Metric Space Indices for Dynamic Optimization in a Peer to Peer-Based Image Classification Crowdsourcing Platform

Abstract

Large-scale computer platforms that process users’ online requests must be capable ofhandling unexpected spikes in arrival rates. These platforms, which are composed of distributedcomponents, can be configured with parameters to ensure both the quality of the results obtainedfor each request and low response times. In this work, we propose a dynamic optimization enginebased on metric space indexing to address this problem. The engine is integrated into the platformand periodically monitors performance metrics to determine whether new configuration parametervalues need to be computed. Our case study focuses on a P2P platform designed for classifyingcrowdsourced images related to natural disasters. We evaluate our approach under scenarios withhigh and low workloads, comparing it against alternative methods based on deep reinforcementlearning. The results show that our approach reduces processing time by an average of 40%.