From the archive
Tuesday, 3 November
Folded from 18 sources that day.
Sorted by quality and by how many sources covered it, the same for everyone.
-
Designing a planned switchover for minimum disruption
Describes how Neki's sharded Postgres system performs planned primary switchovers: preflight checks, WAL convergence before promotion, and client routing recovery, without failure detection.
-
KVCache Meetup Korea 2026: Event Recap
A recap of the Korean meetup on LLM inference and KV cache, covering CXL memory expansion, GPU and HBM bottlenecks, RNGD request scheduling and LMCache MP mode.