Will solid-state drives accelerate your bioinformatics? In-depth profiling, performance analysis, and beyond

摘要

A wide variety of large-scale data has been produced in bioinformatics. Inresponse, the need for efficient handling of biomedical big data has beenpartly met by parallel computing. However, the time demand of manybioinformatics programs still remains high for large-scale practical uses dueto factors that hinder acceleration by parallelization. Recently, newgenerations of storage devices have emerged, such as NAND flash-basedsolid-state drives (SSDs), and with the renewed interest in near-dataprocessing, they are increasingly becoming acceleration methods that canaccompany parallel processing. In certain cases, a simple drop-in replacementof hard disk drives (HDDs) by SSDs results in dramatic speedup. Despite thevarious advantages and continuous cost reduction of SSDs, there has been littlereview of SSD-based profiling and performance exploration of important buttime-consuming bioinformatics programs. For an informative review, we performin-depth profiling and analysis of 23 key bioinformatics programs usingmultiple types of devices. Based on the insight we obtain from this research,we further discuss issues related to design and optimize bioinformaticsalgorithms and pipelines to fully exploit SSDs. The programs we profile covertraditional and emerging areas of importance, such as alignment, assembly,mapping, expression analysis, variant calling, and metagenomics. We explain howacceleration by parallelization can be combined with SSDs for improvedperformance and also how using SSDs can expedite important bioinformaticspipelines, such as variant calling by the Genome Analysis Toolkit (GATK) andtranscriptome analysis using RNA sequencing (RNA-seq). We hope that this reviewcan provide useful directions and tips to accompany future bioinformaticsalgorithm design procedures that properly consider new generations of powerfulstorage devices.