Cost-Efficient Hierarchical Caching for Cloudbased Key-Value Stores

With the rapid expansion of cloud computing services, optimizing data storage and retrieval in cloud-based systems has become crucial. This thesis explores a cost-efficient approach to hierarchical caching in cloud environments, specifically focusing on key-value stores. Building on previous work in elastic caching, this research introduces a hybrid caching system that leverages in-instance memory, in-instance disk, and persistent cloud storage to create a multi-tiered hierarchy. This hierarchy reflects the varying importance of data, optimizing its placement and retrieval. Using Amazon Web Services (AWS) for implementation, the study evaluates different data eviction and placement strategies to enhance performance. Experiments conducted measure query latency and cache hit rates across different storage mediums. Results indicate that our cost-aware caching strategy achieves approximately a 10 % improvement in query latency compared to traditional Least Recently Used (LRU) approaches, with a significantly higher number of hits occurring in memory. This demonstrates that a hierarchical, cost-efficient caching policy can substantially improve performance in cloudbased key-value stores, balancing speed and cost-effectiveness.