In my opinion, this is a practice that should be strongly discouraged.  While on average reviewers selected by the author give fair, high quality reviews, that doesn't mean that the unscrupulous can't exploit this opportunity to select reviewers that share opinions that are far from the scientific mainstream in order to get dubious arguments into the peer-reviewed literature.  This is especially the case where the paper is on a contentious topic that is only tangentially relevant to the journal, so the action editor may not be able to easily find adequate reviewers from within their own field.  

As an example, there are numerous papers published on climate related issues in energy, astronomy or general physics journals, which can easily be shown to be fundamentally flawed.  Where the journal asks for the author to recommend reviewers, it does raise the question of how much this contributed to the evident failure of the review process.  It seems to me to be better to avoid the problem ever arising.

Ultimately if the action editors cannot identify satisfactory reviewers by themselves, the work probably doesn't belong in the journal in the first place.

Being able to specify people who *shouldn't* be used as reviewers is, of course, another matter entirely.  

To answer the question directly, suggests the names of reviewers that you consider to have the required expertise in your field and who can give you a rigorous, but constructive review.  Don't choose people you know personally if there is someone equally well qualified that you don't know.  I recall reading that when you receive reviews you are getting advice for free from experts who's time you couldn't afford to buy, so why not attempt to get the most value from it as you possibly can?