Table 12. Timing comparison of FSA in regular and parallelized modes.

FSA options	100	200	300	500	1,000 seqs
FSA	6,407 s	27,534 s	—	—	—
FSA –parallelize 10	819 s	5,713 s	22,113 s	—	—
FSA –fast	1,650 s	3,781 s	6,207 s	12,249 s	—
FSA –fast –parallelize 10	201 s	513 s	924 s	2,511 s	15,179 s

Runtimes for FSA in regular, –fast and –parallelize modes when aligning the 16S sequences of Table 11 sequences in unanchored mode (–noanchored) with a 3-state HMM (–noindel2) and refinement disabled (–refinement 0). When running in –fast mode on a cluster with 10 processors (3.00 and 3.20 GHz; 8 GB of RAM), FSA can align 500 16S sequences in 20% of the time required without parallelization. The parallelized FSA was run on a cluster managed by the Condor batch queueing system [66]; nodes were connected by a 100 Mbps Ethernet network. Note that these runtimes are much slower than users can expect from default FSA usage, which uses anchoring for speed (Table 11); we used unanchored mode to make clear the benefits of parallelization.