Multi-Tier Data Access & Hierarchical Memory Optimization

Building robust and reliable storage systems and memory hierarchies is very important in modern computing environment because hierarchical memory architecture is present almost everywhere. This hierarchy is not limited to attached devices, but it expands to middle-tiers, web caches, and network and internet storage. So optimizing access of the data stored along all these hierarchical levels is very challenging. In this paper we study different performance metrics of the hierarchical memory in order to optimize the cost, access time, and reliability of hierarchical data access. We first present formulas for calculating the probability of exceeding a threshold access time when using hierarchical memory architecture, then we study the behavior of the different performance measures as we vary the number, size and cost of the intermediate memory levels and we come to conclusions that can help the designer build an optimal storage hierarchy.