Archil's shared-storage approach to stateless analytics
An Archil team member says a shared cache would avoid the cold starts that slow queries on new servers.
TLDR
An Archil team member argues that local SSD caches cause cold starts and tie analytics data to individual servers, complicating containerized workloads. They say Archil's shared storage would let ingestion and query services operate without host-local state, with fresh data and cached S3 data accessible across pods. The piece also identifies a hurdle: POSIX file opens traverse directories, making lookups slow when each step requires a check against S3.
Combined views
—
1 Source, first seen ago
