InfoQ Homepage Performance Content on InfoQ
-
Cloudflare Opens Quicksilver to Customers as Workers KV Instant, inside a 1 MB Envelope
Cloudflare has put Quicksilver, the internal key-value store behind nearly every request it serves, into private beta as a Workers KV mode. Reads resolve in 1.62 ms at p99 against 287 ms for classic KV. The constraints decide the use case: 10,000 keys, one megabyte per namespace, and one write per namespace per second.
-
Cloudflare Brings On-Demand CPU and Memory Profiling to Production Workers
Cloudflare has introduced on-demand CPU and memory profiling for Workers and Durable Objects, allowing developers to investigate performance bottlenecks in production using interactive flamegraphs.
-
Microsoft Commits Maia to Multi-Gigawatt Inference, and its Efficiency Metric Draws Questions
Microsoft will scale its Maia accelerator family to multi-gigawatt deployments from 2027, making it the primary inference system for its own products. Maia 300 collapses scale-up and scale-out onto one Ethernet fabric and moves KV state between prefill and decode workers without leaving it. Practitioners have questioned the tokens-per-watt claim that leads the announcement.
-
Instagram Modernized its UI with an AI-Native Architecture, Cutting Token Costs by 33%
Instagram Direct migrated its Android UI from the legacy Android View system to Jetpack Compose, creating an AI-native UI codebase that is easier for AI coding agents to understand, modify, and maintain. According to Instagram engineers, the migration reduced UI code volume by 50% and token costs per AI agent session by 33%.
-
Github Migrates Copilot Runtime to Rust with AI-Assisted Rewrite
GitHub migrated more than 800,000 lines of Copilot runtime code from TypeScript and Node.js to Rust in about 14.5 weeks using AI-assisted development. The incremental migration used N-API interoperability, automated testing, human review, and 128 pull requests while continuing to ship releases.
-
GKE Pod Snapshots Cut Model Load Times, and Move the Work to Snapshot Lifecycle Management
Google has published benchmarks for GKE Pod snapshots, reporting up to 89% lower startup latency and a 70B model loading in 37 seconds. The feature checkpoints CPU and GPU memory through gVisor into Cloud Storage. Practitioners have asked whether invalidation is the harder problem, since snapshots match on a spec hash, machine series, and kernel and driver versions.
-
Cloudflare Cuts 100 TB of Memory from 1.1.1.1 DNS Cache
Cloudflare redesigned the in-memory representation of its Big Pineapple DNS cache, reducing the per-entry footprint by 56% and freeing roughly 100 TB of working-set memory across its fleet. The Rust-based changes also increased cache insertion throughput by 43% and reduced lookup latency by 19%, while enabling Cloudflare to increase cache capacity without additional memory.
-
Cloudflare Measures Origin TLS Preferences, Cutting Handshake Retries from 52% to 3.7%
Cloudflare has replaced its static X25519 guess for origin TLS handshakes with per-origin measurement. HelloRetryRequests on scanned origins fell from roughly 52% to 3.7%, removing over 150 ms from p90 latency. Post-quantum connections completing in one round trip rose from 0% to 99.2%, though only 12.8% of origins support it.
-
Lambda SnapStart Comes to Container Images, Ending a Packaging Tradeoff
AWS has extended Lambda SnapStart to container image functions, which hold up to 10 GB against 250 MB for zip archives. Teams previously chose between dependency headroom and sub-second startup. A Reddit thread from a month earlier shows what that cost: stripping whitespace and docstrings from installed packages to stay under the limit.
-
NVIDIA Personal AI Router Distributes AI Tasks across Local Compute
NVIDIA Personal AI Router (PAIR), now available in beta, lets you combine the inference capacity of multiple computers on your local network and automatically distribute AI requests among them. It is primarily designed for local multi-agent AI workloads, where multiple independent model calls can otherwise overwhelm one GPU.
-
Shopify Introduces Gisting: Compressing LLM System Prompts into Learned Tokens
Shopify's engineering introduced gisting, a novel technique for compressing long LLM prompts into a smaller set of learned "gist" tokens, improving throughput and reducing inference cost.
-
Project Valhalla's First Preview: JEP 401 Redefines == for Java Objects
JEP 401, integrated into JDK 28, introduces value objects. These new class instances feature final fields, altered behavior for equality checks, and stricter construction and synchronization rules. It aims to enhance efficiency and reduce memory allocation costs. However, the preview is disabled by default and requires specific configuration at compile and run time.
-
GitHub Increased Instant Navigation from 4% to 22% by Rethinking Client Side Architecture
GitHub redesigned GitHub Issues navigation using a client-side architecture that combines caching, predictive prefetching, and service workers to reduce perceived latency. The approach uses IndexedDB, in-memory caching, and background synchronization to serve data faster. GitHub reported instant navigation improvements from 4% to 22%, with latency reductions across multiple navigation paths.
-
Android Studio Quail 2 Redesigns Agent Mode, Streamlines AI-Assisted Coding
The latest release of Android Studio, Quail 2, now stable, expands Gemini/AI Agent Mode inside the IDE by enabling multiple AI conversations in parallel and further advancing Google's push to integrate AI-powered workflows directly into Android Studio. It also enhances debugging and profiling capabilities and makes it easier to explore experimental features.
-
DoorDash Uses Envoy and Valkey for a 1.5M RPS Proxy Cache with 99.99999% Availability
DoorDash has developed Entity Cache, a transparent proxy caching platform built on Envoy and Valkey to reduce redundant service-to-service requests across its microservices architecture. Operating within DoorDash’s service mesh, the platform serves over 1.5M requests per second with 99.99999% availability through caching, event-driven invalidation, failure handling, and performance optimizations.