GKE Launches VPA CPU Startup Boost to Cut JVM Pod Startup Times in Half

·1 views

Google Kubernetes Engine (GKE) has released a Public Preview of the VerticalPodAutoscaler (VPA) CPU Startup Boost feature, available on clusters running version 1.36.0-gke.4447000 or newer in the Rapid Channel. The feature addresses a long-standing problem where JVM-based applications require far more CPU during startup than at steady state, forcing teams to permanently over-provision resources. VPA CPU Startup Boost temporarily elevates a pod's CPU allocation at launch and then uses Kubernetes In-Place Pod Resize (IPPR) to scale it back to baseline once the application passes its readiness checks — without restarting the container. Google claims the feature delivers up to 2x faster startup latency while eliminating the persistent CPU waste that inflates infrastructure costs and reduces node packing efficiency. Teams can configure the boost within a standard VPA custom resource, either standalone or alongside continuous vertical autoscaling.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Budget Alerts Don't Stop LLM Overspending — Here's What Actually Does

A developer recounts how a clear budget alert warning of 84% usage failed to prevent a $1,900 charge against a $600 cap after an overnight batch job ran unchecked. The article draws on a 1975 psychology study to argue that depletion warnings can actually increase the perceived urgency to consume remaining resources, not conserve them. A common coding pattern — checking spend before an API call but recording usage only after — leaves ledgers perpetually outdated, especially under concurrent workloads or fast-looping AI agents. The author demonstrates how twelve parallel workers can each independently read a stale budget figure, all clear the threshold, and collectively overshoot the cap by several times. The proposed fix replaces passive balance-checking with a database-level reservation system that locks and holds funds before any API call is made, similar to how financial transactions handle concurrency.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Web NFC API lets browsers read and write NFC tags without a native app

The Web NFC API enables modern web applications to interact with physical NFC tags directly through the browser, eliminating the need for native iOS or Android apps. The API supports reading NDEF messages, writing text, URLs, and vCard contact data, as well as locking tags to read-only mode. Currently, native support is limited to Chrome for Android version 89 and above, and requires an HTTPS connection along with an NFC-capable Android device. Developers must trigger scanning via a user action and should include feature detection to confirm browser compatibility before use. The API represents a significant step toward hardware interaction on the open web without app store dependencies.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

SD-WAN Wins Most 2026 Network Decisions, But MPLS Still Holds Niche Value

The long-running debate between MPLS and SD-WAN has largely settled in favour of SD-WAN for most organisations by 2026, as business traffic has shifted from private data centres to cloud and SaaS platforms. MPLS was built to deliver private, predictable connections to a central data centre, but that value diminishes when most traffic is destined for the public internet anyway. SD-WAN reduces costs by intelligently combining cheaper internet links and routes cloud-bound traffic more directly, improving performance for the majority of use cases. However, MPLS retains a genuine edge for latency-sensitive, mission-critical workloads such as real-time financial transactions, voice, and industrial control systems where guaranteed performance is non-negotiable. Most real-world deployments in 2026 are therefore hybrid, using SD-WAN for general and cloud traffic while retaining MPLS selectively where its dedicated reliability still justifies the premium.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Linus Torvalds Backs AI Coding Tools, Signaling a Shift for Open Source

On July 14, 2026, Linux creator Linus Torvalds stated on the kernel mailing list that Linux is 'not one of those anti-AI projects' and called AI 'a clearly useful tool,' lending significant credibility to AI-assisted development. The comment came amid a thread debating whether LLM-based tooling should be used to help kernel maintainers. A software developer writing on DEV Community had already noted a personal inflection point in December 2025, when working with Claude Opus 4.5 shifted their workflow toward reviewing and directing AI agents rather than manually implementing every detail. The author argues that while AI-generated code can still be buggy, insecure, or poorly structured, the falling cost of generating and evaluating candidates makes the overall process genuinely productive. Torvalds' endorsement is seen as a project-level signal that goes beyond individual experience, marking a broader acceptance of AI tools in serious open-source engineering.

0 comments Read more at DEV Community