Fig. 1
From: DECA: scalable XHMM exome copy-number variant calling with ADAM and Apache Spark
DECA parallelization and performance. a DECA parallelization (shown by dashed outline) and data flow. The normalization and discovery steps are parallelized by sample (rows of the samples (s) × targets(t) read-depth matrix). The inputs and outputs of the different components are shown with thinner arrows. b DECA and XHMM execution time starting from the read-depth matrix for s = 2535 on both the workstation and on-premises Hadoop cluster for different numbers of executor cores. Mod. XHMM is a customized XHMM implementation that partitions the discovery input files and invokes XHMM in parallel. c DECA execution time for coverage and CNV discovery for different numbers of samples using the entire workstation (16 cores) and cluster (approximately 640 executor cores dynamically allocated by Spark)