--- author: "Umesh Malik" canonical: "https://umesh-malik.com/blog/tag/performance" description: "Explore articles tagged with Performance by Umesh Malik — AI Engineer, LLM & GenAI Developer. Learn Performance best practices, practical tips, and in-depth guides." title: "Umesh Malik's Blog - Performance Articles | Performance Tutorials" tokens: 1232 generator: "scripts/generate-page-markdown.mjs" --- [← Back to Blog](https://umesh-malik.com/blog) # Performance 9 articles [![Two CPU pools under a sched\_ext policy: latency-critical request threads separated from background work, with a watchdog fallback to EEVDF](https://umesh-malik.com/blog/fix-eevdf-latency-regression-sched-ext-cover.png) Web Engineering • Aug 16, 2026 ### Fix EEVDF latency regressions with sched\_ext: Meta's 28% p99 win Your p99 got worse after a kernel upgrade and nothing else changed? That is an EEVDF latency regression. Meta cut p99 28% with a two-pool sched\_ext policy. 12 min read Read more →](https://umesh-malik.com/blog/fix-eevdf-latency-regression-sched-ext) [![Cover showing the vLLM VRAM budget split into model weights, runtime overhead and KV cache, with the KV cache block math that converts free VRAM into concurrent sequences](https://umesh-malik.com/blog/vllm-throughput-tuning-flags-cover.png) LLM Engineering • Aug 8, 2026 ### vLLM throughput tuning: configure these four flags, not a bigger GPU vLLM throughput tuning starts with KV cache blocks, not a bigger GPU. The four flags that decide your tokens/sec, and the one that quietly backfires. 10 min read Read more →](https://umesh-malik.com/blog/vllm-throughput-tuning-flags) [![Layer-by-layer LLM inference streaming one transformer layer at a time from disk onto a 4GB GPU](https://umesh-malik.com/blog/run-70b-llm-on-4gb-gpu-airllm-cover.png) LLM Engineering • Aug 4, 2026 ### Run 70B LLM on 4GB GPU: AirLLM's Real Tradeoff Run 70B LLM on 4GB GPU hardware with AirLLM's layer-by-layer inference. The VRAM math is real — you just pay for it in disk bandwidth. The honest tradeoff. 11 min read Read more →](https://umesh-malik.com/blog/run-70b-llm-on-4gb-gpu-airllm) [![Cover showing an HTML document streaming placeholder markers that get filled out of order by template-for elements, with no JavaScript](https://umesh-malik.com/blog/streaming-html-out-of-order-without-javascript-cover.png) Web Engineering • Jul 14, 2026 ### Streaming HTML Out of Order Without JavaScript (2026) Streaming HTML out of order without JavaScript: how Declarative Partial Updates and Declarative Shadow DOM reorder content natively in Chrome 148. 13 min read Read more →](https://umesh-malik.com/blog/streaming-html-out-of-order-without-javascript) [![A lightning bolt splitting between Next.js and Vite logos, symbolizing Cloudflare's revolutionary Vinext framework that's 4.4x faster](https://umesh-malik.com/blog/cloudflare-vinext-cover.png) Web Engineering • Feb 25, 2026 ### Cloudflare viNext: The $1,100 Next.js-on-Vite Rebuild Cloudflare viNext rebuilt Next.js on Vite for $1,100 in 7 days: 4.4x faster builds, 57% smaller bundles, already powering CIO.gov in production. 27 min read Read more →](https://umesh-malik.com/blog/cloudflare-vinext-next-js-vite-revolution) [![A desktop workstation running a local LLM for coding — 80 billion parameters, 3 billion active — representing the shift from cloud AI to local AI coding](https://umesh-malik.com/blog/local-llm-coding-cover.png) AI Coding Agents & DX • Feb 22, 2026 ### Qwen3-Coder: Run an 80B-Parameter LLM on Your Desktop Qwen3-Coder runs 80B parameters on a desktop with only 3B active per token — and plugs into Claude Code. Why the cloud-only era of AI coding is ending. 16 min read Read more →](https://umesh-malik.com/blog/local-llm-coding-revolution-qwen3-coder-desktop) [![Editorial cover: Node.js V8 pointer compression cuts heap memory in half](https://umesh-malik.com/blog/nodejs-memory-cut-in-half-cover.png) Web Engineering • Feb 18, 2026 ### Node.js Pointer Compression: Cut Heap Memory ~50% V8 pointer compression finally lands in Node.js: one Docker image swap cuts heap memory ~50%, improves P99 latency, and can save $80K–$300K a year. 17 min read Read more →](https://umesh-malik.com/blog/nodejs-memory-cut-in-half-pointer-compression) [![Core Web Vitals dashboard showing three gauge meters for LCP, INP, and CLS with before and after optimization results](https://umesh-malik.com/blog/core-web-vitals-cover.png) Web Engineering • Nov 12, 2025 ### How to Fix Core Web Vitals: LCP, INP & CLS (2026) A hands-on guide to optimizing Core Web Vitals (LCP, INP, CLS). Covers measurement, diagnosis, and specific fixes with before/after examples from real projects. 6 min read Read more →](https://umesh-malik.com/blog/core-web-vitals-optimization-guide) [![React performance optimization showing 10 techniques including memoization, code splitting, and virtualization](https://umesh-malik.com/blog/react-performance-cover.png) Web Engineering • Nov 20, 2024 ### React Performance Optimization: 10 Proven Techniques React performance optimization techniques — memoization, code splitting, virtualization, and 7 more battle-tested strategies from real apps. 5 min read Read more →](https://umesh-malik.com/blog/react-performance-optimization-techniques)