Yes, there is a trick, you can just localize the '$/' variable, set it to '>', and loop through your data. I haven't seen your actual data, but I made a few assumptions, and came up with the code below, which I believe is memory efficient.

use strict;
use Data::Dumper;

my @seq;  # array of sequences
my %seq;  # hash lookup of sequences based on id's

{
  local $/ = '>';
  while (<DATA>) {
    s/^>//g; # strip out '>' from beginning
    s/>$//g; # and end of line
    next if !length($_);           # ignore empty lines

    my ($header_info) = /^(.*)\n/; # capture the header
    s/^(.*)\n//;                   # and strip it out

    my @rec = split /\|/, $header_info;

    s/\n//mg;                      # join the sequence strings
    push @rec, $_;                 # make sequence the last element

    push @seq, \@rec;              # push into array of sequences
        
    $seq{$rec[1]} = \@rec;         # or store in a hash lookup
  }
}

print Dumper(\@seq);
print Dumper(\%seq);

__DATA__
>gi|2695845|emb|Y13255.1|ABY13255 Acipenser baeri mRNA for immunoglobu
+lin heavy chain, clone ScH 3.3
TGGTTACAACACTTTCTTCTTTCAATAACCACAATACTGCAGTACAATGGGGATTTTAACAGCTCTCTGT
+ATAATAATGA
TACCCCCGCGACCTTCTCGTGGACTGATCAATCTGGAAAAGCTTTT
>gi|2695846|emb|Y13255.1|ABY13255 Acipenser baeri mRNA for immunoglobu
+lin heavy chain, clone ScH 3.3
TGGTTACAACACTTTCTTCTTTCAATAACCACAATACTGCAGTACAATGGGGATTTTAACAGCTCTCTGT
+ATAATAATGA
TACCCCCGCGACCTTCTCGTGGACTGATCAATCTGGAAAAGCTTTT
[download]