Missrate

MissRate is a metric used in computer architecture and data systems to quantify how often a memory access cannot be served from a given fast storage level, such as a CPU cache. It is defined as the ratio of misses to accesses and is typically expressed as a fraction or a percentage: MissRate = Misses / Accesses.

In CPU caches, a miss occurs when the requested data is not found in the current cache level. Misses are often categorized as compulsory (cold) misses, capacity misses, or conflict misses. A lower miss rate indicates better locality and cache effectiveness. The impact of a miss is the miss penalty, the time or cycles required to fetch the data from a lower level of the memory hierarchy, such as L2, L3, or main memory.

Miss rate is commonly measured over a workload or interval using hardware performance counters or profiling tools. Related metrics include hit rate (the complementary measure) and MPKI (misses per thousand instructions). In broader contexts, such as web caching or database buffer pools, MissRate similarly denotes the fraction of requests that are not served from the cache, and it is influenced by cache size, workload, and eviction policies.

Reducing miss rate can involve increasing cache size, improving data locality, changing replacement policies, or employing prefetching strategies. Understanding MissRate helps performance engineers assess how memory hierarchy behavior affects overall system performance and where optimizations may yield the most benefit.