The VMware, Nutanix mud wrestle is hilarious, but which one is crying with fear on the inside? So Nutanix is in yet another spat with VMware. The big ones are getting to be a yearly affair. That's great for them: these little soap operas seem amusing, but underneath it all, there are some very serious issues being hashed out. VMware wants control of our data centers. The whole stack, top to bottom. It wants to be the …
Nutanix is a noisy startup with VERY limited IP. As the article notes, there isn't anything Nutanix does that a dozen other companies either do already or could do very easily. VMware may be annoyed with how noisy Nutanix is, but they are not scared of Nutanix at all. ZERO.
Untrue. Nutanix is making very real wins into VMware's customer base. They are not pleased with Nutanix, which is why VMware are constantly putting time and effort into thwarting them. You don't waste millions on trying to sabotage a company that doesn't matter. You don't have an entire section of your partner portal devoted to defeating a company that doesn't matter. You don't kick a company out of your conferences that doesn't matter.
What Nutanix does can be done many companies, but it takes many companies to do all that Nutanix does. More to the point, Nutanix's customers are overall more satisfied with Nutanix than they are with VMware. (With a few notable exceptions.)
So I agree wholeheartedly that Nutanix aren't the be-all end-all, but they absolutely have VMware's attention, and they absolutely are drawing a very public - and expensive - fear response.
And? I am sure 90% of VMware's customers are worth way less than their top 5%. As it is for most companies.
*shrug*
What is it with the VMware religious types that they assume that because I think VMware needs to take the high road I am automatically 100% pro Nutanix for everything? Lotta black-and-whiters out and about.
re: 50% of revenue coming from a small account base.
AND of their US federal sales most of it comes from the DOD and most of THAT comes from the U.S. Army. this comes as no surprise as more than half of the Nutanix federal sales team comes from the VMware army sales team. the problem here that I see is that they have no diversification in customer base. what happens when the U.S. army has to cut their It budget by 20-30% as is happening now? that scares many but should especially worry nutanix.
Good lord. Can you just come out and identify yourself as a VMware employee or a member of their PR or social media agencies. I am not too happy with the way your company is dealing with this entire situation. I am sure Nutanix is happy to stoke the fire to get the awareness. This is just embarrassing.
I've always found that Federal is the #1 target for Hyperconverged. If there is anyone with legendarily slow deployment issues its Gov sector. An easy button for infrastructure will win anytime vs. the 5 sub deep labor at $500 an hour, paid $25 an hour in reality. This isn't just Nutanix, everyone in this space should be selling in the federal space, because its an arbitrage of slow confusing procurement and bloated labor to a SingleSKU and a day to deploy.
"What Nutanix does can be done many companies, but it takes many companies to do all that Nutanix does."
So we have the crux of the article; a paid advertisement for Nutanix. You're effectively saying, "Don't lock in with VMware, they'll hold all the cards" whilst saying "Lock in with Nutanix, which doesn't lock you in with VMware".
Genius!
Uh...what? Are you on mind-altering substances? Half the article is about how Nutanix isn't needed either.
The article isn't "don't lock in with VMware" at all. I do say "don't lock in with a company you don't trust not to screw you." I also point out that where decent competition exists, lock-in isn't impotant...you will have other vendors you can call upon.
VMware is trying to build a complete stack that controls all aspects of your datacentre. Storage, compute, networking, automation, hybrid linkage, data protection and more.
Don't buy that massive vertical integration unless you trust VMware completely. There are no full stack alternatives. What's more, once locked in to VMware's software solution for more than just basic hypervisor stuff, it's really, really hard to move away.
It's easy to move away from Nutanix. They're disposible and easily replaced. So you don't have to worry about "lock-in" from them.
If you cannot understand the difference between that and "shilling for Nutanix" you have issues.
The ability to afford the software in question? VMware's great, if you aren't spending your own money. They get a whole lot less appealing when it's your budget that has to cough up the goods. And they don't really negotiate unless you're really, really large.
What's more, they seem hellbent on taking over every single aspect of the datacenter and then pushing all partners and competitors out. So, do we really want a company that likes pricing things only for the Fortune 2000 in charge of every single aspect of our datacenter?
Competition is good. Monopoly is not. VMware makes great stuff, no question...but not everyone needs it all. Who cares if your datacenter can "fly to the moon" when all you need is "commute to work"? But when the options are packaged such that they would make US cable companies blush, well...
...you start dreaming of competitors...or at least "a-la carte".
Here is where I agree fully. If VMware really wanted to take action they could drop the half million SKU's and sell a product called vSphere that when licensed at the enterprise level included Compute/Storage/Networking/SelfService all in one integrated product. e.g. vCloud Suite/NSX/VSAN but instead they wish to follow the path of everyone who has fallen before them (Oracle/IBM/Microsoft etc.) claiming it wont happen to them yet it will it always does, remember Oracle came about from this. No charge separately evolve products separately and call your customer terrorists so that you don't have to negotiate with them.
(Remember: "We don't negotiate with terrorists")
But most of all this is entirely the fault of open source! for too long they keep creating alternatives but pushing companies away saying that the features they ask for are not important or that they need to develop them themselves never mind how hard they make this. Then there is the fact that instead of working together they argue then fork things and while they tell you this is good it is bad for the customer. Imagine if the KVM team put priority on building a management layer and features to match vSphere suddenly there would be an alternative then mix in the XEN team/code to help and build a single hyper-visor.
This is where Nutanix has them scared as if they can put a management layer on top of KVM/NUTX/Midonet then they have a web based fully managed Infrastructure play that does not need VMware while farming large amounts of R&D out to opensource.
They tried fining us for being out of compliance with our license (too many VM hosts according to them) but they couldn't tell us what exactly our license was. We have one VMware stack left and when that system gets upgraded it bye-bye VMware. So I'm all for what's happening and seeing new competitors show up.
There's an 800 pound gorilla in this debate that barely gets a mention, and that is EMC. The largest shareholder and one could say a very strict father to VMware.
Case in point. No dedupe is available on VSAN. Is that because dedupe is a tough technical nut to crack? No, dedupe is solved and commoditized, and while no one from EMC or VMware will admit it, it doesn't take a huge leap to come to the conclusion that VSAN was hobbled to prevent EMC's SAN market share erosion.
Everytime VMware has a new feature or function, they have to run it up to EMC to make sure they minimize overlaps and competition with a legacy storage vendor.
I believe this will ultimately give these startups the air cover they need to leap over VMware and come up with game changing innovative technology or at the very least an eco-system better or equal to VMware's that is feature/function competitive.
In the end, it's the politics created from being such a large company that forces it to fall behind.
While I am not actively calling for anyones demise, All technology companies seem to have their dawn, and their SUNset (pun intended). VMware is no different.
Hardware is dirt cheap. Software is not. It's easy to move between hardware platforms. Not so easy to move between software platforms. Hardware vendors have so much competition that there is no effective lock in except for mainframes.
Software, well...how much power do you trust your software vendor with? So much that you're willing to bet your whole business on them without retaining any realistic bargaining position?
If you trust them that much, go hard. I personally don't trust vendors. I thusly require that my vendkrs be relatively easily replaced, should they try shenanigans.
> Software, well...how much power do you trust your software vendor with? So much that you're willing to bet your whole business on them without retaining any realistic bargaining position?
<SFX: Twilight zone theme>
Something I've been mentioning on and off for a couple of decades, as in (to the FD) "do you really want to write a load of blank cheques to Microsoft". Something I was mentioning to a colleague only this morning - how "cloud" is (in some ways) a full circle back to the 1960s where a powerful ${something} supports loads of "dumb" terminals - and how so many seem to be rushing to cloud without considering the risks (lock in, cloud evaporating, your account evaporating from cloud, ...). And how some vendors (Adobe for example) have moved to cloud/subscriptions so they can extort money for as long as you want to run your creative business - or someone else comes along and usurps them (don't see much sign of that, yet).
Thought provoking post Trevor, good stuff. I've watched the various VMware/Nutanix battles running on Twitter while thinking that as a customer I don't care - all competition is good (as an earlier comment stated).
Your article hits on a very good point though - as hyperconverged offerings gain increasing traction in datacentres (and let's not forget that they still only account for approx 5% of datacenter sales from the last figures I saw) their very nature means the less vendors the better. If you end up running VMware/VSAN, some Pernix/Proximal, some Nutanix/Simplivity/Scale, and maybe even some old school vBlocks/Flexpods, you've defeated the purpose of simplified operations - you'll simply have created new 'hyperconverged silos' to replace the original compute/storage/network silos. So which vendor you trust is a key concern.
Maybe that's how it'll be - what's old is new again, there's nothing new under the sun...
Not sure I can agree. First, you're lumping a lot of stuff together that doesn't belong together. vBlock/Flexpod is converged, not hyperconverged. It's a bunch of legacy crap packaged under one SKU. Pricy, pricy...and still a pain to manage.
Proximal/Pernix is server side caching. Nothing at all to do with hyperconvergence. As for the rest, well...I don't see how competition is bad.
Nutanix/SimpliVity/Scale/Maxta/what-have-you simply take away the need to worry about keeping the lights on. The infrastructure comes pre-canned and you just don't faff about.
Does it matter if one cluster is SimpliVity and the other Nutnaix? No! The interfaces are so simple - and integrated into the vendor management tools - that you don't ultimately have to care. You play one vendor against another, drive down costs, and so long as each cluster is the same vendor, you're good.
And you know what? Each hyperconverged vendor has their own specialty, so you might want different vendors, as each cluster would be better at some kinds of workloads than others. Just like with traditional infrastructure, but less faffing about wasting precious administrator time resizing LUNs.
Hyperconvergence was 5% early last year. It's predicted to be 10% early next year. ANd 30% the year after that. Both Gartner and EMC predict hyperconverged infrastructure will make up 50% of the non-object storage in a datacenter before the decade's out. Hyperconvergence isn't a fad, and it isn't going away. It's the new normal.
So that whole thing where you got paid to make sure your storage was set up right, that it worked with your servers, that you resized LUNs and so forth? Gone. Get a new job. Storage administrators are done, except for the very few who do object storage.
And that's a good thing! Storage administrators are smart folks, and it's better to retrain them for something else and get their brains working on advancing business needs instead of keeping the lights on. That's called progress.
So don't fear hyperconverged setups. Don't even fear multiple hyperconverged vendors in one datacenter. The whole point is "easy button" simplicity...and for the most part, they deliver.
I've looked at the hyper converged platform forms and truthfully I'm not a believer. 2 reasons : You're never going to fill CPU, memory and disk at the prescribed rate that your vendor (eg. Nutanix). Are going to give you. I might be leaving 1 or 2 components on the table and yet go back to the purchasing well to address your needs. I'm much happier with a solution like VCE or FlexPod out of disk add a tray out of memory buy so.e servers but I don't need to buy 3 components when I only need one. Also if hyper converged is so replace able I don't want to divorce my storage and compute vendor at the same time.
VMware is certainly the 800lbs gorilla. They have crossed several lines with many customers. I think an implosion probably will occur because of their ignorance. I see many parallels between VMware and BlackBerry ironically. Hopefully VMware wakes up and starts being a "partner" with their customers before it's too late.
"I've looked at the hyper converged platform forms and truthfully I'm not a believer. 2 reasons : You're never going to fill CPU, memory and disk at the prescribed rate that your vendor (eg. Nutanix). Are going to give you. I might be leaving 1 or 2 components on the table and yet go back to the purchasing well to address your needs"
You are assuming all hyperconvergence solutions are based on predefined hardware blocks. Maxta, for example, is not. You can build your hardware as you see fit, scaling compute performance and storage capacity independently. There's no waste there. A solution that can be designed to meet the needs of your business and your budget, with all the data management simplicity that hyperconvergence offers.
Disclaimer: Maxta partner and believer in storage innovation.
Maxta is currently focusing on pushing it's reference architecture appliances through it's channel partners and downplaying the ability to install it's software on any hardware. This is because unless you are a service provider (who is used to the integration/testing work necessary to DIY) it's relatively easy to unbalance your nodes and end up with a cluster that doesn't perform adequately. Providing support to the uninitiated is costly, and - to be perfectly blunt - the overwhelming majority of users simply don't need anything but an appliance.
Disclaimer: Maxta marketing contractor.
VMW legal should put in an official request for US Government TAA Compliance Document to Nutantix.
.........The issue has gained visibility over the last year after office supplier firm Safina Office Products Inc. blew the whistle on competitors OfficeMax Inc., Office Depot Inc. and Staples Inc. for delivering non-TAA compliant items. They collectively paid settlements of more than $30 million to the government. As a result, the General Services Administration's Office of the Inspector General stepped up its efforts to find violations of the TAA across all industries
You say
"VMware wants control of our data centers. The whole stack, top to bottom. It wants to be the only vendor in your data center making any margin, and everyone else can be commoditized into interchangeable (and irrelevant) suppliers of steel and silicon.
Nutanix wants to make VMware (and all other hypervisor vendors) irrelevant and commoditized, and wants to be the one supplying the software that makes all the margins."
I wonder how that will be possible, for both of them, without owning an operating system? Without owning application services, like containers or Infrastructure-as-a-Service? Without owning Middleware? Without a PaaS Offering?
The true king in the data center is not the commoditized hypervisor, not the kool-aid hyper-converged storage solution - it's the application driving all the requirements. And to the application your fancy infrastructure silos become less and less important. A hypervisor is a commodity - everyone does it - 2/3 of the customers buy vSphere Standard, every KVM solution has more features than that today and it runs on Linux. Storage is becoming a commodity - it's a piece of software that runs on top of an standard x86 OS on top x86 hardware - no propietary box from a vendor who tries to lock you into hardware. Networking is becoming commodity - all I need are dumb L2 switches with fat backplanes - control plane, routing and firewalling is all done in software on top of standard x86 OS.
Your idea about quickly changing SDS vendors is a dream - in reality switching storage from propietary software vendor A to propietary appliance vendor B will involve a data migration. And data migrations are costly - they are disruptive, potentially and their always be trade-offs.
Your idea of just using someone else's KVM instead of VMware is a dream - in reality OS (and thus applications) are certified on Hypervisors. Nobody certifies on KVM as is - because the paravirtualized interfaces to the hardware matter to the software on top and differ from each implementation. KVM itself is not a full hypervisor solution - it's a kernel module that exposes Intel VT-x and AMD-v in a common API - there is qemu, libvirt, virtio and much more - you need to completely control and certify this stack. Try to find out if running Windows or RHEL/SLES/Ubuntu on top of Nutanix KVM is actually supported by the Microsoft or Red Hat / SuSE / Canonical of this world - you'd be surprised.
This shows that the real control lies with those vendors who control the OS. With everything running on top of standard x86 OS these vendors have the potential of delivering the next wave of innovation.
VMware is a mainly hypervisor with nice management and cloud desire - their potential on delivering the next wave of features is limited to what they can do with ESXi. Nutanix is a storage hypervisor - their target surface is even smaller. At some point VC cash will be burnt completely and when the marketing dust settles we will see what true advantages they are able to deliver despite a 6% increase of efficiency in what by tomorrows standards will be recalled as legacy environment.
I've migrated from VMware to KVM. It wasn't that big a deal. If you have trouble migrating between hypervisors or data vendors, the startups will gladly help you. For cheap, too.
But you'll notice the article talks about migrating from one KVM hyperconverged solution to another, and from one VMware hyperconverged solution to another, and so forth.
If you want to pay more and get progressively less, go ahead. If you want to have no bargaining position with a vendor who controls more and more of your datacenter, go ahead. You will not be alone in your view of the world, and you are not alone in your view of the world.
But the number of individuals and companies who retain that view are diminishing. Maybe that means nothing to you today. Maybe it won't mean anything to you tomorrow. At some point, however, you might eventually notice that a rather large shift has occurred while you were refusing to look. Perhaps then you'll stop to consider what all those people know that you don't.
Cheers and beers.
I think you got the wrong idea of my intention. I am not advocating to preserve the status-quo.
Quite the contrary. I believe that moving from one propietary lock-in (VMware) to another (Nutanix) is not going to help. Maybe for you a transition is easy because your set-up is easy. But in general it's a non-trivial disruptive and potentially length undertaking. Because all propietary solutions will eventually lock you in. Maybe not today, because Nutanix, Simplivity and all the other niche players still have to fight for market share - but as that increases and the market consolidates it will become more and more difficult.
I would be interested what KVM setup you are running.
I believe in two things:
1) Only vendors with an OS and application-level service offering (IaaS, PaaS) will be able to truly make a difference and truly "own" a data center - e.g. getting the significant margin by delivering disruptive innovation (not the 6% increase in efficiency) on top of the OS.
2) Only open systems, open architecture will avoid lock-in in the long term. Since everything is moving to software anyway, that equals Open Source.
You are not explaining how Nutanix is lock in. Nutanix sells a variety of appliances that use different hypervisors. So long as you are not using a custom hypervisor designed by Nutanix, how is it lock-in? The whole bloody point is that you can take your workloads and move them to another appliance vendor using the same hypervisor at the snap of a finger.
Your belief that "only vendors with an OS and application-level service offering (IaaS, PaaS) will be able to truly make a difference and truly "own" a data center" is wrong. Full stop. There are lots of parts of a datacenter that matter. If a datacenter becomes heavily invested in automatino and orchestration, for example, any move away from that automation/orchestration platform becomes practicably impossible. And that's just one example.
"Only open systems, open architecture will avoid lock-in in the long term" is also completely and utterly incorrect. You can be just as locked in to open source as you can to anything else. The "openness" of the platform doesn't prevent lock-in. Standards - the establishment and adherence to - prevent lock in.
Consider for a moment the following two scenarios:
1) There is one and only one hypervisor. It is open source. If you want a hypervisor, you must use this, or one of it's not-very-far-afield derivatives. If you want change you need to bribe developers and play petty ego politics and hand-hold groups of grown children as they squabble about minutia. Only the largest companies have the money to effect change. Features that benefit small companies don't get built.
2) There is a defined standard for both the virtual machine container (VHD/VMDK/XML) and the "integration tools", and there are multiple hypervisor vendors available. VMs can be moved between hypervisors with ease. If you don't like the features, development path or what-have-you of a given vendor then you simply choose another.
The latter is an ecosystem without lockin. The former is an open source ecosystem with lots of lock-in. Competition and standards are what prevent lock in. And it's compeition that ultimately drives standards adoption. We're starting to see this now. Startups are emerging to migrate seamlessly between hypervisors. In some cases, even to enable vMotion between them. (See: Ravello, as one extreme example).
Lots of hyperconvergence players is hugely good because they make it easy to transition between them. Don't like one? Throw them away and get another. Nothing - at the moment - locks you in to Nuanix in any way. Lots locks you in to VMware.
Nutanix is lock-in simply because it's a proprietary stove-piped solution. Your data sits in a silo in a data format inaccessible by tools and protocols other than what the vendor allows you. They certainly won't help you get it out when - as you say - one easily migrates them to just another "hyper-converged" vendor. And they certainly won't open the data path wide enough to make it easy for a competitor to enable smooth migration. Your point has yet to be proven - I have not seen whitepapers and success stories around and certainly haven't heard such a story from anyone sharing it at a conference. It's still way to early to tell.
Despite this, you are eventually locked-in because the vendor dictates the use-case and the path to grow and expand the solution. Want to use Nutanix storage for something else than virtualization? You want to re-purpose all that storage you bought for something like an enterprise dropbox? You want to use it to store massive log data? You want to use for analytics using HDFS? You want to use it to run containers? Or built a storage-as-a-service platform? Good luck with that.
What you describe as "no lock-in" is only true if you look at just a single use case - virtualization with merely two hypervisors - Hyper-V and ESXi. As far as I can see Nutanix does not have a credible KVM story. They offer no certification for OS on KVM. No one writes drives for their KVM platform. They don't contribute to the Linux kernel either. It may be interesting for a test bed for a home-brew solution but not for a production environment.
Standards don't prevent lock-in either. There are so many examples where this is visible, it's even hard where to start. But take OVF as an example. A true standard to make moving VM between hypervisors and deployments at ease. Can you export your VM from ESXi in OVF and seamlessly import into Hyper-V or KVM-based stacks? No. There are vendor extensions that prevent this. Happy editing that XML stuff. Good luck with relying on third-party migration tools that cover only 60% of all possible configuration permutations.
Also, Open-Source by definition does not prevent lock-in - true. The more you adopt it, the more you change the ecosystem the more your processes, tools and skills are gravitating towards this model. That's also sort of a lock-in, right. But at least it's an ecosystem that fosters open architectures and by definition everything to build an interface which does not exist yet is available to everyone.
Your example to seamlessly migrate VMs between hypervisor is an example of the degree of obsolescence of the architecture you are trying to improve.
This challenge has been addressed otherwise already. New workloads don't rely on vMotion - certainly not between different hypervisors - they rely on scale-out. You can see this with containers and OpenStack. Both approaches don't offer this feature (at least not in the beginning and certainly not as a major feature) but both see a tremendous uptake in adoption and maturity. Your proprietary appliance however does not cope well with this because it's geared towards a legacy architecture. This won't go away tomorrow - but the industry is moving away from it at an accelerated pace.
Here is a real example of seamless migration that makes business sense and saves actual money: An open-source storage solution can be migrated even off-premise, into something Amazon EBS based. Because it runs on top of an x86 standard OS that is there since forever. It relies on the interface that the OS provides - which is much more ubiquitous than your storage hypervisor tied to ESXi.
It can also be scaled-out to provider platforms, enabling hybrid deployments or quickly providing burst capacity. Try doing this with your hyper-converged. The convergence ends at the datacenter row.
Your understanding of open-source innovation is also incomplete. If you need a feature that does not exist yet you are free to contribute. Good luck begging to your proprietary vendor.
If your contribution however does not get accepted either your implementation does not live up to the quality levels of the project or you are trying to implement something for which other solutions have been found already and you are just not seeing it yet. There is an aspect of meritocracy there that ensures high-quality implementation and common-sense.
It certainly has nothing to do with money (at least not primarily, however money is also what pays developers) and there are many examples where small, very small companies have significantly driven open source projects - in fact it's the norm.
I guess what I am saying is: you are basically right: Nutanix, Simplivity and Maxta are certainly doing their job - in their space (on-premise standard x86 virtualization using ESXi with manual administration and a static IT environment with no self-service capability). This space however looses relevance. And your believe in being more independent and flexible by relying on this is only true as long as you don't leave that space.
In reality however you keep spending money on this - and Nutanix and the like are certainly not cheap - therefore cementing your legacy architecture lock-in even more whereas you should spend money in resources and skills to adopt true open systems and architectures. It's like evolve or die for a dinosaur.
Paypal recently completed their migration to OpenStack - they are completely VMware-free. They completely rely on software-defined storage and software-defined networking. No need for costly Enterprise Plus licenses and crude storage appliance solutions or SDN appliances. They have freed enough resources doing this so they are now able to look at the next level - workload mobility. Containers enable that, not x-hypervisor vMotion. They enable even more - true agile IT operations and fast-paced feature development for much shorter turn-around.
All this was possible because of innovations based on top of an OS. I don't think it would have been enabled by tools and products from vendors that don't have an OS.
Your entire insane rant falls apart right at the very beginning:
"Nutanix is lock-in simply because it's a proprietary stove-piped solution. Your data sits in a silo in a data format inaccessible by tools and protocols other than what the vendor allows you"
Nutanix uses the same formats the hypervisors use. There is nothing proprietary about the "data formats" at all. There is nothing preventing you from very easily moving from Nutanix off to someone else.
So basically, you're just making shit up. The question is...why?
And again you are not understanding the angle of the question.
Of course on the front end Nutanix' Controller VM supports various the data protocols used by several hypervisors. And from that perspective you can migrate data as "easily" as with every other storage array that is used by Hypervisor X to Hypervisor Y or to another "hyperconverged" solution. You are doing this with the interfaces on top of the hypervisor. Not much gain there. It's as easy or complex as with a storage array that speaks a supported protocol.
The question is: how is data stored in the backend above it's internal SCSI controller? Meaning: is the BACK-END data format that sits on top of the HDD/SSDs in the node making up NDFS known? Is it directly accessible?
Unless I am misinformed it's not because it's propietary. And that's where Nutanix controls access to your data. You need HDFS instead of NFS and SMB 3.0? Out of luck. That's where the lock in occurs.
A real software-defined storage solution sit's on top of a well-known format here, in the Linux space this is usually either LVM or XFS. There you can still access the data even when the SDS layer is offline or unavailable.
So apparently you are always coming back to the use case of using hypervisors to access data which is a major but not the only use case for SDS.
Instead of accusing me of "making up shit" you should probably try to understand what other uses cases exist than just yours. As it turns out the SDS market, where Nutanix claims to be a leader in, only attributes a small portion of it's growth and revenue to online workloads like VMs.
The major money is being made in environments where the data is NOT accessed through a hypervisor.
And that's where YOUR story is completely falling apart and you end up with another silo for a single use case.
My 2 cent.
Mmmm Think about it VMWare is forcing all hyper-converged (and other manufacturers) to develop and work with their own hypervisor(s) this is already starting to happen. Vendors may long term have to collaborate and create one jointly to share costs, initially they will lose business but finally VMWare will end up with competition they never had in the first place and losing market share they currently dominate because vendors will be forced to not work with them
The trouble is, every time I look at hyper-converged systems I come to the conclusion that there are only really a couple of use cases where it fits; VDI and simple web farm setups. For any moderately sized enterprise there are just too many edge cases to make it worthwhile. I don't disagree that for systems like VDI the benefits can be enormous where consistent scaling and performance makes this a natural fit. If you live in a world where all your servers are virtual and don't need consistent high performance then again it's great. For the rest of us though, we have things like heavy database workloads where the performance and consistency trade-offs of hyper-converged solutions are simply unacceptable. We have servers which for various reasons must be physical, and a million and one other things which mean that the hyper-converged system will need to sit next to the other standardised servers. Anyone who has run even a small enterprise data centre will tell you that standardisation, not diversification, is the way to a sane life and stabe systems. If I choose to lock myself in to HP, I know I can update all of my firmware through a single solution rather than visit each silo individually. Locking myself in to Microsoft means that all my servers are patched in the same way, and I only need to test patches once. Locking myself in to NetApp means I can easily replicate my data and backups in a single operation. Those are choices to be made individually, and I'd certainly agree the storage replication is a good area where software is a good addition and things like Veeam can mean not needing same vendor storage on every site.
The real costs, the real complexity is not in the hardware or the software, it's in the human element. Can your IT team learn and professionally operate all these different solutions to an acceptable level, or will they be masters of none? The answer to this and most other data centre questions is generally how susceptible to marketing BS you are.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
