If deletion is so often, you probably should use hash instead of array. Here is some benchmark:

use strict;
use warnings;

use Benchmark("timethese");

my $array;
for (1 .. 100_000) {
        push @$array, $_;
}
my $hash;
for (1 .. 100_000) {
        $hash->{$_} = 1;
}

timethese(1, {'array' => sub {s1($array)}, 'hash' => sub{s2($hash)}});

sub s1 {
    my $array = shift;
    for (my $i = 100_000; $i > 0; $i --) {
        splice(@$array, int(rand($i)), 1);
    }
}

sub s2 {
    my $hash = shift;
    for (my $i = 100_000; $i > 0; $i --) {
        delete($hash->{int(rand($i))});
    }
}
[download]

Result on my XP is:

Benchmark: timing 1 iterations of array, hash...
     array:  8 wallclock secs ( 7.38 usr +  0.00 sys =  7.38 CPU) @  0
+.14/s (n=1
)
            (warning: too few iterations for a reliable count)
      hash:  0 wallclock secs ( 0.20 usr +  0.00 sys =  0.20 CPU) @  4
+.93/s (n=1
)
            (warning: too few iterations for a reliable count)

C:\Perl\bin>perl -w math1.pl
Benchmark: timing 1 iterations of array, hash...
     array:  8 wallclock secs ( 7.44 usr +  0.00 sys =  7.44 CPU) @  0
+.13/s (n=1
)
            (warning: too few iterations for a reliable count)
      hash:  0 wallclock secs ( 0.25 usr +  0.00 sys =  0.25 CPU) @  4
+.00/s (n=1
)
            (warning: too few iterations for a reliable count)
[download]