Tag: performance

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

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

DragonflyDB vs Redis 8.x and Valkey 9.x: architecture deep dive, real performance benchmarks, honest gotchas, and when to swap your cache layer for a dragon.

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.

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.

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.

A kernel upgrade halved one team's PostgreSQL throughput. The fix was kernel tuning: huge pages, THP, swappiness, I/O scheduler. Here's what matters and why.

Learn multi-stage Docker builds to slash image sizes by 90%. Practical before/after examples for Node.js, Python, and Go with real size comparisons.

Learn systemd socket activation to start services on-demand, save RAM, and cut boot time. Includes .socket unit files, real examples, and testing with systemd-socket-activate.