Raju Rangaswami | Searcharxiv

arXiv2026

Idiosyncrasies of Programmable Caching Engines

Programmable caching engines like CacheLib are widely used in production systems to support diverse workloads in multi-tenant environments. CacheLib's design focuses on performance, portability, and configurability, allowing applications to inherit caching improvements with minimal implementation effort. However, its behavior under dynamic and evolving workloads remains largely unexplored. This paper presents an empirical study of CacheLib with multi-tenant settings under dynamic and volatile environments. Our evaluation across multiple CacheLib configurations reveals several limitations that hinder its effectiveness under such environments, including rigid configurations, limited runtime adaptability, lack of quality-of-service support and coordination, which lead to suboptimal performance, inefficient memory usage, and tenant starvation. Based on these findings, we outline future research directions to improve the adaptability, fairness, and programmability of future caching engines.

cs.OS↗

arXiv2017

Analyzing Adaptive Cache Replacement Strategies

Adaptive Replacement Cache (ARC) and CLOCK with Adaptive Replacement (CAR) are state-of-the- art "adaptive" cache replacement algorithms invented to improve on the shortcomings of classical cache replacement policies such as LRU, LFU and CLOCK. By separating out items that have been accessed only once and items that have been accessed more frequently, both ARC and CAR are able to control the harmful effect of single-access items flooding the cache and pushing out more frequently accessed items. Both ARC and CAR have been shown to outperform their classical and popular counterparts in practice. Both algorithms are complex, yet popular. Even though they can be treated as online algorithms with an "adaptive" twist, a theoretical proof of the competitiveness of ARC and CAR remained unsolved for over a decade. We show that the competitiveness ratio of CAR (and ARC) has a lower bound of N + 1 (where N is the size of the cache) and an upper bound of 18N (4N for ARC). If the size of cache offered to ARC or CAR is larger than the one provided to OPT, then we show improved competitiveness ratios. The important implication of the above results are that no "pathological" worst-case request sequences exist that could deteriorate the performance of ARC and CAR by more than a constant factor as compared to LRU.

cs.DS↗