You really don't want to go down the SnapMirror sync or semi-sync route with the requirements you describe. SnapMirror is a great replication solution if you can live with an RPO (Recovery Point Objective) of one hour or more. Our customers who use SnapMirror are typically in the 1hr-24hrs RPO category.

But SnapMirror sync and semi-sync are architecturally slightly different and try to mold SM into something it was probably never really designed to do, and come with their own set of (performance) problems. Of the very few attempts we had at getting synchronous SM to work at customers we've always had to eventually implement a different solution. Very powerful oversized controllers are really the only workarounds to tackle some of the syncSM shortcomings. I guess recent controllers with the larger NVRAM will help, but the reality is that syncSM never really took off and is probably a dead end NetApp technology.

Have a look at the Metrocluster solution from NetApp, which uses SyncMirror as the underlying "protocol" to essentially give you mirrored RAID-DP aggregates spread across two different locations.

In 7-mode, it's a very robust architecture and pretty well known and supported. It's a very popular architecture over here in Europe: we have several tens of MCs in the field amongst about 200 customers. I've been told it's less common in the States, perhaps because it's officially limited to distances of about 160 km/100 miles or thereabouts. Maybe that's more of an issue in the USA.

In Clustered ONTAP, metrocluster will be supported (with a different design) as of version 8.3 which will be out in november.

This is your biggest worry IMHO, if after having looked into this a little further you find yourself convinced that MC is the way to go: the dilemma of either going for tried-and-trusted 7-mode MC, which means no more new features in the coming years, or a plunge into Clustered Data ONTAP territory with a brandnew and largely untested MC in cDOT 8.3. You really don't want to miss out on cDOT nowadays, but you should expect some quirks with MC-on-cDOT that could realistically take a couple of months before they are ironed out.

On Thursday, October 2, 2014, Rhodes, Richard L. <rrhodes@firstenergycorp.com> wrote:

Hi Everyone,

We currently use some low end IBM NSeries for a specific application.

All of our other storage is from a large three letter vendor.

We have never really considered NetApp for our high end Tier 1 storage.

For our T1 three letter vendor storage we replicate bi-directionally between two datacenters. Years ago we used synchronous replication, but at the last rollover have changed to async replication due to write latencies. It’s worked very well. The storage systems are roughly 5-10 seconds out of sync.

I was looking at what OnTap supports for this kind of work. I see that in 7-mode it supports sync and semi-sync. Semi-sync seems interesting for what we do, except that you have to replicate within a Consistency Point. You have until the other half of nvram fills up to complete the current CP with replication. When I looked at C-mode, I can find no mention of sync or semi-sync at all.

Q) Does C-mode support semi-sync?

Q) If not, can normal async snapmirror support replication into the 10-15 second lag range? (note – T1 storage is almost all large Oracle DB’s)

Q) Any other idea on how you would implement a low lag async replication with NetApp?

Thanks

Rick

The information contained in this message is intended only for the personal and confidential use of the recipient(s) named above. If the reader of this message is not the intended recipient or an agent responsible for delivering it to the intended recipient, you are hereby notified that you have received this document in error and that any review, dissemination, distribution, or copying of this message is strictly prohibited. If you have received this communication in error, please notify us immediately, and delete the original message.