Embedded RAM is soldered and unupgradable. zram compressed swap relieves pressure without changing hardware. Compressed memory beats bigger memory zram sizing: allocate 25-50% of physical RAM, with zstd for compression ratio or lz4 for speed; swappiness: set 100 so the kernel happily uses zram, but disable or minimize real disk swap; OOM protection: give critical […]
MQTT is the nervous system of a smart home, yet most guides stop at “install and go”. An unauthenticated broker becomes a liability the moment an untrusted device joins. The bus needs permissions too Enable password files and listeners: two accounts — sensors publish-only, controllers subscribe; Isolate by topic prefix: zigbee2mqtt/# writable only by Z2M, […]
The key metrics are NAT throughput, concurrent connections and offload capability; clock speed is not the first priority. Routers are judged on forwarding, not benchmarks RK3528’s place: quad A53 with dual gigabit, low power and good cost for home and small business routing; RK3568’s edge: richer PCIe and USB3 for Wi-Fi 6 modules or multi-port […]
Switching from Emby or Plex to Jellyfin is rarely blocked by features — it’s years of watch history and favorites. Migration happens at three levels of increasing difficulty. Three levels of migration Media files are easy: keep the same folder structure and let Jellyfin re-scan; Metadata migrates well via Jellyfin’s NFO import, which reads Emby’s […]
Not every job belongs on the main SoC. Hard-realtime control on an MCU with perception and scheduling on the host is the sturdier design. Task split matters more than TOPS Layered architecture: motor loops and safety stop on an MCU; SLAM and visual avoidance on the RK3588 host; Compute allocation: lidar SLAM is CPU-bound, visual […]
Originally each store had its own player updated by walking around with a USB stick, making a promo change a week-long job. From 40 standalone players to one system Architecture: a central CMS composes content while store players pull on schedule and precache, so switching never blacks out; Operations: heartbeats and playback logs report back, […]
RK3576’s four A72 + four A53 cores cover most light server roles at lower power and price. The skill is knowing which loads fit. Match the chip to the load Handles fine: light Docker services, AdGuard, DDNS, Uptime Kuma, small databases, Nextcloud for ten users; Tight: single-stream Jellyfin transcode, Immich with ML off, Frigate at […]
Port chaos follows service sprawl. A reverse proxy with a single 443 entry and host-based routing is the cleanest structure. One entry for all services Deploy: the jc21/nginx-proxy-manager three-container stack, publishing 80/443; Certificates: built-in Let’s Encrypt renewal; wildcard certs need DNS validation (one API token covers a year); Security: expose only 443 and bind the […]
Quantization compresses FP32 weights to INT8 with big speed and memory wins, but a calibration set mismatched to your scene collapses accuracy. The calibration set is the soul of quantization Calibrate from real cameras: 200-500 frames covering day/night, weather and angles beat generic datasets; Per-layer sensitivity: keep accuracy-critical head/tail layers in FP16 mixed quantization; Always […]
Linux lands most interrupts on CPU0 by default, leaving one core saturated while others idle and total throughput capped. Interrupts defaulting to CPU0 is the trap Inspect: cat /proc/interrupts shows which core each IRQ lands on — watch NIC and storage controllers; Pin them: set /proc/irq//smp_affinity to spread NIC interrupts across the A76 cores; Pair […]