TLDR DevOps 2025-10-22

API gateways used to be a monolith that sat between the big scary internet and your production services - but this model doesn't really fit the way software is built today.

This blog by ngrok looks at the "shapes" of API gateways that are more relevant today - including building with agents, exposing ephemeral workloads, self-hosting a PaaS, and more. See all 8 patterns here.

Managing AI traffic?Get early access to ngrok's AI gateway for unified routing, spend management, and failover across any LLM.

Amazon has announced the general availability of vector search for ElastiCache, enabling customers to index, search, and update billions of vector embeddings from major AI providers with microsecond latency and up to 99% recall. The feature supports use cases like semantic caching for LLMs, RAG-based systems, recommendation engines, and anomaly detection, and is available on Valkey 8.2 clusters.

A major AWS outage on October 20 in the US-East-1 region disrupted hundreds of internet services, including Snapchat, Reddit, Fortnite, and Venmo. The outage, caused by DNS resolution failures for the DynamoDB service endpoints, underscores the importance of multi-region architecture and resilient cloud design.

Jenkins' challenges in enterprises stem from poor governance rather than flaws in the tool itself, as unchecked plugin sprawl, inconsistent configurations, and weak access controls create fragility and maintenance burdens. With centralized governance, automation, and compliance frameworks like those provided by CloudBees, Jenkins becomes a scalable, secure, and enterprise-grade CI/CD solution that enables faster, more reliable software delivery.

A deep dive into the BPF LPM trie, a data structure crucial for network packet routing, revealed performance bottlenecks when storing millions of entries, such as entry lookup times taking hundreds of milliseconds and freeing maps locking up a CPU for over 10 seconds. Benchmarks highlighted that freeing a BPF LPM trie with 10K entries can cause soft lockup messages in production, with throughput decreasing to around 1.5 million ops/sec at 1 million entries due to L1 dcache and dTLB miss rates.

Salesforce Commerce Cloud migrated from a self-hosted Prometheus and Thanos stack to Amazon Managed Service for Prometheus, achieving a 40% reduction in AWS costs and eliminating monitoring-related maintenance. The new setup improves scalability, reliability, and cross-cluster visibility.

Learn about the evolution of platform engineering, practical AI use cases, and required skills for modern platform engineers. Attend the webinar for a chance to win a collector's edition LEGO Technic Lamborghini Sián FKP 37. Looking for better ways to provision infrastructure at scale? Book a demo of env zero

WinBoat, an Electron app currently in beta, enables users to run Windows applications on Linux using a containerized approach with a Windows VM inside a Docker container. FreeRDP is used with Windows' RemoteApp protocol for compositing applications as native OS-level windows.

Google Research designed the Coral NPU, an open-source hardware accelerator for machine learning inferencing in low-power wearable devices, and made it freely available. Based on the 32-bit RISC-V ISA, the Coral NPU contains matrix, vector (SIMD), and scalar processor components.

GitHub Copilot CLI brings AI assistance directly to the terminal, allowing developers to clone repositories, debug, manage dependencies, and even open pull requests without leaving the command line. It integrates seamlessly with GitHub accounts, supports secure command approvals, and can be extended with MCP servers to fit custom workflows.

Organizations typically evolve their internal developer platforms through three stages: deployment, secure, and DevOps pipelines, each adding automation, security, and productivity features. This progression forms a practical maturity model where teams first automate delivery, then integrate security, and finally enhance developer experience through documentation, infrastructure automation, and one-click setup.

LinkedIn's generative AI application tech stack was updated to improve AI agents by enabling them to think, plan, and act with users, with the Hiring Assistant being globally available in English by the end of September. Key to this development were defining agents through gRPC service schema definitions and leveraging LinkedIn's messaging system for multi-agent orchestration. Agent interactions were also engineered to be seamless, and a hybrid observability strategy was adopted for agent development.

PaaS used to focus on servers and deployment - but new platforms enable developer creativity. Explore how modern PaaS makes it simple to build and scale. See the evolution of PaaS →

A 15-hour AWS outage in the North Virginia region (us-east-1) on October 19 and 20 impacted over 140 services due to a DNS breakdown within the DynamoDB API endpoint.

Engineering productivity can be measured effectively when done transparently and with minimal burden on engineers.

This tutorial explains how to automate Azure SFTP deployment using Terraform for consistent, version-controlled infrastructure setup across environments.