VMware has launched its virtual storage area network (vSAN) technology; pretty much unleashing a raging bull into a china virtual storage appliance shop and watching all the crockery racks of electronics go flying. Available now as a beta, vSAN can create a virtualised storage array that pools the disks and flash drives of a …
Just don't get the excitement
For some this will be just what they need. If you've got a pretty dumb lump of disk you need to stretch a bit more functionality out, then hell why not?
But Datacore have been selling this has an approach for years. Lefthand has probably been about the most successful of late. ONTAP Edge has never particular set the market on fire. These solutions all work well in maybe a remote environment where you say one or two fat servers with disk in are at a remote site, but for core DC platform? Some of the new ones, who knows, good luck to them, but in 2 years time will anything be fundamentally any different.
You can argue these appliances might worth in xSP environments and thats true to some extent. But you still need to manage the underlying physical disk pools and the VSA does not change where or how you do that.
A final puzzler as I see it - if was going to be so successful, why did the backers of ScaleIO sell it for relatively small price of $300m to EMC? $300m to the man in the street is a lot of cash - for EMC its probably what Tucci spends on his jet and 5-tassled loafers a year. Personally I think the ScaleIO backers realised it was going to be a very tough sell to the man in the street, even if it made great marketing copy and got out before they had to get it to market. I would point the same at XtremeIO to be fair...
Re: Just don't get the excitement
A voice of reason in a sea of insanity..
We used to call these "file servers"
Seriously, what's the big deal here? These used to be called "file servers" but in the 21st century they're "network attached storage"
Either way, if you have direct attached storage on a VM host that you want to expose to the rest of the cluster, you run an NFS or iSCSI server in a virtual machine and call it a day. OpenFiler and FreeNAS both do the job.
Hollis blog highlights big shifts in thinking...
I was struck, upon reading Hollis' blog, just how disruptive this VSA can be. if you haven't, read the blog.
First off, it does away with the requirements for any kind of dedicated storage hardware AT ALL. No more Fibre Channel, iSCSI, NFS, SMB, physical disk pools, array-based software, LUNs etc, you name it, that goes away because the VSA utilizes the drives on compute nodes (aka the same commodity servers that are running ESXi) and only handles VMDKs.
This is a really smart idea on at least four levels:
1) Simplicity: It sheds a lot of unneeded complexity and works with what's there. This happens in many areas, but to name a few, support for third party storage, management plug-ins, dedicated networks, etc etc. You can now manage your VMware resource pool across a single pool of hardware from a single console without worrying about full compatibility with VMware -- an especial concern for folks who have been stuck with legacy storage. It also works with existing storage -- so if the existing compute servers are SAN attached or DAS attached -- no problem, the VSA encompasses that.
2) Performance: Now that reviewers have proven that a single commodity 2 socket server can sustain 400K IOPS, it will seem pretty sensible to stop buying storage to drive performance. VMware's own proof shows a pool delivering more than 100K IOPS -- plenty for many use cases.
3) Risk mitigation: You can mix and match, provision and deprovision server hardware as you choose, without the complications of storage provisioning. Vendor lock-in at the hardware level is a thing of the past. It's darn easy and has little risk to swap out servers in a compute pool, much less easy and much higher risk to rip and replace your storage array.
4) TCO: It's licensed per-socket, just like VMware, so the hidden costs of capacity licensing go away, as does the complexity, as do the costs of array side tool licensing.
Sure, it's v1.0, it has limits and restrictions, it needs a lot of testing in the field to be fully inspiring. But guess what? I suspect VMware will be driving it because it's a better moustrap.
The different between niche and mainstream
And another thought I should have included in my earlier comment.
Loads of commenters have weighed in, saying "so and so had this three years ago, company X implementation is better, what's the news here?"
No doubt the tech has been out there, as the article points out. That's not the news.
The news is simple: Over time, VMware will be making vSAN standard operating procedure for much of their install base. There's a BIG difference between niche vendor and mainstream vendor, third-party add-ons and vendor integrated technology.
Imagine a small vendor coming to you and saying "I have a widget that will make your car run faster, longer, for less money". Sounds good, right? But you look into their offering and observe that it has significant up front costs, and that if it didn't work, it would wreck your car. Plus you have no experience with their widgets, and no idea if they can service the widget well.
You'd be a little apprehensive? So would I.
Now imagine a major car manufacturer coming to you and saying "I have a widget that's built into your car, which will make your car run faster, longer. And oh by the way the widget will be covered under your car warranty. It might not be the perfect widget, but it's OUR widget -- and you've already bought into buying our car."
Assuming you like the car, odds are good you'll like the widget.
Assuming you've accepted the VMware value proposition, and VMware has become your trusted advisor, it will be super easy to buy into the vSAN proposition.
This is the same strategy Microsoft (and hundreds of others vendors) have pursued successfully for years. Let little guys establish the value of a new technology, then add it to your product, let your brand equity make it a credible solution, and rake in the Euros while niche tech becomes largely unnecessary.
If I was a niche SAN vendor who touts virtualization as their major use case, I'd be thinking about my positioning today. No doubt VMware vSAN isn't a threat for the moment -- but that's what Netscape devs said about IE 1.0...
You should look at Maxta. Similar to the vSAN but they present everything as an NFS datastore on a VMware cluster basis. Lots of enterprise functionality like read/write caching in flash, zero footprint snaps and clones, dedupe and compression. We tested it for a client and came away impressed. We have not recommended VSAs in the past because the performance is terrible. We really like Maxta though.
You got a link for this?
StorMagic SvSAN support for Hyper-V
Chris, StorMagic' SvSAN now supports Hyper-V.
Has VMware sunk a torpedo hole into Nutanix ? Maybe that's why they are hedging their bets with Hyper-v options....
Lot's of innovation for those that still believe the the earth is flat.
Try running Windows Server 2012 R2 with Hyper-V and the 'free' Storage Spaces in a scale out Cluster-in-a-Box hardware platform. Then let's talk the true value of performance/resiliency -to- TCO/ROI.
vSAN is not a VSA
This vSAN is not a VSA infact you can enable it simply by checking a box. An great overview of vSan can be found here: http://www.yellow-bricks.com/2013/08/26/introduction-vmware-vsphere-virtual-san/
Personally, I think the vSan is a great addition next to your existing storage, you can add low cost SSD local disks in your servers and use vSAN to create a Tier 1 Datastore to store swap files