To measure throughput use LWP or Net::FTP and measure the time (using Time::HiRes for greater accuracy) it takes to push a large chunk of data across. Divide the data size by the elapsed time and you will have a usable aproximation of end-to-end application-layer throughput.

use LWP::Simple;
use Time::HiRes qw( gettimeofday tv_interval );

my $t0 = [gettimeofday];
my $d=get('http://www.perlmonks.org');
my $elapsed=tv_interval ( $t0, [gettimeofday]);
my $throughput=length($d)/$elapsed;
print "throughput=$throughput Bytes per second\n";
[download]

Another (more accurate) measurement could be gathered by spawning off several such "data pusher" processes in parallel. Then use raw network IO stats (gathered by SNMP or however you want) to determine the throughput during the period of the test.

You can get latency by using a Time::HiRes timer around Net::Ping. My example below takes only a single ping, for real use you'd probably want to take multiple tests and average them.

use Time::HiRes qw( gettimeofday tv_interval );
use Net::Ping;

my $p = Net::Ping->new("icmp");
my $t0 = [gettimeofday];
$p->ping('127.0.0.1',5);
my $elapsed=tv_interval ( $t0, [gettimeofday]);
print "latency=$elapsed\n";
[download]