subsea cable investment, and geopolitical vulnerabilities including South Red Sea maritime blockages.

Vendor Radar

Last 7 days · arrows compare with the 7 before

What changed this week

Last 7 days vs the 7 before

Biggest moves

Trending topics

agent · automation · MCP · observability · OpenTelemetry · Agentic AI · SRE · RAG · Kubernetes · inference

Get the daily briefing

Today's 3 things that matter and every story with why it matters, in your inbox each morning. Free, and you can unsubscribe at any time. Prefer a reader? Follow the RSS feed.