2.81M Ops/Sec Warm Reads
Nanosecond-fast L1 memory tier powered by Window TinyLFU admission and segmented LRU. Completely eliminates redundant serialization on the hot path.
L1 RAM → L1.5 Off-Heap (/dev/shm) → L2 Redis/Valkey. Delivering 2.81M ops/sec warm reads with zero stampedes.

[Incoming Read Request]
│
▼
┌──────────────────────────────────────────────────┐
│ L1: Smart Memory (W-TinyLFU + CMS) │ ~350 ns (2.81M ops/s)
│ Window LRU (1%) + Segmented Main SLRU (99%) │ Zero V8 allocation
└────────────────────────┬─────────────────────────┘
│ Miss / Eviction Spill
▼
┌──────────────────────────────────────────────────┐
│ L1.5: Off-Heap /dev/shm & NVMe Spill Tier │ ~20 µs (RAM speed)
│ POSIX tmpfs bypasses V8 GC & cloud EBS IOPS │ readOnlyRootFilesystem
└────────────────────────┬─────────────────────────┘
│ Miss
▼
┌──────────────────────────────────────────────────┐
│ L2: Redis / Valkey Distributed Tier │ ~1.2 ms (Network)
│ AES-256-GCM AEAD at-rest + msgpackr records │ Pipelined MGET/MSET
└────────────────────────┬─────────────────────────┘
│ Complete Cache Miss
▼
┌──────────────────────────────────────────────────┐
│ Upstream Database / API (Singleflight Lock) │ Exactly 1 execution
│ Inflight Promise Map coalesces 10,000 callers │ Zero stampedes
└──────────────────────────────────────────────────┘First-class drivers and turnkey adapters built for zero serialization overhead across modern runtimes.