Analysing Dependability and Performance of a Real-World Elastic Search Application

Abstract

Increased complexity in IT, big data, and advanced analytical techniques are some of the trends driving demand for more sophisticated and scalable search technology. Despite Quality of Service (QoS) being a critical success factor in most enterprise software service offerings, it is often not a generic component of the enterprise search software stack. In this paper, we explore enterprise search engine dependability and performance using a real-world company architecture and associated data sourced from an ElasticSearch implementation on Linknovate.com. We propose a Fault Tree model to assess the availability and reliability of the Linknovate.com architecture. The results of the Fault Tree model are fed into a Stochastic Petri Net (SPN) model to analyze how failures and redundancy impact application performance of the use case system. Availability and MTTF were used to evaluate the reliability and throughput was used to evaluate the performance of the target system. The best results for all three metrics were returned in scenarios with high levels of redundancy.