#!/usr/bin/perl
use warnings;
use strict;
use Socket;

#dosn't work well for images, but you can fix that

my $url  = "http://zentara.net/index.html";
my $infile = $url;
$infile =~ tr#\/#-#;
print $infile;
my $host = "zentara.net";
$| = 1;

my $start = times;
my ( $iaddr, $paddr, $proto );
$iaddr = inet_aton($host);
#$iaddr = ( gethostbyname($host) )[4];
$paddr = sockaddr_in( 80, $iaddr );
$proto = getprotobyname('tcp');
unless ( socket( SOCK, PF_INET, SOCK_STREAM, $proto ) ) {
    die "ERROR Dude: getUrl socket: $!";
}
unless ( connect( SOCK, $paddr ) ) { die "getUrl connect: $!\n"; }

my @head = (
    "GET $url HTTP/1.0",  #maybe better to use 1.0, instead of 1.1 for
+ "no keep-alive" ??
    "User-Agent: Mozilla/4.78 [en] (X11; U; Safemode Linux i386)",
    "Pragma: no-cache",
    "Host: $host",
    "Accept: image/gif, image/x-xbitmap, image/jpeg, image/pjpeg, imag
+e/png, */*",
    "Accept-Language: en"
);
push( @head, "", "" );

#Build Header and print to socket
my $header = join( "\015\012", @head );
print "sending request\n$header\n\n";
select SOCK;
$| = 1;
binmode SOCK;
print SOCK $header;

my $body = '';
open (FH,"> $infile") or warn "$!\n";

while (<SOCK>) {
    my $data = $_;
    $data =~ s/[\r\n\t]+$//s;
    $data =~ s/^[\r\n\t]+//s;
    last if $data =~ /^0$/s;
    my $len = length($data);
   print STDOUT "len:$len\n";

    $body .= $data;
    last if $data =~ /\<\/html\>$/is;
    if ( $data =~ /\<\/body\>$/is ) {
        $body .= qq|</html>|;
        last;
    }
    
    print FH $data;

}
unless ( close(SOCK) ) {
    return ("getUrl close: $!");
}
select STDOUT;
close SOCK;
close FH;
my $end = times;
my $diff = $end = $start;
print "Took $diff to access page\n";
[download]

I've heard LWP is slow and too CPU-intensive task for crawlers

Then the first step should be to test that. Maybe it was too slow for somebody on his own, weak machine, but it's no issue for you.

If it's to slow for you, try search cpan for spider and crawler, maybe some of the results might help you.

If you care very much for CPU time, consider using curl or wget, which are written in C and probably less CPU intensive.