The operations stories enterprise IT leaders actually read.
Volume 342, Issue 1 · 20 articles from 4 issues
Solid-State Transformers Power Next-Gen AI Data Centers
Solid-state transformers are emerging as critical infrastructure for next-generation AI data centers, converting grid AC directly to 800V DC distribution.
Unlike traditional iron-core transformers, SSTs use high-frequency magnetics with silicon carbide semiconductors for improved efficiency, compact form factors, and bidirectional power flow.
While companies like Delta report deployments achieving 98.5% efficiency, industry skeptics including SPOC Energy and Hitachi Energy caution that the technology still lacks full maturity for broad rollout, citing supply-chain depth and manufacturing challenges.
Analysts expect phased adoption starting in 2027, with scaled SST deployments becoming more common among hyperscalers and cloud providers.
AI infrastructure demands require evaluating data centers beyond traditional metrics, including power, cooling, sustainability, and community risks.
Traditional data center procurement criteria no longer suffice for GPU-intensive AI workloads, which demand extreme power density, advanced cooling, and low-latency interconnects.
Businesses must assess technical capabilities across power density, liquid cooling architecture, network performance, and scalability alongside sustainability factors like water consumption and energy sources.
Environmental impact assessments, regulatory compliance readiness, and community opposition represent critical long-term risks that can delay projects by billions in investment. Companies should distinguish between secured and planned capacity while evaluating provider transparency on emissions and resource usage.
A guide to structuring SOC incident handling as managed cases from intake through closure, augmented by agentic AI.
The article lays out how security case management gives SOC teams structured incident handling from alert intake through final closure, with each incident consolidating the alert, evidence, ownership, decisions, review points, handling steps, and outcome into one operational record.
It walks through the full lifecycle: intake, enrichment, triage, investigation, escalation, response, and closure.
Swimlane Turbine's agentic AI is presented as a layer that analyzes context, generates assessment plans, and coordinates remediation across tools while keeping humans in the loop.
SOC leaders gain visibility into investigation quality, team ownership, automation controls, and operational bottlenecks. Comprehensive case documentation also underpins executive reporting.
Vicarius survey finds 79% of organizations were breached via vulnerabilities already sitting in their own inventory.
A Vicarius survey of 300 IT and cybersecurity leaders exposes a sharp gap between finding vulnerabilities and actually fixing them.
Scanning tools surface weaknesses effectively, but 58% of remediation still requires direct human intervention and 82% of organizations cannot consistently remediate within the team that discovered the issue.
The research found 79% of organizations suffered security incidents involving vulnerabilities already in their inventory, with about half of those known for 30 to 90 days.
Organizations that require a verified rescan before closing a vulnerability experienced breaches at rates 40% lower than those using looser definitions of fixed.
The finding points to verification rigor as a meaningful lever on breach risk.
AI workloads are pushing traditional datacentre management to its limits. Modern DCIM is now essential, using predictive analytics to ensure resilience, efficiency, and sustainabilit
Organisations worldwide are racing to deploy artificial intelligence (AI) workloads, but there’s a catch.
These advanced applications are placing unprecedented demands on datacentre infrastructure. The immense power consumption and cooling requirements of AI-capable servers are pushing traditional management tools to their limits and create an urgent need for infrastructure and operations (I&O) leaders to develop a more intelligent and resilient approach.
This is where modern datacentre infrastructure management (DCIM) tools matter more than ever.
David Cannon argues the valuable work is now orchestrating technology wherever it sits, not running a service desk.
David Cannon contends that service management has spread well beyond IT, with technology now bought and run across the organisation by people outside the IT department.
In his framing, the valuable work has shifted from running the IT service desk to orchestrating and governing technology wherever it sits. Specialised tools have proliferated across functions without anyone integrating them, leaving gaps that no single team owns.
The article argues organisations need to recognise this shift explicitly and assign clear responsibility for connecting disparate systems and teams instead of clinging to IT-centric operating models.
Reviews 19 AgentOps platforms for observing AI agent performance, cost, and failure modes in production.
Agent observability, or AgentOps, has become essential as enterprises push AI agents and LLMs into production.
The article explains that these tools track agent performance, resource consumption, spend, and failure modes such as hallucinations and model drift.
Many products adapt familiar DevOps monitoring patterns to AI-specific problems like non-deterministic LLM behavior and the need for detailed prompt logging.
It profiles 19 platforms spanning purpose-built entrants such as AgentOps.ai and Braintrust alongside established observability vendors like Datadog and Dynatrace. Each is framed around a different organizational priority, from rapid iteration to enterprise-scale stability and cost control.
Gartner's ITSM Magic Quadrant returns after four years with sixteen vendors and major position shifts.
Stephen Mann compares Gartner's 2026 Magic Quadrant for ITSM platforms against the 2022 edition after a four-year gap.
The evaluated field expanded from ten vendors to sixteen, with seven appearing in the quadrant for the first time. Freshworks climbed from Challenger to Leader while Ivanti dropped into Niche Players, changes Mann reads as a proxy for market momentum rather than a verdict on product quality alone.
The analysis highlights how aggressively vendors have chased new customers while foregrounding AI capabilities and customer success programmes.
As AI needs rise and cloud costs escalate, organizations are reassessing colocation, adopting a balanced approach that prioritizes workload-specific strategies.
After years of aggressive cloud migration, enterprises are adopting balanced hybrid infrastructure approaches that prioritize workload-specific strategies rather than cloud-first defaults.
Rising cloud costs, AI infrastructure demands, and compliance requirements are driving renewed interest in colocation facilities.
Modern colocation has evolved beyond traditional outsourcing to offer hybrid connectivity, high-density computing, and managed services that complement on-premises and public cloud environments. Organizations now evaluate infrastructure choices through a workload-first lens, aligning each application with the environment best suited to its technical and business requirements.
CISA and 16 partner agencies expand SBOM minimum elements, but critics say the framework lacks real risk-management improvements.
Government partners from around the world released new guidelines for the minimum elements organizations should include in a software bill of materials, authored by CISA and 16 other government entities across four continents.
A couple-dozen changes to SBOM fields make them more comprehensive, but some argue the framework lacks real risk-management improvements.
The updated version supersedes the NTIA's 2021 guidelines, was first drafted in 2025, and was informed by suggestions from 90 commenters including Google, Microsoft, and AWS.
A key concern is that CISA's guidelines are not legally enforceable requirements-the responsibility lies with regulators and customers to require suppliers to follow best practices.
Power and energy operators need fast, validated OT recovery, because prevention alone cannot keep the grid running.
Power and energy organizations face a widening attack surface as connected digital technology is layered onto legacy OT systems, making cyber resilience an operational requirement.
PC-based OT systems such as SCADA servers, HMI and engineering workstations, historians and substation control PCs are essential to grid stability, and downtime affects public safety, reliability, compliance and trust.
The FBI logged more than 2,100 ransomware incidents against U.S. critical infrastructure in 2025. OT recovery differs from IT: remote substations, air-gapped networks, legacy operating systems, vendor-certified builds and little on-site IT staff.
Acronis positions Cyber Protect for OT for low-impact backups, bare-metal and Universal Restore, backup validation and one-click guided recovery.
Applies Kahneman's fast/slow thinking model to security operations, pairing autonomous AI triage with analyst copilots.
The article maps Daniel Kahneman's dual-process cognition framework onto security operations, arguing SOCs should be structured the way human reasoning actually works. It cites research indicating 98% of enterprise alerts can be resolved autonomously, echoing the finding that humans handle roughly 95% of cognition automatically.
Fast autonomous systems should absorb routine alert triage, while AI co-pilot framing fits the complex analysis that genuinely requires human judgment.
Organizations that bring investigation in-house instead of outsourcing to MDR providers build a stronger institutional knowledge base. The conclusion is that effective SOCs need purpose-built autonomous systems for volume plus frontier models held in reserve for cases demanding deliberate analysis.
Sysadmins remain cautious about AI adoption despite corporate pressure, maintaining human oversight for critical systems.
A 2026 Action1 survey reveals that sysadmins' 2024 expectations for AI automation in patch management, vulnerability prioritization, and incident response proved overly optimistic.
While troubleshooting has seen meaningful AI adoption, high-impact functions requiring business context and risk understanding lag significantly.
Sysadmins express concerns about data privacy, security risks, and AI accuracy, with 53% refusing to allow AI to deploy patches to production systems without supervision and only 11% permitting AI to override existing patching policies.
Explains how to turn endpoint hardening from a one-off checklist into a governed, measurable ongoing program.
The article draws a sharp line between treating endpoint hardening as a one-time checklist and running it as a program with governance and measurement.
It describes the interconnected components required: asset inventory, configuration baselines, patch management with risk-based prioritization, privilege governance over local administrator rights, and exception lifecycle management.
A central point is that continuous enforcement, not periodic audit, is what prevents configuration drift.
Three readiness questions are offered to gauge program maturity.
The piece closes on the importance of retaining historical data and generating evidence so compliance and incident-response questions can be answered immediately rather than reconstructed during an investigation.
A walkthrough of six built-in Windows PowerShell commands for diagnosing and fixing common network connectivity problems.
The article highlights six command-line utilities available in Windows PowerShell for network troubleshooting: ipconfig for adapter and DNS cache management, ping for testing reachability of routers and external hosts, nslookup for evaluating DNS resolution performance, netsh for resetting the TCP/IP stack and managing saved Wi-Fi profiles, tracert for mapping packet routes and isolating latency at each hop, and Restart-NetAdapter for quickly rebooting an adapter.
Each tool is explained with practical variants and the scenarios where it helps isolate whether a fault is local or upstream. The author recommends running PowerShell as administrator for full functionality.
A 14-step framework for rolling out IT automation across an enterprise, emphasizing governance and phased adoption over technology alone.
The piece lays out a structured, phased approach to enterprise IT automation rather than attempting to automate everything at once.
Foundational steps cover setting clear automation goals, establishing governance structures, and identifying high-value workflows that genuinely need improvement.
The analysis phase evaluates workflow complexity and the exceptions that still require human intervention, while the implementation phase covers tool selection, staff training, and proving value through pilot projects.
Ongoing management steps address change management integration, measurable KPIs, systematic expansion, and periodic effectiveness reviews.
The author stresses prioritizing repetitive, error-prone, resource-intensive tasks and giving teams time to master tools before scaling.
Organizations must monitor failed authentication attempts to detect attacks, resolve configuration issues, and meet compliance requirements.
Failed authentication attempts reveal critical security and operational insights that most organizations ignore. The article explains how these events indicate credential stuffing attacks, brute force attempts, configuration errors, and user mistakes-each requiring different responses.
By centralizing authentication logs, creating real-time dashboards, and setting up anomaly alerts, teams can detect threats earlier and fix performance issues.
Without this visibility, organizations remain vulnerable to undetected attacks while missing compliance obligations and usability problems affecting customer conversion rates.
Uptime’s 2026 survey shows AI raising costs and density while narrowing management focus, as operators hedge with phased builds and struggle to forecast capacity.
Uptime Institute's 2026 survey of 1,600+ participants reveals AI is driving significant uncertainty across data center management.
Rack density reached 11 kW-a historic high-as operators grapple with rising costs (79% concerned), capacity forecasting challenges (76% concerned), and talent shortages (53% reporting difficulty hiring).
For the first time, third-party facilities host more IT workloads (46%) than enterprise-owned sites (44%), while operators increasingly adopt phased builds to hedge against uncertain AI adoption rates. Despite these pressures, outages continue declining, with only 47% of operators experiencing impactful outages over three years.
A joiner-mover-leaver gap between HR and IT let a fired employee delete files, lock accounts and corrupt a database.
In this PWNED column war story, Yad Senapathy recounts a company of more than 1,000 employees that terminated a worker and never disabled his access. Days passed with the person off payroll but credentials still active, because HR assumed IT would act once the termination was processed while IT waited for a formal HR request.
The ex-employee logged back in and deleted files, locked other users out of their accounts and corrupted a database. Shared admin credentials, account controls and project tracking systems each granted onward access, producing a domino effect.
Damage ran to hundreds of thousands of dollars plus weeks of project delay - the lesson being unowned deprovisioning with no deadline.
As network silicon becomes the bottleneck, 102.4 Tbps switching targets latency and multi-site AI cluster connectivity.
Data Center Knowledge contributor Drew Robb examines next-generation switching architecture for AI clusters. AI clusters depend on predictable bandwidth and low tail latency, and network silicon has become the bottleneck as clusters scale.
The article covers 102.4 Tbps Ethernet switching and what that capacity enables. It distinguishes scaling up within a rack, scaling out across a cluster, and scaling across multiple sites.
Multi-site connectivity is treated as a distinct problem because training jobs increasingly span facilities.
More from Operations · All IT section newsletters · This week