Sorry 'nall, but that is a ludicrous suggestion.

On a file of around 1GB of randomly generated DNA data:

C:\test>dir randDNA.txt
10/06/2012  01:13     1,036,000,000 randDNA.txt

C:\test>head randDNA.txt
ATAGTAAGGGACCTCAGCGAGTGTCATAAATATAAGTTGTCGAGAGGTAACGATAGACGCCAATCACTTT
+TA
GCATCCAGGGGGTCAGTGTCCTAGCGACGTGGAACAACGACTACGCTTCGTAGGTCTCACCGTATAGATG
+CC
CGGGAGGCCTGCAAAGGAGTGAAGGGTAACGCCTGAACCCTTTGGCCTATCTACGTCGAGATTTCTACCG
+GA
GCGGAGATCTCCCCCCGGATTTCGTCAAATTCTGGAAATAAGTGTAGCAACCGAACGGTATAGCCAGATA
+AT
GCTCGAGCACACGCGGACGGTCTCAGAAACTAATTTTCTTAAGCTGGAACAGGCAACCAAAGATTTTAGA
+TT
ATCGGACGTAGCCAGAAGTGCGGATTTACAGCAACGCCTTTCTCAAAAGTTGCCGTCCCGCGGCACTAAT
+AC
ACCGATATGAAGGCGCTGAAACGATTATGTGTAGTGACGTGCCTTTCAGCGGCTATGGACGCTATCCCCG
+CA
GTCATGAGTCCAATTTGGGGTTAGCTGAAATAACCTGCTGTCCCCTAAAATTGTCGCATTCAAGCAGGGT
+GG
CGGGTACACATGCTAGCATCCGGACGCTATAAGGGCTCCCTTAGTAACATTTCCACTTTCTTGATATTTG
+TG
GGTGCGTTTAACGACGTCATTACTATGAGAGTCGGTATAGCCATCACATAATGACTCGAGCTTACGTCCT
+AC
[download]

This short script reads that into a single scalar and searches it for a single short sequence and prints out the 15000+ offsets where it is found in just over eleven seconds:

#! perl -slw
use strict;
use Time::HiRes qw[ time ];

my $start = time();

local $/;
my $DNA = <>;
$DNA =~ tr[\n][]d;

my $seek = 'AGAGAGAA';

my $p = 0;
printf "%s found at position %d\n",
    $seek, $p while $p = 1+index $DNA, $seek, $p;

printf STDERR "Took %.3f seconds\n", time() - $start;

__END__
C:\test>DNAsearch1 randDNA.txt | wc -l
Took 11.281 seconds
  15313
[download]

This does the same thing using your 100 bytes-at-a-time method:

#! perl -slw
use strict;
use Time::HiRes qw[ time ];

my $seek = 'AGAGAGAA';

my $start = time();

my $file = shift;
open DNA, '<', $file or die $!;

my $size = -s( *DNA );

for my $o ( 1 .. $size - 100 ) {
    read( DNA, my $DNA, 100 );
    $DNA =~ tr[\n][]d;
    my $p =0;
    printf "%s found at position %d\n",
        $seek, $p while $p = 1+index $DNA, $seek, $p;

    seek( DNA, $o, 0 );
}

printf "Took %.3f seconds\n", time() - $start;

__END__
[ 1:32:07.24] C:\test>DNAsearch2 randDNA.txt | wc -l
1441865

[ 4:12:28.90] C:\test>
[download]

It has been running for 30+ minutes now and I don't expect it to finish anytime soon, so I'll leave it running and report back tomorrow.

Updated above: Over 2 1/2 hours and 1.4 million hits instead of 15,000.

Think it over again. You can do much faster by moving your window to the end of the previous section less the length of the substring to match plus one. That will make sure that any small substring split over the present window and the next, will now be completely in the next window.