Both have pluses and minuses depending once again on what you are attempting to accomplish. What is your RPO (recovery point objective) and RTO (recovery time objective) for your DR plan? Does it vary by application or is it uniform for the entire company? What type of DR plan are you implementing? One that works in the case of a major disaster, or one that lets you recover individual files easily and without system admin intervention?
Downsides for each of these solutions can be (not always) a lack of highly granular recovery that comes with continuous data protection (CDP). There are variations of these types of solutions that offer a CDP-like recovery. Server replication typically requires agents on every server and a "collector" that captures all of the replicated data locally or remotely. This can add to your server and admin costs. The SAN-based solution can be fabric based or array based. Fabric based can be appliance based, intelligent switch based, or combinations of the two. The fabric- or array-based solution tends to be simpler at the cost of less granularity for recovery than the server-based solution. Unfortunately, like all generalizations, this one is not always true.
My suggestion is to pick the solution that you are most comfortable implementing and operating, the one that best fits your DR plan objectives, and the one that best meets your budget constraints.
This was first published in July 2005