Try it this way. Create a set number of sessions, then dispatch events (using snmp_dispatch_once()) until you get a reply. Then create another and dispatch events until you get another reply, and so on. Adjust the value of $MAXCONCURRENT to control your memory usage.

(Note:Untested code):

#!/usr/bin/perl -w
use warnings;
use strict;
use Net::IP;
use Net::SNMP qw( snmp_dispatch_once oid_lex_sort );
#use Smart::Comments '###';

my $startip     = $ARGV[0] || die "Missing Starting IP";
my $endip       = $ARGV[1] || die "Missing Ending IP";
my $community   = $ARGV[2] || die "Missing community string";

my $ips = Net::IP->new( "$ARGV[ 0 ] - $ARGV[ 1 ]" );

my $MAXCONCURRENT = 50;
my $running = 0;

### Updated do{} while (taken from docs) to while(){} per reply.
while( ++$ips ) { 
    my( $session, $error ) = Net::SNMP->session(
        -hostname       => $ips->ip,
        -version        => 'snmpv2c',
        -nonblocking    =>  1,
        -community      => "$community",
        -timeout        =>  3,
        -retries        =>  1,
    );

    if( defined( $session ) ) {
        my $serialno    = '.1.3.6.1.3.83.1.1.4.0';
        my $mac         = '.1.3.6.1.2.1.2.2.1.6.2';
        my @msoids      = ( $mac, $serialno );
        my $result = $session->get_request(
            -varbindlist    =>  \@msoids,
            -callback        =>  [ \&getms, $session, $ips->ip ]
        );

        $running++; ## Count the sessions started
    }else{
        warn "Session not defined for %s: %s\n", $ips->ip, $session->e
+rror_status;
    };

    ## start another unless we have the max running
    next unless $running > $MAXCONCURRENT;

    ## Dispatch events until we get a reply from one
    snmp_dispatch_once() while $running > $MAXCONCURRENT;

} 

exit;

sub getms {
    my $obj = shift;
    my $session = shift;
    my $hfcip = shift;

    $running--; ## One more done

    if (!defined($obj->var_bind_list)) {
        warn "$hfcip SNMP Error.", $obj->error_status,"\n";
        return;
    }
    ## print values for the oids
    
    $session->close;
}
[download]

This should work, and it's better than the forking idea because it keeps the dispatcher busy with monitoring at least $MAXCONCURRENT requests at a time. Let us know if you still have memory issues with this approach.

ok thx a lot for the script browseruk, thx for the fast response! ;). now when i run it it gave me an error "Can't "next" outside a loop block" so used while( ++$ips ){}; instead of the do(). now as for the memory usage, it now took a lot more. When at 10.0.43 it was already at 60mb and increasing.(remember that all i got so far is timeouts, so 60mb really is too much in my opinion) im interested in the fork method because of this "you can fork off one subnet at a time to a child process and its memory will get reclaimed when it exits.". so if i understood correctly, you would scan 255 ips with a child, then when done all thememory they used will get returned?. So, if the script takes 10mb scanning 10.0.0.0-10.0.0.255 the 10.0.0.0-10.0.1.255 scan should take aproximately the same??