Formalization of Experimental Methods for Evaluating Memory Allocators, Taking into Account the Semantics of the Object Lifecycle

Abstract

Memory allocator evaluation still lacks a stable methodological basis, even though allocator behavior depends on workload heterogeneity, concurrency patterns, and the temporal semantics of allocated objects. This article addresses that gap through a formal analytical framework that treats object lifetime as an explicit experimental variable instead of a residual statistic. The study systematizes recent allocator research, identifies points at which benchmark practice loses methodological precision, and proposes a phase-based evaluation model that links allocation events with lifecycle classes, workload transitions, and metric interpretation. The presented materials include 11 recently peer-reviewed sources on allocator characterization, lifetime profiling, semantics-aware allocation, tiering, and benchmark methodology. Comparative analysis, conceptual synthesis, typologization, and analytical generalization shape the methodological basis. The analytical section derives a lifecycle-centered evaluation schema, distinguishes workload classes that require different observables, and formulates reporting rules for reproducible allocator assessment in server workloads, managed runtimes, heterogeneous-memory systems, and application-specific environments.