This can be done with Biopieces www.biopieces.org) like this:

read_fasta -i data_in.fna | split_seq -w 100 -s 20 | write_fasta -o data_out.fna -x

Perl option:

#!/usr/bin/perl

use strict;
use warnings;

my $len = 100;
my $over = 20;
my ($seq_id, $seq);

while (<>) {
    chomp;
    if (m/^>/) { $seq_id = $_; } else { $seq .= $_; }
}

for (my $i = 1; $i <= length $seq; $i += ($len - $over)) {
    my $s = substr ($seq, $i - 1, $len);
    print "$seq_id ($i-", $i + (length $s) - 1, ")\n$s\n";
}

This can easily be done with genometools as:

gt shredder -minlength 100 -maxlength 100 -overlap 20 ecoli.fasta > ecoli_shredded.fasta

Note that there are also -coverage and -sample options for shredder that will allow you to control how your fragments are generated. Another good option is dwgsim, which is capable of doing sampling with various kinds of mutations, but this may be more than what you need. The Biopieces (mentioned by martinahansen) or genometools solutions are probably more appropriate based on your question.

You can use pyfasta to do this

pyfasta split -k 100 -o 20 input.fasta -n 1