Helastic: On combining threshold-based and Serverless elasticity approaches for optimizing the execution of bioinformatics applications

Abstract

Recent advances in Next Generation Gene Sequencing (NGS) technologies brought an abundance of bioinformatics and phylogenetics data. The available datasets create new opportunities for studies about the genetic relationships among organisms, which previously relied mainly on manual observations. The state-of-the-art shows that the software employed in this area is based on technologically outdated solutions with ample space for adopting modern computing techniques such as cloud resource elasticity and dynamic load balancing. This article aims to fill this gap with the proposal of Helastic, a model to explore cloud elasticity on jModelTest. The latter is a widely used software for performing statistical selection of nucleotide replacement models in phylogenetic analyzes. Helastic’s contributions appear in a dual elasticity layer that combines the traditional threshold-based, reactive approach with Serverless (also referred to in the literature as Function-as-a-Service, or FaaS). Design decisions include interoperability as a requirement, enabling existing jModelTest applications to benefit from Helastic without significant code changes. We evaluate our proposal through a prototype, which was tested on both elastic and non-elastic scenarios. Data regarding execution time and resource usage are presented in this article. Results demonstrate our solution’s feasibility and the benefits of working with a dual-elasticity approach rather than a single resource rearrangement technique.