Classification runtime of Cliffy and Kraken2 on simulated Illumina reads. The figure shows the time required for Cliffy (using approximate document listing) and Kraken2 to classify 10 million simulated Illumina reads (simulated with art_illumina [Huang et al. 2012] using MiSeq v3 error model, 250 bp paired-end). We simulated reads for three different biomes and, for each one, simulated reads from four different hypervariable regions.