Optimal Balancing of Data Transfer Time and Cost in High-Performance Networks

The proliferation of large-scale applications has led to the generation of vast datasets across diverse scientific domains. The subsequent need to transfer such expansive data across geographical distances is essential for collaborative data storage and analysis. While reserving bandwidth on dedicated links within high-performance networks (HPNs) has proved as an efficient means for such extensive data transfers, certain crucial challenges remain to be investigated. In this paper, we delve into the intricate tradeoff between cost and completion time of data transfers using bandwidth reservation on fixed paths with fixed bandwidth of the HPNs, the most common type of bandwidth reservation or data transfer paths. Our focus centers on the scheduling of two types of bandwidth reservation requests (BRRs) that encapsulate this tradeoff: (i) minimizing data transfer cost within prescribed deadlines, and (ii) achieving the earliest data transfer completion time while adhering to predefined cost constraints. We propose two algorithms to optimize the scheduling of individual BRRs of these two types. We then compare the proposed algorithms with existing ones from the perspective of different performance metrics, and efficacy of the proposed algorithms is verified through extensive simulations.