Tag: performance

Linux Real-Time Kernel: When PREEMPT_RT Actually Earns Its Keep

PREEMPT_RT hit Linux mainline in 2024. Here's what real-time actually means, when you need it, and when it'll just slow you down.

Linux PSI: The Pressure Metrics Load Average Wishes It Were

Linux PSI (Pressure Stall Information) exposes real CPU, memory, and IO stall percentages, here's why it beats load average for monitoring.

Multi-Arch Docker Builds With QEMU & buildx

Build Docker images that run on amd64 and arm64 from one command. QEMU emulation, native multi-node builders, CI caching, and manifest lists explained.

cgroups v2 in Practice: Limits, Pressure, Containers

cgroups v2 unified hierarchy explained: read /sys/fs/cgroup/, limit CPU/IO/memory, use systemd slices and PSI pressure files.

crun vs runc vs youki: OCI Runtimes Compared

runc, crun, and youki all run your containers, but they differ wildly in startup speed, language, and cgroups v2 support. Here's which to pick.

KV Cache Quantization: Free LLM Context, Almost

KV cache eats your VRAM at long context, not the weights. Q8/Q4 KV quantization in llama.cpp and vLLM cuts it 2-4x with almost no quality hit.

Dragonfly: P2P Container Image Distribution at Scale

Dragonfly turns your container registry into a P2P swarm. Stop hammering one registry with 200 nodes pulling 4GB GPU images at once.

Speculative Decoding: Faster LLMs With a Tiny Sidekick

Speculative decoding, Gemma 4 MTP, and DeepSeek DSpark all make LLMs 2-6x faster losslessly. How each works, and which to use for local vs. serving.

io_uring: Linux's Async I/O Future Is Already Here

io_uring slashes syscall overhead vs epoll and AIO. Covers the submission/completion queue model, liburing examples, security caveats, and when to skip it.

iperf3 + nload: Network Diagnosis

Your gigabit link drops to 200 Mbps and you don't know why. iperf3 measures throughput honestly, nload visualizes traffic, together they find the bottleneck fast.

fio: Real Disk Benchmarking

Stop trusting marketing IOPS numbers. fio measures what your disk actually does on your workload, sequential, random, mixed. Here's how to read the output honestly.

Sysctl Tuning: The Linux Kernel Settings Nobody Told You About

Linux ships with conservative kernel defaults. These sysctl settings tune your server for networking, memory, and file I/O, with explanations, not just paste-and-pray values.