The first considers similarity across the full
extent of the sequences (a global alignment); the second focuses on regions of similarity
in parts of the sequences only (a local alignment). Some of the publicly available implementations of pairwise
comparison programs (e.g. BLAST and FastA) are fast because they look for
local alignments, and are made to run even faster by incorporating heuristics.
The rationale for local similarity searching is that functional sites are
localized to relatively short regions, which is conserved irrespective of
deletions or mutations in intervening parts of the sequences.