This paper presents a comprehensive analysis of the workload characteristics of Facebook's Memcached deployment, one of the world's largest key-value stores. The study collects detailed traces from five different Memcached use cases, capturing over 284 billion requests over several days. The analysis covers various aspects, including request composition, size, and rate; cache efficacy; temporal patterns; and application use cases. Key findings include a GET/SET ratio of 30:1, higher than previously assumed, and the observation that some Memcached applications behave more like persistent storage than a cache. The paper also proposes a simple model to generate more realistic synthetic workloads, highlighting the importance of understanding workload characteristics for improving the performance, scalability, reliability, and cost efficiency of key-value stores. The analysis reveals that while strong locality metrics do not always guarantee high hit rates, there is still room for efficiency and hit rate improvements in Memcached's implementation. The study concludes with suggestions for future work, including improvements in memory allocation and cache replacement policies.This paper presents a comprehensive analysis of the workload characteristics of Facebook's Memcached deployment, one of the world's largest key-value stores. The study collects detailed traces from five different Memcached use cases, capturing over 284 billion requests over several days. The analysis covers various aspects, including request composition, size, and rate; cache efficacy; temporal patterns; and application use cases. Key findings include a GET/SET ratio of 30:1, higher than previously assumed, and the observation that some Memcached applications behave more like persistent storage than a cache. The paper also proposes a simple model to generate more realistic synthetic workloads, highlighting the importance of understanding workload characteristics for improving the performance, scalability, reliability, and cost efficiency of key-value stores. The analysis reveals that while strong locality metrics do not always guarantee high hit rates, there is still room for efficiency and hit rate improvements in Memcached's implementation. The study concludes with suggestions for future work, including improvements in memory allocation and cache replacement policies.