Determination of Optimal Thread Pool for Cloud Based Concurrent Enhanced No-Escape Search

Abstract

In this era of high demand for cloud-computing, concurrent and high-performance processing is the viable option to enhance performance and use available resources efficiently. Thread pool architecture is widely implemented in order to improve resource utilization and enhance performance. Studies in this field suggest that Thread pool size is mostly determined by heuristics, trials, or practical experience. This process lacks precision and theoretical justification. In this paper, a novel Hyperbola-based Thread-Pool Analysis (HTA) technique for determining optimal Thread Pool size by considering available bandwidth (upload/download speed) and workload (file-size) has been proposed for any cloud-based concurrent process. Here, HTA has been developed in the context of No-Escape Search (NES), a cloud-based content indexing and search system, proposed by authors in an earlier work. Results of HTA indicate an accuracy of 97.63% in estimation of optimal thread pool size. Incorporating HTA with No-Escape Search resulted a significant decrease in insertion time by 2236 folds and retrieval time by 1747 folds. Additionally, this paper also presents design of enhanced NES with features like pictorial representation of files for the visual summary, more efficient deletion algorithm using lazy delete and duplicate files detection in order to ensure efficient indexing.