alexaborisov/fastcache-standalone-noavx512 repository overview

FastCache Standalone FastCache is a high-performance in-memory key-value store written in C++ and accessible through gRPC APIs. The Standalone Edition is built for single-node deployments where clustering overhead is unnecessary — local development, edge services, or production workloads that fit on one machine. The focus is maximum single-thread throughput and zero-copy memory operations without compromise.

Key Features High-throughput core. Direct C++ execution path with custom thread-local memory allocators and Google Protobuf Arenas for zero-copy deserialization. No interpreter, no GC pauses, no intermediate copies.

Multi-tenant isolation. Multiple tenants run on a single engine instance with dynamic data isolation built into the core. No separate processes or instances per tenant.

Rich data structures. Native support for Key-Value, Vector, List, Queue, Set, Ordered Set, Map, Ordered Map, and hardware-level Atomics (add, CAS, exchange). Not just strings and hashes — real containers with ordering, ranges, and concurrency built in.

Fine-grained TTL & locking. Per-key and per-element time-to-live policies. Three lock levels — READ_LOCK, WRITE_LOCK, GLOBAL — each with a TTL that auto-releases the lock if the holder crashes or disconnects.

gRPC native protocol. Clean, strongly typed RPC schema for async streaming and request-response operations. No proprietary wire formats — just Protobuf and gRPC.

Limitations No AVX-512 support. The Standalone Edition is compiled without AVX-512 optimizations. It runs on any x86-64 CPU — no specific instruction set requirements. If your workload benefits from AVX-512 (large vector operations, bulk SIMD processing), consider the full HurriCache edition.

Cache Contract The gRPC schema is available at /app/cache.proto inside the Docker image. Use it to generate clients in any language gRPC supports.

Clients Official clients and integrations: https://github.com/orgs/hurricache/repositories⁠