Alas, sync/semi-sync seem to have either been deprecated either due to a limit in cDOT/Spinnaker and/or in an effort to sell more MetroCluster. In the end, we never had a use-case for either in 7mode and the closest we have in ONTAP 9 is more like 15min. Anyway, if your RPO is really down to 0, MetroCluster might be the best bet. If you can survive with an RPO in single-digit minutes, you can definitely experiment with more frequent SnapMirror replication - we've had success down to 5min or so. It really comes down to the quality of the network between your two sites and the rate of change in your data. We've had our 15min jobs "stack-up" when a huge amount of data gets written to the source - the nice thing is ONTAP just logs a "snapmirror update failed because replication is already underway" message and things move on. It gets caught up by the next run and things are good.
In an NFS scenario, you could even play around with XCP running in a near-continuous-scan mode - almost like running DoubleTake outside of NetApp constantly replicating changes from Point A to Point B. I have no idea if that's a use-case the XCP team envisioned with the tool, but it might do the trick as well.
Hope that helps. Maybe a smart SnapMirror guru can chime in with the maximum replication frequency for SnapMirrors - I can't find anything other than limitations on InfiniteVol and transition mirrors (i.e. 60min).
Chris
