ByteDance details USDB storage system: 85.8% lower write latency for recommender sequences
_reachsumit · x · 2026-11-04
ByteDance researchers published a paper on USDB, its unified storage system for long user behavior sequences serving both real-time inference and batch training.
Key designs:
- CascadeLSM time-tiered compaction materializes application-specific histories
- Differentiated replicas share persistent files; partial dumps reconstruct inference inputs from recent state plus versioned history
- A hybrid columnar format accelerates projected reads
Production migration records show 85.8% lower mean write latency and 42.1% lower normalized infrastructure cost. It outperforms an optimized RocksDB sequence baseline on long-history reads, and aggregation pushdown sustains 8.5x ordinary-read throughput under a constrained receive budget.
More from Infra
- Self-improving software is inevitable: closing the agent-observability loop — bibryam · 2026-11-04
- Gigabyte's AI TOP motherboard packs 2TB RAM and 9 GPUs for home AI labs — vincent_koc · 2026-11-04
- Building a 7-GPU 132GB VRAM local inference rig: consolidate on EPYC or split across two systems? — infectiousstupidity · 2026-11-04
- imec neural probe integrates photonics directly into 130nm CMOS backend — jwt0625 · 2026-11-04
- Gary Marcus: AI data center demand has already passed its peak as projects get killed — GaryMarcus · 2026-11-04
- DeepMem: byte-level multiscale memory cuts GPU memory 54% with lossless decoding — Afshin Khadangi · 2026-11-04