Design and Implementation of a High-Performance In-Memory Key-Value Cache Server in Zig Using Robin Hood Hashing

This paper presents the design and implementation of a high-performance in-memory key-value cache server written in Zig 0.14, leveraging the Robin Hood hashing algorithm to optimize probe sequence lengths and memory utilization. We employ Zig's std.AutoHashMap and std.HashMapUnmanaged abstractions, combined with a custom allocator strategy based on std.heap. GeneralPurposeAllocator, to achieve deterministic memory behavior and minimal overhead. The Robin Hood hashing scheme balances probe sequence lengths by allowing new entries to "steal"slots from entries with shorter probe distances, reducing clustering and improving lookup latency under high load factors. Our benchmarking on an isolated Intel i7-8700@3.2 GHz demonstrates average GET latencies of ~20 μs and SET latencies of ~25 μs at a 0.9 load factor - improvements of 20% and 17%, respectively, over linear probing implementations - and a 15% reduction in memory footprint per million entries. These results validate Zig's suitability for systems programming in caching scenarios and highlight the benefits of Robin Hood hashing for in-memory data stores.