I'd always opt for replication over a code solution. If you can keep the data divided so that it only get's written to on one server, that's optimal and you can stick with transactional replication which is simple and dependable.