Figure 5:

(A) Runtime breakdown of the full run (T = 300) on a 36-core CPU, 262 GB RAM server. Owing to some technical issues, no usage log is available for steps prior to the sub-Pfam extraction. CPU usage log shows that most of the time, BiG-SLiCE only uses 1 CPU core, giving room for further improvement, e.g., via SQL parallelization. Spikes in the RAM usage (peak = ∼150 GB) came from the periodic “dumping” of the in-memory database (used to speed up runtime) into an SQLite db file. (B) Runtime comparison between multiple runs, with T = 300 bearing the full load of performing input processing and feature extraction. Here, runtimes are separately shown for both the clustering (GCF model construction + membership assignment) and other steps (input parsing, hmmscanning, and feature extraction).