Storage Replica and Storage Spaces Direct Were Killed By Licensing
The licensing announced in Dec 2016 ( http://www.microsoft.com/en-us/server-cloud/products/windows-server-2016/ ) for WS2016 killed Storage Replica and S2D before GA. It makes absolutely no financial sense to deploy these features if Datacenter is required. It would be cheaper (and an easier sell) to go with the incumbent non-commodity hardware solutions that S2D and SR were meant to replace. The math is easy.
Any hope in seeing S2D coming down to the Server Standard SKU. Even if it is limited to four nodes, that is still better than nothing for smaller business running a few VMs. Datacenter licensing does not make sense for anything fewer than 14 VMs per node.
Limit is tiny, 2TB, single hard disks are barely that size today! Use case for such a small volume will be very limited and niche. Would be useful to have larger / no limit. Just limit node qty
Richard Stretton commented
Limit is so tight its not very useful. 2TB? Ouch! As the price per TB is in free fall, a TB restriction is like an expiry date. If SR has to be limited in standard then node / member limitation makes more sense. SMB here uses 4x2tb NVMe SSD's for production in RAID 5, and some backups on to 8x12TB disks. 2 nodes, that's all they ever need, nothing special, not major budget stuff, but both volumes happily exceed 2TB. In fact on modern drives I find it hard to imagine a volume of 2tb or smaller.
Storage Replica now included in Standard edition in limited version, in Windows Server 2019. See coming preview build. :)
I agree with the other comments. S2D & SR limited to Datacenter edition puts it way out of reach for SMBs or for use in a Remote Office/Branch office situation. Even if they provided it with limited node and storage on Server Standard, that would be fine for these use cases.
Brent Vierling commented
I was looking forward to Storage Replica being brought to Windows....until I saw DataCenter requirement. Hopefully this will be changed in the next version!
The beauty of Microsoft has always been that they make enterprise features affordable for the SMB. Not so with this decision. As an IT Manager of a SMB, I fully agree with Aidan and I am extremely disappointed that Microsoft chose to make these game changing features available to it's enterprise customers.
Benoit Tedeschi commented
If you plan to host desktop VM (windows 10) instead of server vm, the hyperconverged scenario is not suitable too :(
Esben Barnkob commented
The datacenter license is likely already purchased since your cluster contains or will contain, so many VMs that you need the datacenter license anyway. Also the datacenter license can be transferred from your previous virtualization platform if covered by SA.
People need to compare this to VMware vCloud Suite + VSAN Enterprise + Microsoft CIS Datacenter, and suddenly the "Datacenter licensing requirement" is no longer crazy. Actually there is a great business case for dropping 2/3 of the licenses of VMware, which amounts to about 2/3 the cost of the virtualization layer.
Andrea Matesi commented
In order to get the solution out of the door (or "out of the window" should I say?), I guess MSFT should offer these features as part of their "Windows Storage Server"-family of products.
That said, it looks like a very typical MSFT move (MSFT did the same w/CAs-PKIs - "only available on Enterprise at the beginning).
Ryan Helmer commented
Why on earth tie this to Datacenter? There are many of us that do NOT want our VM hosts handling both storage and compute duties. $25k in licensing--and that's before hardware costs--means a Windows-based storage system never even makes the shortlist. Look at the numbers and tell us how we can justify the cost vs. competitors. It's just not possible. It only makes sense in an HCI environment, but why condemn this solution to that niche market? Wasn't the whole point of the Windows Storage stack to show the compelling performance and cost of commodity hardware and Windows software?
Please don't dig in your heels on this; please find another way. The people caring enough to comment here are the ones that have been the main cheerleaders for Windows-based storage systems. This decision leaves the people that have been selling this without a reason to recommend it.
Agreed, there needs to be a limited (at least by the spirit of the license) version of these technologies in the standard version of Server. With the destruction of SMB server MS has already limited small companies as is. Things like failover servers and shared storage are not only for huge corporations any more.
They are killing their own very good product... what a shame..
Hein Oudshoorn commented
First SBS now this. Just admit all microsoft wants is put SMB in the cloud.
David Tocker commented
We will give our money to HP then - I can't justify $26k of software to replace $35k of HP StorVirtual including the drives and the two servers that it runs on.
****, I can spin up a baby 3PAR for that money.
I was looking at commodity servers and HP's cloudline, but its back to the DL360's and dedicated storage.
Vmware Virtual commented
This is die-hard good feature on server 2016 , but i think the market share won't expand due to this restriction. The better way to balance between customers and MS is Standard Edition can enable these features but with limited node to accomplish it .
I agree with Aidan.
Microsoft brought, in my opinion, one of the best solutions in 2012 and ruin it going forward in 2016.
I really thought this was going to become a mainstream solution with more hardware vendors supplying components.
But now there is no reason to bring it up to small businesses. Actually there's no reason to bring it up to large businesses either.
Andreas Furtenbacher commented
You couldn't kill S2D and SR more effectively as with this decision. Imagine a "non-hyperconverged" S2D installation! 4 servers all with Datacenter, just for storage? Definitely not! The big storage vendors will send flowers to MS for this...
Mike Wynne commented
I was really looking forward to this. I was even looking forward to building a NVMe based SAN. I had originally planned to do this with GlusterFS on CentOS servers. After reading of the new server 2016 features, I was sold on S2D and replica. We have a 3 server Hyper-V Cluster w/ an offsite Replica that may get migrated to Xen server as result. Since RDMA and resilient file handles are not support in Samba we might not have a choice, once we move from our EMC to a Linux based SAN.
I was really loving the way Microsoft was moving with this and some of its newer technologies. They would have eventually made huge ins with the SAN storage market. Instead you will see more Linux DYI systems popping up as GlusterFS, ZFS, Ceph and Samba begin to mature. They have a few missing features in samba 4 that should be implemented this year. ZFS is rock solid and GlusterFS is getting better every day. Vendors like Kaminario will also be much better choices for SAN vs the ridiculous licensing costs incurred by these wonderful features on datacenter only licenses. I really hope they change their minds. While Xen is nice, I really like the stability and ease of use in Windows Failover Clustering these days. The Core licensing while a bit steep does not bother me or shock me near as much as the Storage features being placed in datacenter. The cloud features I am perfectly fine with them tacking on to datacenter.
Well, that certainly prices it out of range for the use cases I was thinking of. I'm sure that there's environments where this may make sense... It's just no longer an easy way to move clients off existing poor storage into a fail over environment if they are budget constrained.
I was really hoping that Microsoft would use this as a way to encourage smaller business to transition off vmware and onto hyper v. The cost of performing a migration + licensing for the same features (Which an included vsan would have broken parity with VMWare) makes it really hard to get existed about changing infrastructure, and seeing as our clients have been sold vmware for the past five years exclusively, retraining and menu costs without a deal making feature is hard.
What I'm honestly surprised by is that they don't just limit scale out to a number of nodes or amount of storage, but instead kept it entirely removed.
Probably will check it out on preview 4 though, because maybe they'll license it a la carte or something reasonable between now and release.