Converged infrastructure – the trend of using one piece of kit for some pretty varied tasks, including networking, computing and storage, software and automation – really took off in 2013, mostly thanks to a group of robust virtual SAN players entering the market. Barring an unexpected new entrant, 2014's big names in the vSAN …
I'd politely disagree about HP's VSA. For small/medium business it's great. It's super cheap, can still scale to a reasonable size and now features tiering. Throw in a few SSDs and turn on AO and performance is pretty decent. Usable space due to network RAID isn't great but I certainly would not put HP's VSA in the same category as VMware's.
Agreed about VSAN though, still far too much of a beta product. However if VMware do something like put it in with the Horizon View bundle then I can see it being really useful in VDI environments (and would be a good extra feature against Citrix).
Not really a Virtual SAN but my favourite Virtual Storage Appliance at the moment is definitely Atlantis (the persistent model). If they expand beyond VDI and bring it to stuff like SQL then it could be really great for certain use cases.
I'd agree, as a virtual SAN the HP one is pretty much right. Pricing starts at free with all the Gen8 servers and it can scale very well if necessary by adding real hardware SANs into the mix.
Personally I think Microsoft will win this one within a few years. Hyper-V is spanking VMware at the moment and they have been busy adding storage features to Server 2012/R2 while also beginning the path to cloud integration with Azure and StorSimple. I'll be amazed if Windows Server v.next doesn't roll a lot of this in to allow scaleable storage on every Hyper-V node with automated cloud archiving and migration. If they pull it off they will probably win by default for the same reason they have with Hyper-V - admins are lazy and it's easier to just use what you've got licences. The product doesn't even have to be better, just look at Hyper-V!
Never been a fan of vSAN's for production..
But recently I've rolled Citrix VDIIAB on VMWare, I can see these type of appliances being very useful in that Arena, I'm very interested in their replication etc, as normal methods (such as CBT) simple won't work for things such as personal disks.
Here's the problem I'm seeing with vSANs (from all the vendors), and that's the node licensing costs, especially in the SMB world. I may only need two nodes to carry all my compute/storage requirements, but from a redundancy standpoint that's not enough. So I pay VMware for more node licenses, more maintenance, etc. just to get vSAN reliability, and I start wondering if all I'm really doing is changing the name on the check.
Can't really commit to anything...but watch this space. I am working with some of these folks to get setups even my customers could afford...
Not all of them are licensed by node, I can think a couple off the top of my head that are licensed by capacity (HP's VSA and EMC's ScaleIO). Also VSAN licensing is kind of unknown at the moment until VMware bring the product out of beta.
Also I would never recommend anything less than a three node cluster in any virtualised environment. If you bring a host down for updates/maintenance then your entire environment will be running only on one box which is never desired. I would always quote three lower specced hosts, in which case quoting three nodes for any virtual SAN is enough to ensure resilience (in the case of quorum).
More to the point: 3 nodes is the node count for a VMware Essentials Plus package, which is the more economical virtualisation offering on the planet for SMBs. I include open source in that. I'm rarely one to buy TCO arguments, but [redacted] is vSphere easier to use than the competition.
Great post, Trevor. :-) Have a happy New Year!
Cheers, Jeremiah. Happy holidays to you and yours, too.
What happened to Ceph?
This is a general scalable data store with front-ends to make it act as a block device, a filesystem, or an object store. And it's free.
Or is there some capability which a "vSAN" is supposed to have which this doesn't?
Re: What happened to Ceph?
The "general object store that has a shim to provide NFS and CIFS" market is a whole other ball game. I hope to cover it next year. Short version: speed is the issue. They can't even come close to proper VSANs on IOPS. Not just raw throughput, but latency as well. If anyone has a general object store + shim for which this isn't true, I'd love to take a boo.
The big reason that things like the VSAN sets do so well is the flash acceleration tier to their operation. That, and they share the system they're running on with a hypervisor and so do both virtualised compute and virtualised storage workloads in the same system. At the moment, that looks to be a definably different market from things like Hadoop or Ceph.
I didn't see a mention of vSpex or FlexPod? Both allow a customer to use commodity servers, infrastructure, and templates for storage; and build out a reference architecture without the VCE like premium you say costs a few virgins...
It'll still use the big iron of a VNX or FAS filer, but you have a reference template to build off of that is known to work and can save you a bit of money if you don't need converged support.
Honestly, there are so many virtual storage entrants that to list every single one would be an ebook. I tried to stick to the ones I thought were going to be the major players in 2014.
Zadara Virtual Private Storage Arrays
We're in complete agreement that virtual SANs are poised for a disruptive year in 2014, and really appreciate seeing more attention paid to the virtual SAN space. Perhaps not so humbly :-) we would like to add ourselves to the list of relevant players. We provide virtual SAN and NAS, running on standard, servers, delivered as a service, on-demand and elastically. Our VPSA (Virtual Private Storage Array) service, in full production since Q2 2012, has multiple petabytes of instantly customizable SAN and NAS, with a who's who list of customers. Please feel free to add us as yet another proof point that virtual SANs are already making a significant impact on the Enterprise Storage landscape, on- and off-premise. Happy Holidays, Trevor and all!
Re: Zadara Virtual Private Storage Arrays
Hey Noam. Ping me an e-mail (there's a "mail the author" link in the article, or use the contact form on trevorpott.com. I'd be happy to review your offering and chat with your people about what makes Zadara different and why they might be a contender in 2014.
What is IBM doing in this space ? Hiding under a desk hoping it'll just be a marketing exercise ?
IBM's storage strategy in the entry/mid-range is Storwize and Storwize. :)
They do modular in XIV but no real virtual SAN offering at the moment.
IBMer here, speaking for myself not my company of course. IBM is most certainly not simply sitting and waiting for others to innovate in the "virtual SAN" or broader software-defined converged infrastructure space. Unfortunately, Trevor's article mixes some admittedly nebulous concepts together under the overextended umbrella of "virtual SAN", losing some important nuances of each. And meanwhile my employer is not known for amazing product-level marketing like some of our competitors and flashy startups, but we do play well in all the categories the article mentions. Here's my attempt to clarify.
I believe the term "virtual SAN" should be more narrowly described as a specific *type* of software defined storage implementation that can provide some SAN functionality with software running in VMs on ordinary servers, usually (but not always) with DAS behind them. This is an area where IBM has some exciting developments, including our GPFS Storage Server offerings for file-based "virtual NAS" (if you will), FlashCache Storage Accelerator and Easy Tier Server for block-based caching, and lots more in the pipeline. IBM also works with a few companies like Atlantis Computing to provide "virtual SAN" functionality with features that benefit specific use cases like VDI. And of course IBM has a very close relationship with VMware...
BTW, "virtual SAN" is sometimes confused with "SAN virtualization", which I'd argue is a similar but distinct term used to describe the idea of "abstracting traditional storage array functions into a single common control and (sometimes) data plane". Virtual SAN implementations usually have some kind of SAN virtualization features. IBM was one of the *original* innovators in that space with our SAN Volume Controller product, which is now a very mature offering (ten years and tens of thousands of deployments using the technology) with leading features including real-time compression (from our Storwize acquisition), tiering (Easy Tier), and OpenStack integration.
More broadly, some of the companies on Trevor's list cover more ground than either of those buckets - in particular Nutanix and VCE. I'd classify those companies as "converged infrastructure" or (trendier) "software defined environment" players, and IBM plays well in that space with our PureSystems family of offerings. You get integration all the way up to the "business problem" level if you desire, without losing flexibility and choice. That total solution integration is a key weakness in competitive converged infrastructure offerings which are more often than not just a rack of semi-integrated components. But that's a discussion for a broader converged infrastructure article, not just the storage pieces...
Finally, one of my colleagues wrote a good summary of IBM's place in the SDS world which mentions some of these points and has some pretty pictures too:
All that said, tl;dr: Trevor, calling this article a list of "virtual SAN" players is perhaps a bit misleading, and you missed IBM. :-)
Erik: unless I'm missing a great deal here, IBM's offering isn't a virtual SAN in the commonly accepted sense. There's a lot of weaseling around trying to shoehorn existing product into a new category, but the fact that when anyone I've talked to over the past 8 months talks about a "virtual SAN" they are talking about converged storage and compute where the VMs are being run on the same systems as are providing the storage. Most critically so that one can add both compute and storage to the network simply by adding another node."
Attempts to stretch the definition beyond that, or twist the definition to say "that's not a virtual SAN, this is" are reminiscent of nothing so much as hangers-on putting time and effort into muddying the term "cloud computing" to make themselves seem relevant. They certainly managed to muddy the waters, but...they didn't really win that battle either, now did they?
If IBM actually has something that plays in the area, I'd be more than happy to take a look at what they offer and include them in a follow up article sometime in the next few months. So far I've seen nothing that fits, despite having looked. I'll go look at GPFS Storage Server one more time, but as it was demoed to me before it was emphatically not a virtual SAN.
Also: "virtual SAN" has bloody nothing to do with SAN virtualisation. They are completely distinct concepts. I think anyone serious about storage will be able to tell the difference between A and B here. They aren't "similar" in any way.
Virtual SANs are about A) commoditising storage by making it possible to run it on the commodity compute nodes we already use for visualization and B) increasing speed/decreasing latency by putting the storage a VM uses in the same compute node it runs on while still offering full redundancy equal or superior to that which you would find in a traditional SAN.
SAN virtualisation is about taking a traditional SAN and cutting it up into logical SANs and delivering these logical SANs to compute nodes across the network. It doesn't drive down the cost and it doesn't move the storage closer to the compute.
Again, here I do not see "Pure Systems" as being remotely the same. It's like VCE: old school infrastructure that is shipped as a single blob. It is thus "converged" in enterprise-tech-speak, but is not "converged" in a practical sense, nor does it have any of the advantages of a virtual SAN. It's just...wanting to be VCE. For whatever that's still worth. For my money, Nutanix is going to crush VCE like a bug.
Trevor, I appreciate the thorough response. I did not mean to kick a hornet's nest here. :-)
I agree with your "virtual SAN" definition as elaborated upon in your comment, and I too hate it when vendors muddy the generally accepted meaning of such terms - it hurts everyone in the industry. I was simply trying to provide more color for the earlier commenter who did not appear familiar with IBM's efforts in the universe of things that some people might associate with the term "virtual SAN".
Regarding your above PureSystems comments: I agree that PureSystems is not (currently) a vSAN play - rather it is a converged infrastructure play (as identified in my comment). But you named VCE as one of "2014's big names to watch in the vSAN space", so surely PureSystems would not be out of place in that list as well... right?
Regarding your above SAN virtualization comments: I never intended to imply that virtual SANs == SAN virtualization technology, but I have encountered many less-than-perfectly-informed customers who mix the two concepts due to lack of knowledge of one or the other, so I tried to clarify that distinction in my previous comment. BTW, since you expressed a dubious opinion about the value of SAN virtualization... actually our SVC virtualization platform does much more than just cut a big SAN into baby SANs, as you imply above. For example, SVC can provide inline real-time compression, which really can save lots of money with the right workloads, even if the SVC hardware and software is an incremental cost above the customer's existing array licensing.
"But you named VCE as one of "2014's big names to watch in the vSAN space", so surely PureSystems would not be out of place in that list as well... right?"
Honestly, no. As I said in the article, I was highlighting those I thought would be major players. In the case of VCE or Nutanix because of raw market domination. In the case of someone like Maxta because they are fundamentally game changing via commoditisation.
If I tried to list every single CI or VSAN player it would be an ebook, not an article. There are well over 100 at last count; the whole area is growing at a level that makes me think of the mobile device management market. That's why I narrowed it to companies I think would make a bang in 2014 here.
The sad truth is, I just don't think IBM is currently very relevant here, nor do I believe they have anything innovative enough to become relevant in the future. You'll have to forgive me here, but I see IBM's PureSystems setup as a "me too" play that mostly will have negative appeal outside IBM's captive market and stands a damned good chance of having even IBM's captive clients turn to VCE or Nutanix.
As for "SVC is more than cutting a big SAN into little SANs because compression"...I must be missing something. Compression and deduplication were neat in 2006. It was right around then that this "younger, hipper feature" emerged that let people defer SAN refreshes for a cycle. Everyone of any value does this today, it's not exactly a point of differentiation. Indeed, today's deduplication is host-based flash caching: add something like Proximal Data to the host and use it's internal flash to buy you extra IOPS. Ultimately, this can save you from needing an upgrade RFN if you're at the IOPS wall: buy you a year or two to get your act together.
I don't believe there's a "dubious value" in SAN virtualisation...just that SAN virtualisation is pedestrian. I can cut a Starwind SAN into baby SANs, if I massage the interface! It's got HA, deduplication, what-have-you. The past is the past: not only is none of that special, it's commodity now.
Big Blue is many things, but it should never allow itself to become merely pedestrian. Surely a company that creates subterranean seismically-stabilized research labs in order to create atomic-resolution scanning tunnelling microscopes capable of demonstrating the strong force has more virtual confetti and storage oomph than SAN virtualisation and inline deduplication to sprinkle about the holidays!
Your point that people mix the two concepts up, however, is well taken. I really should do an article to explain the difference. It's amazing how much interest this one article has driven. I've a lot of interviews to do next year with both vendors and end users. I look forward to seeing what 2014 brings and to seeing how everything unfolds.
I've no ego in being "right" here. I called the players as I saw them and there is just as much information in being wrong as in being right. If I'm wrong, I'd have no problem saying "hey guys, IBM actually did some decent cleaning up and/or innovating in 2014!" Indeed, insert any name in place of IBM: I've no bias about winners or losers here...but no, I don't think IBM is going to make a bang in this area in 2014.
2014 is going to see the old guard fight viciously to defend their aging product lines and business models as newer, more capable companies redefine entire markets. Cisco will be fighting a noisy rearguard action against all comers; everyone from the OpenDaylight project to Juniper to IBM are pushing SDN offerings that will truncate Cisco's relevance and collapse it's market share.
Microsoft will continue to machine gun itself in booth feet while Oracle doubles down on extracting every single bent copper from it's hostages. EMC will continue to bleed market share to the likes of Nutanix and Tintri while VMware will be forced to completely reinvent their pricing model or die.
None of this will mean the death of giants in 2014...but the giants look ever so much more inflexible today than they did only a few years ago. A sea change is coming. With luck, it will not be a sudden collapse and refactoring of the market like the dot-com burst. Instead, it seems likely that the great behemoths of IT will fade slowly, going quietly into that good night over the course of the next decade or two.
But they have peaked. They have allowed themselves to become wrapped up in pride and hubris; to let innovation occur beyond their borders to an extent that is shocking. Some - like VMware - simply clone the ideas of their own "partners" (the lawsuits once the patents are granted will be hilarious good fun.) Others attempt to buy up these companies one by each and lament their failed integration as the soul-destroying corporate culture of the IT behemoth evaporates the very things that drove innovation in those startups to begin with.
There is a divide here. The staid, high-margin, enterprise-tech-driven model reliant on "services" and 4-hour engineer rollouts is meeting the world of DevOps. Like a freight train meeting a mountain.
In the new order everything is dynamic. Everything is scriptable. This new world sees 4 hours as "not nearly good enough" and requires double and triple redundancy for everything. The new world is burstable. It is programmatic. It is responsive and automated. It is built on commodity everything and change is a feature, not a bug that needs 5 months of change management to plan around.
This is the transition period. It is the time where the old school IT market starts shrinking and the vendors that supply it start winking out one by one: going bankrupt, getting bought, or leaving the market and doing something else. Oracle has proven you can dine on that for some time...but the future is being written by the likes of Puppet and OpenDaylight, not Cisco or VCE.
So, against that backdrop...why would I see PureSystems being a big player in 2014? VCE is the name enterprises and governments trust when it comes to old-school infrastructure. They'll give you a pod of stuff that Just Works and support it in a traditional enterprise fashion. All the big names are on board, it ticks all the old-school CTO paranoia boxes.
Anyone looking for futureproof IT isn't going to be looking at anything VCE like - not from VCE, IBM, Oracle or anyone else - so what chance to any of these other enterprise players have?
IBM's version of "innovation" seems to be "the all flash datacenter." That's great, if you happen to have a continent full of virgins and access to one of the three volcanos on earth with a phonolitic lake. Even for the Fortune 500 that's a bit rich. None of that, of course, addresses how one is to overcome the bottlenecks or latency issues inherent in centralized storage or the transition from North-South networking to East-West networking. (IBM does have some OpenFlow stuff that shows promise, however, it does seem to require that you have a stead supply of virgins to afford it.)
IBM is - to my mind - like the US military. It's really quite spectacularly prepared to fight the last war it was engaged in. Every now and again the R&D department vomits up something truly amazing (like the internet)...but it's the wars of right now, today, that it can't ever seem to quite win. Every now and again both will hold a "Mission Accomplished" ceremony, but disentangling real world victory from meaningless marketing mumbo jumbo designed to fool the plebes is itself a full time job.
IBM is working on hundreds of astonishing new technologies that I think will make the latter half of this decade amazing. (Some of the AI stuff begin done with WATSON, for example.) Storage, however...just isn't one of them. Yet.
Here's hoping that for all the length of that comment, I'm wrong multiple times. Life's far more interesting if you can't predict it quite right. Cheers! Have a great new year. :)
I have commented on Maxta before, but that stuff is great. We have a few clients using it. Still may not be ready for a large scale-out environment, but the product is solid. As pointed out in the article, you can add nodes and storage as needed, they are not correlated. It is basically a hybrid array you can run in sw.
Beware the network guys!!!
*** Disclaimer*** Riverbed employee and formerly at a storage vendor. ***
From Cisco we got the USP platform and that at the last glance (prepared to be corrected) was the market leader in server build out infrastructure.
Riverbed will be pushing Granite hard in 2014 as a storage mechanism where your multi million dollar storage array sits in the DC and services the branch office by projecting the storage out from the DC. No more moving parts at the remote office (Storage vendor OS, firmware, disks, cables, floor space, networks, power...), no more ogranising site visits or trying to get tapes back to the mothership.
Shameless plug? Absolutely. Storage arena game changer? Who knows, but it's worth noting that companies are moving out of there prefered "traditional" territories and looking at other areas.
It doesn't have to be a startup that changes the landscape...
Re: Beware the network guys!!!
@bear_all I agree with you 100% that traditional companies can evolve. Do I believe that 2014 will be the turnaround year for Cisco? No.
The reason for this is that Cisco simply isn't yet willing to cannibalize it's own core competency yet. It's still fighting ridiculous internecine civil wars and is as yet unable to meet the challengers at the gates.
That said, of all the tech giants, I believe Cisco has made the best acquisitions recently, (though I can think of at least five others it needs to make. Preferably now, while the startups are cheap.) What remains to be seen is whether or not Cisco's internal bureaucracy manages to strip away the "souls" of those organizations, or whether it will truly be able to undergo a cultural metamorphosis which allows it to actually become an agile competitor once more.
Cisco has spend the past decade and a half bleeding it's top minds. They have turned around and created the very technologies and companies that will be Cisco's downfall. (Arista, anyone?) They left Cisco for the same reasons that VMware has been shedding bodies for the past 6 years, and why IBM went through yet another purge in the augties: the company culture became staid, stiff, resistant to ideas from "the muck" of the employees.
Workers were to be drones carrying out the grandiose vision of the Supreme Ruler and his core staff of imagineers, not innovating and adapting on their own. Even executives had to cut through red tape with a machete, ultimately leading to a bleed of top talent.
Other companies can do almost all of the things we used to turn to Cisco for...but for far, far cheaper. Cisco's biggest mistake is fighting desperately to keep margin up on those areas. Especially trying to use lock-in tactics and walled gardens to do so. Buyers are wising up to that crap.
For Cisco to keep margins high it needs to be able to find a new market, become the best of the best in it, charge it's margins there, while dropping it's margins elsewhere to "market margins + name cachet". That will give buying into the "complete stack" of Cisco real value. Unfortunately "name cachet" is not the 2x-3x multiplier Cisco believes it to be.
Cisco has been trying to reform. To change this about itself and to expand it's areas of operation such that it will survive the collapse of it's core networking market. Has it done so yet? No. Will it manage it? I honestly can't say. Will 2014 be the year it returns to being an adaptive, innovative behemoth that will redefine markets? I sincerely doubt it.
...but 2015 might. It really just might. Cisco is definitely on my "companies to keep an eye on" list because unlike many of the others it is trying to change. It is simply a question of whether or not Cisco manages to do this before it bleeds all but a hard core of it's customers.
- ASTEROID'S SHOCK DINO-KILLING SPREE just bad luck - boffins
- BEST BATTERY EVER: All lithium, all the time, plus a dash of carbon nano-stuff
- Stick a 4K in them: Super high-res TVs are DONE
- Review You didn't get the MeMO? Asus Pad 7 Android tab is ... not bad
- FTC to mobile carriers: If you could stop text scammers being jerks that'd be just great