Iron out the kinks and all-flash servers just might have a future Nutanix's NX-9000 is an all-flash, scale-out, hyperconverged server/storage/networking system, with a single logical pool of SAN storage aggregated from the NX nodes in a cluster. In other words, it's a clustered, disk-less server. It's disk-less, and so less prone to mechanical and electro-mechanical errors and needs less …
Disclosure: I work for VMware in the EVO:RAIL Team :-)
Interesting post. Firstly, I'm not sure if an all-flash based system is requirement to make hyper-converged a true alternatively to conventional SAN storage. So long as the working set (r/w) remains in cache, and writes are de-staged to disk - its not clear to me why an all-flash array or all-flash hyper-converged system would be that compelling. Unless the customer has very deep pockets, and wants an absolute cast-iron guarantee that they would limited by a spindle. Of course, the are customers like that - otherwise these products wouldn't exist. As you rightly say, for the cost per GB, you would need a very compelling use-case...
As with all these things - whether a technology can truely upset the market depends largely on the use-case. The customers see value in both conventional server/switch/storage - which allows them to add additional disk shelves for increased capacity - without buying more compute (that they may not need) - in what would be a classic scale-up play... Where they see value in hyper-converged is in SMB, ROBO and VDI - where they want an easy deployment option with relatively low on ramp compared to other infrastructures - and where the compute and storage requirement scale out at the same rate - as is the case with VDI.
I'm not sure you analysis of the other vendors passes muster in all cases...
Cisco - have recently signed a distribution deal with Simplivity
Dell - is both an VMware EVO:RAIL Qualified Partner and also Nutanix supplier - in the main both a hybrid models of SSD/HDD
EMC - Will have an VMware EVO:RAIL system by Q1 of next year, they also offer Scale IO as server SAN (although I see you mention that...)
This post has been deleted by its author
I understand the point of this article. The kinks are in large part why my systems still run SAN (and disk less -- boot form SAN too). I don't see this changing in the next 1-2 years anyway.
Lots of companies out there have been trying to integrate the server side end with arrays, it's not easy to do, and gets almost exponentially harder when dealing with writes and clustering. That's a lot of logic required, a lot of complexity.
I was pretty excited when Qlogic announced their Mt Rainier technology mainly because they said they were going to offer cached writes too. My workload is 90% write from a storage perspective, so stuffing ssds in the systems for read caches isn't going to do crap.
For a while at least my talks with QL didn't get too far and they did not have the write caching ready yet (with no ETA at the time this was 1-2 years ago). Looking now it's hard to tell if that is done yet or not(they still talk about it in their documentation), and in any case I'd feel most comfortable deploying something like that if it had support from the SAN vendor as well (array based snapshots data consistency etc).
But for me, at least for now it is a moot point, my all flash 3PAR 7450 arrives in a couple of weeks and that will allow our systems to scale nicely, without having to worry about complex sophisticated clustering software on the hosts to offset the performance penalty of using spinning rust. Our data sets are small and are not growing very quickly.
So, people have been working on this for years already.....
Nate... big fan of your comments. You have influenced some of my purchases, but why did you say?
> I don't see this changing in the next 1-2 years anyway.
Is it because your company still needs to depreciate your assets or are you waiting for a vendor like HP to come up with a comparable offering?
@Vmware. How many vSAN nodes can you have ? 16 maybe 32 soon. Nutanix unlimited and just scaled up. Is that why vmware asked Nutanix and Veeam not to come to vmworld few years ago because they know nutanix and other vendors like Veeam were blowing them out of the water. Just Google for vsan horror stories. I have worked with vsan and Nutanix. Clearly vSAN has some catching up to do. Don't be the guinea pig by going with latest vSAN. Nutanix Replication Factor of 2 or 3 is worry less.
Tehehe... Nutanix employee? Not surprised. But... you are right. This is the same reasons why Microsoft takes their time to win a space. It takes a couple of revs to get everything correct. For example: Windows 3.1, Word, Windows Server 2003, IE 6, Hyper-V^H^H^H^H^H etc.
Question is whether Nutanix will be able to maintain their lead vs. VMware as VSAN gets mo' better. I think so. Our channel partner talked about their roadmap. I can't say much, but it is very exciting.
Veeam is a different beast. They are just out there to disrupt the market -- rules of IT vendors be damned. The winners are Veeam and the IT organizations that use them. LOVE 'EM. Not sure if Nutanix is in the same league of disruption, but I am confident they will rise up to the occasion. Their deal will Dell could prove they are willing to give the rules a flaming finger. As an investor, I would think this is the same as investing in the likes of a Google or Salesforce.com.
The group of vendors that should be most nervous are the low-end SAN vendors like HP, NetApp, Nimble, and so on. They are most susceptible to VSAN, Microsoft Storage Spaces, Maxta, and Nutanix (especially if they go 100% turbo^H^H^H^H^H I mean software). I am excited about 2015.
A word of advice to the "flaming finger" vendors -- please don't be a nay sayer. Especially about each other. You need each other. Focus on your common foe (... and help me get rid of my pesky colleague who loves the sound of his MDS switches).
All of the companies you mentioned can perform the "acquire a startup" maneuver. It's just software and someone is probably already developing it as we speak.
As far as networked storage arrays for the bulk of the inactive data - I'll just put this here: http://www.theregister.co.uk/2014/07/10/the_cloud_gateway
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
