Volume 343, Issue 2 · 10 articles from 2 issues
Datadog CI/CD Optimization correlates pipeline, test and runner data to cut CI wait times, flaky failures and compute costs.
Datadog argues that AI-assisted development only speeds delivery if CI keeps pace with growing pull request volume.
Datadog CI/CD Optimization brings pipeline and test data together with the commits, logs, runners and infrastructure behind each run, helping platform, DevEx and DevOps teams shorten feedback loops.
Teams can address flaky failures and cut unnecessary test execution, keeping compute spend from rising as fast as pipeline volume while developers and coding agents get faster feedback.
Datadog BYOC adds cross-cluster queries so teams and AI agents can search data across regional deployments at once.
Datadog Bring Your Own Cloud (BYOC) runs each deployment as a Kubernetes cluster with its own object storage inside customer-controlled infrastructure, supporting data residency and cost needs while teams keep using Log Explorer, dashboards, monitors, Bits AI and the MCP Server.
Deployment boundaries previously forced engineers to repeat queries per cluster, for example to check whether an error in Frankfurt also affected Virginia or Mumbai.
New cross-cluster queries let both teams and AI agents search across BYOC clusters in a single query.
Datadog's Visual Labeler lets teams name web app user actions in Product Analytics without code or new deployments.
Datadog Product Analytics shares SDKs with Real User Monitoring, which autocapture clicks and taps but name them after page elements rather than user intent.
The Visual Labeler in Datadog Action Management lets users select an element in their web app and give the interaction a business-meaningful name without writing code or deploying, and it now runs as a browser extension that works on authenticated pages.
That removes the need for engineering tickets to answer questions such as how many users started checkout.
Datadog Observability Pipelines adds Exabeam packs to filter and route high-signal security logs to the Exabeam SIEM.
Exabeam's SIEM and UEBA detections rely on behavioral baselines built from identity, VPN, workstation and email logs, while firewalls, endpoints and web gateways generate routine volume that inflates ingest and retention costs.
Datadog Observability Pipelines provides packs for processing and routing security logs to Exabeam, letting teams filter upstream to keep high-signal events and reduce noise.
The post explains how this approach sharpens baselines for insider threat, lateral movement and credential compromise detection.
Datadog moved internal host vulnerability scanning to its own Agent, keeping scan freshness above 99%.
As Datadog's host fleet grew, its original vulnerability scanning system struggled to reach and assess every host.
Over the past year the company migrated host vulnerability scanning to the Datadog Agent, making Datadog Cloud Security the system of record for host vulnerability findings.
The Agent-based workflow keeps scan freshness above 99%, measured as the share of in-scope hosts scanned in the previous 24 hours, and fits into Datadog's broader vulnerability management program.
A strategy profile of Datadog's evolution from monitoring tool to unified observability platform moving into AI-driven remediation.
This long-form profile traces Datadog's path from an unglamorous monitoring tool to a roughly $100 billion platform.
It likens the company to a building manager that first brought separate instruments onto one wall, then added alerting, on-call and incident response, and is now adding AI to investigate and fix problems.
It covers the product strategy, competition between observability platforms and AI SRE tools, the bottom-up engineer-led go-to-market and branding, and customer complaints about complicated usage-based bills, noting that usage pricing exposes Datadog to customers' AI-driven optimization cycles.
It closes with lessons for leaders on moving from seeing problems to resolving them reliably.
Datadog explains how it classifies thousands of internal services by criticality to focus deployment and reliability work.
Datadog runs thousands of internal services and long treated them uniformly for deployment and reliability decisions.
As its data center footprint grew, sequential rollouts lengthened and a single strict SLO target led to over-investing in some services and under-protecting others, so the company needed to identify which services are truly critical.
The post describes how Datadog defines service criticality at scale to concentrate reliability work where failures would hurt customers most.
Datadog explains how Kubernetes Dynamic Resource Allocation can consolidate workloads onto fewer GPUs and lower spend.
As GPU spending grows, Kubernetes' legacy device plugin model limits utilization gains, since partitioning must be configured before scheduling, requests can't distinguish GPUs by capability, and a workload can reserve a whole GPU while using part of it.
Kubernetes Dynamic Resource Allocation (DRA) offers a more flexible model.
Datadog shows how DRA can consolidate workloads, shorten topology-sensitive jobs and reduce infrastructure costs.
Datadog launches Scout 1.0, its first AI model specialized for investigating production issues, powering Bits AI SRE.
Datadog introduced Scout 1.0, its first AI model specialized for investigating production issues, which powers its Bits AI SRE and was trained to cut inference cost so investigations can run far more often and earlier in an incident.
Datadog says Scout 1.0 medium matches leading frontier models at 3-8x lower estimated inference cost on its target scenarios.
With test-time scaling, the xhigh variant matches a top frontier model at 4x lower cost.
Datadog RUM adds Operations Monitoring at no extra cost to track whether critical app steps succeed and how long they take.
Datadog Real User Monitoring captures views, actions, resources and errors, but individual events don't directly show whether a critical step worked.
Operations Monitoring, included with RUM at no additional cost, lets teams monitor application-controlled steps across real user traffic.
An operation is a technical step, such as product checkout or sign-in, that users expect to complete reliably, and teams can track its success and duration.