Cache-Centric Multi-Resource Allocation for Storage Services
summary
HARE is a cache-centric allocator that models demand vectors as functions of cache size and jointly optimizes cache plus cache-correlated resources using an iterative harvest/redistribute method: steal cache from cache-insensitive tenants until saved backend units exceed compensation, then split the harvested units by owned share. Implementations: HopperKV (Redis cache fronting DynamoDB) and BunnyFS (microkernel NVMe filesystem). Results show up to 1.9x throughput for HopperKV and 1.4x for BunnyFS; uses sampled ghost cache MRCs and periodic reallocation to tolerate noise.
AI-written summary · about