You're reading the Tuesday, November 3, 2026 edition. Today's briefing →
Toronto
Tuesday, November 3, 2026No. 142

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·Primary source

    Attackers exploit PHP flaws on exposed PAN-OS management interfaces

    Two critical PHP vulnerabilities (CVE-2026-109618, CVE-2026-109619, CVSS 9.2 each) allow unauthenticated remote code execution on PHP systems.

    Why it matters Network appliances with exposed management interfaces are primary targets; PHP2Shell enables unauthenticated RCE directly on firewalls and edge devices. Restrict management interface access immediately, scan for exposure, and apply patches across all affected products (Palo Alto, Fortinet, Cisco, GitLab).

  2. Routing·Primary source

    Cloudflare cuts DNS zone build memory use by 86%

    Cloudflare built a targeted DNS Zone Builder that cuts full zone rebuilds by 40%, reduces peak memory from 79 GiB to 7 GiB (86% reduction), and halves median build completion delay from 7.7 to 3.3 seconds by determining the smallest safe rebuild rather than always rebuilding entire zones.

    Why it matters Large-platform performance gains now come from classifying internal changes precisely to skip expensive work without serving stale DNS data; the 85% peak-memory cut and halved median build delay represent capacity and tail-latency gains inside Cloudflare's infrastructure that operators should understand when evaluating DNS platform scalability.

  3. Startups·Industry news

    CrowdStrike buys Fiddler AI and Dreadnode for $495 million

    CrowdStrike signed definitive agreements to acquire Fiddler AI (AI agent observability, ~$400M) and Dreadnode (automated AI security testing, ~$95M) to integrate into its Falcon platform.

    Why it matters Consolidates agent observability and pre-deployment security testing into a single platform; signals that AI agent governance is now table-stakes for security infrastructure teams managing agentic deployments.

Today's briefing

What happened, and why it matters

13 stories · 9 topics · Updated 12:48 PM ET

Oracle adds multiplanar RDMA fabrics for Vera Rubin and TPUs

techblog.comsoc.org · Oct 27, 2026 · Primary source

Oracle announced Acceleron networking improvements alongside NVIDIA Vera Rubin and Google TPU architecture for AI cloud deployment. New capabilities include RDMA-based fabrics, resilient multiplanar networking, and distributed collective communications for moving data across large clusters.

Why it matters Oracle's multiplanar RDMA design and collective communications library provide alternatives to single-plane architectures for large distributed training, affecting how practitioners architect fault-tolerant AI infrastructure at hyperscale.

Oracle unveiled Acceleron AI networking advances at AI World 2026 on October 27, spanning three elements: NVIDIA Vera Rubin accelerators, Google TPU architecture, and Oracle's high-performance cloud networking. The company reported adding 850 MW of data center capacity and deploying more than 300,000 GPUs in fiscal Q1 2027. Oracle Acceleron combines dedicated network fabrics, programmable networking, direct data paths and host-level acceleration. New capabilities include the Oracle Collective Communications Library, which helps customers keep distributed AI workloads running when network paths fail by spreading collective traffic across OCI's backend Multiplanar Network. Transparent Front-End Multiplanar Networking brings multiplanar networking to front-end networks with built-in transparent path selection, failover, and transparent congestion control for front-end RDMA traffic. NVIDIA and Oracle announced plans to expand OCI Superclusters to multi-gigawatt scale using NVIDIA Vera Rubin NVL72 rack-scale systems. This represents a significant shift from single-plane RDMA architectures toward active multiplanar redundancy in production AI infrastructure, with implications for how practitioners design collective operations and handle network fault isolation.

Read the original at techblog.comsoc.org ↗

CoreWeave builds 240 MW InfiniBand AI cloud in India

Yahoo Finance · Oct 28, 2026 · Industry news

CoreWeave announced a 240 MW AI data center expansion into India in October 2026, powered by NVIDIA Blackwell systems connected via NVIDIA Quantum-X 800 InfiniBand networking. New points of presence planned across Asia-Pacific and UAE with full loop expected by mid-2027.

Why it matters Neocloud fabric procurement patterns now show explicit hedging: CoreWeave uses Quantum-2 NDR InfiniBand for training workloads and Spectrum-X RoCE for GB300 deployments, reflecting how providers choose between proprietary and open Ethernet at different GPU generations.

CoreWeave's planned network expansion targets Asia-Pacific and the Middle East, with new data center points of presence in United Arab Emirates, India, Indonesia, Singapore, Australia, South Korea and Japan. The rollout is scheduled to begin in Q4 2026, with the network expected to form a full loop by mid-2027, extending its existing North America and Europe infrastructure to four continents. The India facility is powered by NVIDIA Blackwell systems with NVIDIA Quantum-X 800 InfiniBand networking. This expansion signals neocloud momentum in deploying high-speed InfiniBand fabrics at scale across geographies. For practitioners assessing neocloud capacity for dedicated training workloads, this reflects continued heavy reliance on NVIDIA's integrated fabric strategy (Quantum-X for proven InfiniBand, Spectrum-X RoCE for newer GPU generations) rather than waiting for Ultra Ethernet Consortium standardization to mature.

Read the original at finance.yahoo.com ↗

Cloudflare cuts DNS zone build memory use by 86%

Cloudflare Blog · Nov 2, 2026 · Primary source

Cloudflare built a targeted DNS Zone Builder that cuts full zone rebuilds by 40%, reduces peak memory from 79 GiB to 7 GiB (86% reduction), and halves median build completion delay from 7.7 to 3.3 seconds by determining the smallest safe rebuild rather than always rebuilding entire zones.

Why it matters Large-platform performance gains now come from classifying internal changes precisely to skip expensive work without serving stale DNS data; the 85% peak-memory cut and halved median build delay represent capacity and tail-latency gains inside Cloudflare's infrastructure that operators should understand when evaluating DNS platform scalability.

Cloudflare rebuilt its DNS Zone Builder with a targeted build planner and unified scheduler to rewrite only the parts of a zone that changed, instead of rebuilding entire zones for every setting change. Full builds dropped by about 40%, with 92.95% of builds now avoiding a complete zone rebuild over a 24-hour window. Peak working-set memory fell from 79 GiB to 12 GiB, an 85% reduction, and the largest p99 build delay spike fell from 14.6 minutes to 1.21 minutes. The extra headroom let Cloudflare shorten internal wait times, halving median build completion delay from about 7.7 seconds to 3.3 seconds. The DNS pipeline now handles nearly three times the changes it did in 2022. Full builds still happen for first-time zone publication or when an event's effect is unclear, but these make up only 7.04% of builds and average just 0.35 seconds. This work addresses scalability at the DNS authoritative side—relevant to operators running large zone counts or evaluating DNS infrastructure partners for performance and resource efficiency.

Read the original at blog.cloudflare.com ↗

Ivanti zero-days chain to unauthenticated root, added to KEV

Rapid7 / CISA / SecurityWeek · Oct 30, 2026 · Primary source

CISA added CVE-2026-111357 (CVSS 8.8, privilege escalation) and CVE-2026-118527 (CVSS 4.3, predictable file upload) to KEV on October 30; both chain with CVE-2026-104935 to enable unauthenticated root-level code execution on Ivanti Connect Secure. Threat group UTA0576 exploited the full chain beginning October 3, 2026.

Why it matters Remote-access gateway compromise chains allow attackers direct enterprise core access; organizations must immediately upgrade all Connect Secure appliances to patched versions (25.1.1.3 or 22.8R2.6) and rotate credentials before federal deadline of November 2–13, 2026.

On October 29, 2026, Ivanti updated its security advisory for Ivanti Connect Secure (ICS) to include two additional vulnerabilities and released fixes for both. On October 30, 2026, CVE-2026-111357 and CVE-2026-118527 were added to the U.S. KEV catalog. In its investigation, Volexity reports that UTA0576 used an exploit chain to gain unauthenticated root access to an appliance running 22.8R2.4, beginning on October 3, 2026. CVE-2026-111357 has a CVSS severity score of 8.8 and results from missing authentication for a critical function. An authenticated attacker with local access could exploit it to gain higher privileges. CVE-2026-118527 involves predictable identifier generation. Affected: Connect Secure 25.1.1.2 and prior, 25.1.2.2 and prior, and 22.8R2.5 and prior. The exploit chain: skip the login with CVE-2026-104935, upload a file to a guessable path with CVE-2026-118527, run code as the web user, escalate to root with CVE-2026-111357, then read stored domain credentials. CISA gave Federal Civilian Executive Branch agencies until November 2 to fix CVE-2026-111357, three days after the October 30 listing, and until November 13 for CVE-2026-118527.

Read the original at rapid7.com ↗

GitLab RCE and Cisco firewall DoS flaws exploited in wild

CISA / The Hacker News / Cisco · Oct 30, 2026 · Primary source

CISA added CVE-2026-97229 (CVSS 10.0, GitLab RCE via integer overflow in CI/CD regex validation) and CVE-2026-76418 (CVSS 8.6, Cisco ASA/FTD null pointer dereference DoS) to KEV on October 30. GitLab flaw allows unauthenticated code execution; an attacker probed Alpine Linux's GitLab on October 20–21.

Why it matters Perfect-score GitLab RCE threatens any developer pipeline using self-managed instances; Cisco firewall DoS flaws allow attackers to disable perimeter security mid-attack. Both require immediate patching by November 1, 2026 (federal deadline).

CVE-2026-97229 (CVSS score: 10.0) is an integer overflow or wraparound vulnerability in GitLab Community Edition and Enterprise Edition that could allow an unauthenticated user to execute arbitrary code on the GitLab server via a specially crafted regular expression in a CI/CD configuration due to insufficient regular expression validation. Alpine Linux reported that an unknown attacker tried to exploit the vulnerability on October 20 and 21, 2026, but the attempts failed. CVE-2026-76418 scores 8.6 and affects Cisco's Secure Firewall ASA and FTD software—the appliances many organizations use as their front door to the internet. A flaw in how the software handles a return value can let an unauthenticated remote attacker force the device to reload unexpectedly, allowing denial of service. Affected releases include GitLab 19.2 versions before 19.2.8, 19.3 versions before 19.3.5, and 19.4 versions before 19.4.2. CISA orders federal agencies to fix both vulnerabilities by November 1st, 2026.

Read the original at thehackernews.com ↗

Attackers exploit PHP flaws on exposed PAN-OS management interfaces

CISA / Palo Alto Networks / BleepingComputer · Oct 27, 2026 · Primary source

Two critical PHP vulnerabilities (CVE-2026-109618, CVE-2026-109619, CVSS 9.2 each) allow unauthenticated remote code execution on PHP systems. Palo Alto reported active exploitation targeting internet-exposed PAN-OS management interfaces on October 29; threat actors extract credentials and enumerate Active Directory after gaining RCE.

Why it matters Network appliances with exposed management interfaces are primary targets; PHP2Shell enables unauthenticated RCE directly on firewalls and edge devices. Restrict management interface access immediately, scan for exposure, and apply patches across all affected products (Palo Alto, Fortinet, Cisco, GitLab).

Two critical vulnerabilities, CVE-2026-109618 and CVE-2026-109619, known as PHP2Shell, allow unauthenticated remote code execution on PHP systems serving web requests. Threat actors without privileges can exploit these in attacks to gain RCE on unpatched systems. Palo Alto Networks warned that threat actors are exploiting the recently patched PHP2Shell vulnerabilities against publicly-accessible management interfaces. Active exploitation has been observed targeting the Palo Alto Networks management interface exposed to the internet. As of October 30, 2026, patches have been released for all vulnerable versions of PAN-OS 12.2, PAN-OS 12.1, PAN-OS 11.2, PAN-OS 11.1, and PAN-OS 10.2. The affected devices run specific versions of firewall software; the attacker then launched attacks against Active Directory. Recovered artifacts confirm credential extraction and directory enumeration.

Read the original at thehackernews.com ↗

CrowdStrike buys Fiddler AI and Dreadnode for $495 million

SecurityWeek · Nov 3, 2026 · Industry news

CrowdStrike signed definitive agreements to acquire Fiddler AI (AI agent observability, ~$400M) and Dreadnode (automated AI security testing, ~$95M) to integrate into its Falcon platform. Both deals expected to close in Q4 FY27.

Why it matters Consolidates agent observability and pre-deployment security testing into a single platform; signals that AI agent governance is now table-stakes for security infrastructure teams managing agentic deployments.

CrowdStrike announced two separate acquisitions on November 3, 2026, at its Fal.Con Europe conference. Fiddler AI, based in Palo Alto, specializes in explaining what AI agents do and why—providing post-deployment traceability and compliance visibility. The technology will become a product called Falcon Observe, expected in Q4 FY27. Dreadnode, founded in 2023 and headquartered in Austin, provides automated AI security testing that runs agents and models through 100+ known attack strategies including jailbreaks and prompt injection, mapping to OWASP, MITRE ATLAS, and NIST frameworks. CrowdStrike did not publicly disclose pricing; the Wall Street Journal reported a combined $495M valuation ($400M for Fiddler, $95M for Dreadnode). Consideration will be primarily cash with some stock subject to vesting. For practitioners managing AI deployments in regulated environments, this consolidation means native integration of agent observability with endpoint security—eliminating the "the model did it" defense and enabling auditable agent execution trails. Both closures are subject to customary conditions.

Read the original at securityweek.com ↗

Quantum eMotion names Wayne Teeple chief technology officer

Quantum eMotion press release · Oct 27, 2026 · Primary source

Quantum eMotion appointed David Wayne Teeple as Chief Technology Officer, a former board member since 2021 with over 30 years of cybersecurity and technology leadership experience. Teeple will oversee technology strategy, product development, engineering coordination, and technical integration of quantum-safe cryptography and identity security platforms.

Why it matters Establishes experienced CTO leadership for quantum-safe cryptography product commercialization and technical integration of planned Plurilock Security acquisition; directly impacts security infrastructure modernization roadmaps.

Quantum eMotion Corp., a quantum-safe cybersecurity solutions developer, appointed David Wayne Teeple as Chief Technology Officer effective October 27, 2026. Teeple, who served on the company's board since 2021, brings over 30 years in cybersecurity, telecommunications and IT. He is the founder and former CEO of Phirelight Security Solutions Inc., a Canadian cybersecurity firm acquired by RHEA Group in 2018, and previously served as a Communications and Electronics Officer in the Canadian Armed Forces. As CTO, Teeple will lead technology strategy, product development roadmap, engineering coordination and technical integration, with specific focus on advancing QRNG technologies, SecureKey cryptographic solutions, and Sentry-Q and eShield-Q platforms. His remit includes technical integration responsibilities tied to Quantum eMotion's proposed acquisition of Plurilock Security Inc., which would combine quantum-safe cryptography with identity security and AI-driven cybersecurity. Teeple received 500,000 stock options at $1.47 per share exercise price, with 200,000 contingent on achieving specified performance milestones.

Read the original at investornews.com ↗

Network operations agents move from pilots to pricing debates

Cybersecurity Institute · Nov 1, 2026 · Analysis

Market shifts from whether agents will run networks to how they will be paid for; Gartner predicts half of organizations will run IT operations agents with minimal human involvement by 2030, with HPE folding SD-WAN/SASE into Mist and Aruba Central.

Why it matters NetOps teams must now address token-metered pricing models, agent autonomy governance, and the NOC/SOC convergence as agentic operations become operational standard rather than experimental.

Gartner forecasts that by 2030, half of organizations will run IT operations agents with minimal human involvement, up from almost none today. The operational focus has shifted from capability to economics: buyers on token-metered agent pricing will overrun budgets by more than half if not carefully managed. HPE folded EdgeConnect SD-WAN and SASE into Mist and Aruba Central, placing branch security in front of the network operations team. Itential and Gluware pitch autonomy earned one task at a time, while Forward Networks extended its network model into security exposure. Cisco is using its channel to bring agentic operations into its installed base, with partners compensated to run Cisco's console as managed service and build agents on Cisco's own models. Cisco's MCP servers enable outside agents to read live network data, and ThousandEyes testing is included in eligible access points. Privileged Infrastructure Access (built with Teleport) gives people and agents device access that expires when the task ends.

Read the original at cybersecurityinstitute.com ↗

Verizon CTO says 6G must be autonomous from day one

RCR Wireless · Oct 27, 2026 · Industry news

Verizon CTO argues 6G standards must embed autonomous network operations and AI-driven use cases from the outset, rather than bolting them on later. Verizon executed 70 million closed-loop automation processes in 2025 and is targeting Level 4 autonomy, requiring architectural shifts to handle AI-native uplink demand.

Why it matters Early 6G design decisions now will determine whether autonomous network operations are native or retrofitted—a critical inflection for NetDevOps and AIOps practitioners planning network evolution strategies.

Verizon CTO Yago Tenorio says 6G needs an architectural rethink to handle AI-driven uplink demand, avoid repeating 5G's standardization mistakes, and embed autonomous network operations and commercial use cases from the outset. This isn't a 2030 problem—it's a standards-shaping problem now. Verizon ran 70 million closed-loop automation processes in 2025. The operator is already demonstrating Level 4 autonomy (where AI identifies anomalies, determines root causes, and executes corrective actions) and wants that operational model baked into 6G specs rather than requiring teams to stitch vendor-specific autonomous platforms together post-launch. Verizon wants use cases such as ISAC to shape the standard, while embedding network autonomy from launch rather than bolting it on later. For operators planning RAN transformation, this signals the industry direction: autonomy-first architecture is becoming table stakes for next-generation networks.

Read the original at rcrwireless.com ↗

Bell plans 1.2 GW AI data centre hub in Saskatchewan

TechSabado · Oct 28, 2026 · Industry news

The 300-megawatt Sherwood campus is the first phase of Bell AI Fabric's growing presence in Saskatchewan. Bell and the Government of Saskatchewan signed an MOU creating a path to a potential 1.2-gigawatt Canadian AI infrastructure hub in the province. Bell plans to have at least two buildings at the facility up and running by the first half of 2027.

Why it matters Telco-native sovereign AI infrastructure with closed-loop cooling and national fiber backbone is entering production. Practitioners evaluating edge compute and enterprise AI workload placement now have a domestic Canadian alternative in the grid timeline.

Total capital investment associated with Bell AI Fabric's Saskatchewan presence, including data centre infrastructure, customer compute hardware and related power generation, could exceed $50 billion, the largest capital investment in Saskatchewan's history. Consistent with all Bell AI Fabric projects, the proposed facilities will use closed-loop cooling technology that requires no municipal water. Bell plans to use its national fiber backbone to connect AI computing infrastructure with enterprise customers. The company has also introduced AI-powered network operations using Google Cloud to improve fault detection and incident resolution. For enterprise workload placement and telecommunications operations, this creates a sovereign, water-efficient infrastructure path tied to a telecom backbone—relevant for organizations with data residency requirements or seeking a Canadian-operated alternative to US-headquartered cloud providers.

Read the original at techsabado.com ↗

Google releases Gemini 4 Carbon for vulnerability finding and patching

Mind and Machine Weekly · Nov 1, 2026 · Industry news

Google launched Gemini 4 Carbon, built to find software vulnerabilities and write patches, restricting early access to vetted cyber defenders through its Fairwind programme and reporting it leads external cyber benchmarks. OpenAI published a revised Model Spec reorganized around faithfulness, harm minimization, honesty and transparency, with new guidance on context and memory.

Why it matters Gemini 4 Carbon's specialized vulnerability-detection capability released two weeks after Gemini 4 Argon signals rapid iteration on frontier model families; teams evaluating large models for security tooling now have a purpose-built option to benchmark.

Google unveiled Gemini 4 Carbon, a frontier model purpose-built to find software vulnerabilities, test whether they can be exploited and write patches, arriving just two weeks after Gemini 4 Argon. Google reports it leads external cyber benchmarks and tops the Artificial Analysis Intelligence Index v5 at 59, with early access restricted to vetted cyber defenders through the Fairwind programme. The rapid release cadence—two distinct frontier models in two weeks—reflects Google's strategy of specializing models for specific high-value workloads rather than single general-purpose frontiers.

OpenAI published a revised Model Spec, the public document describing how it wants its models to behave, reorganized around faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care, adding guidance on context, memory and personalization while retiring detailed style rules. Anthropic widened its AI for Science programme beyond individual researchers to 20 scientific foundations, institutions and research projects, distributing free Claude credits to scientists. These updates reflect continued iteration on model behavior and access policies as labs refine their public commitments and developer partnerships.

Read the original at mindandmachineweekly.substack.com ↗

Salesforce acquires Cresta for $2.7 billion voice AI agents

CMSWire · Nov 2, 2026 · Industry news

Salesforce signed a definitive agreement to acquire Cresta for a reported $2.7 billion to add voice-first AI agents and real-time conversation intelligence to Service Cloud. Cresta's autonomous agents handle complex customer calls end-to-end while also providing real-time guidance to human representatives.

Why it matters Customer service automation is shifting from text chat to live voice; this acquisition fills a critical gap in Salesforce's Agentforce portfolio and signals enterprise demand for production-grade voice agents with human fallback.

Salesforce announced the definitive agreement on November 2, 2026. Cresta was founded at Stanford in 2017 and reported $100M ARR in 2026 with deployments at Marriott, Comcast, and Alaska Airlines. At $2.7B against Cresta's $100M ARR, the valuation marks a pricing benchmark for contact center AI startups at enterprise scale. This is Salesforce's fifth customer experience acquisition in 2026, and the deal addresses a capability gap: organic development of low-latency, enterprise-ready voice agents is difficult and time-consuming. Transaction expected to close in Q1 FY28 (Feb-April 2027). Cresta will continue supporting existing contact center and telephony platforms post-acquisition.

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

Listening

Podcasts and talks
  • Packet Pushers

    NAN135: Unpacking the NetOps Stack

    Eric Chou interviews Mauricio Rojas, Principal DevOps Consulting Engineer at Nokia, discussing how Kubernetes, GitOps, and open-source tooling fundamentally transform modern networks via the NetOpsKube project and Broadband Network Gateways.

  • Buttondown (Network Operations newsletter)

    NANOG 98 and AutoCon 6: AI and automation dominate network operations agenda

    NANOG 98 (October 19–21 in Miami) showcased AI as central to network operations with sessions on AI networking, automation, routing security, and AI datacenter design. AutoCon 6 (November 16–20 in Tucson) gathers practitioners for deep-dive talks on production automation, tooling, and infrastructure operations.

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 · MCP · automation · AIOps · RAG · observability · Agentic AI · OpenTelemetry · eBPF · SASE · fine-tuning

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.