Remote block devices work perfectly fine provided that they are single-access. Yes, there may be increased "latency" to fetch the block if the cable is longer a fifty foot SCSI cable has 100 times the latency of a 1 foot SCSI cable and if changing the "cable" from copper to aluminium will also increase latency (signal propagation in aluminium is slower than copper) as will replacing the copper wires with glass fibres (signal propagation in glass fibre is significantly less than signal propagation in copper). Multi-access remote block devices will have the same problems regarding multiple update consistency that affect remote filesystems -- they can be made to work perfectly but if designed to do so then they become so slow as to be unuseable. The issue really boils down to the "distribution" of the responsibility to maintain consistency and how long it takes to do this properly. Clearly at some point the "time it takes" to maintain consistency amongst multiple accessors of the same resource has its effects. Keeping that time to a minimum by keeping all the multiple access local is the most efficient.