Nutanix 'let chaos reign', groans CEO as shares tumble more than 20% amid dismal forecast Nutanix is staring down the barrel of virtually no growth for the next quarter, an admission that sent its share price into a tailspin as analysts grilled the HCI vendor over inadequate marketing spend and sales hires. The company last night reported financial for Q2 ended 31 January of fiscal '19 with revenues up 17 per cent …
It simply couldn't be a result of a very crowded HCI field, many well-funded competitors, market saturation, and cloud of all flavors, now could it? Nutanix started great, got some success and turned arrogant. " The Future of Enterprise Cloud" indeed. Pride before the fall and all that. Plus, don't forget, with the Nutanix proprietary solution, there is no lock in, ever. Oh, until you buy it that is.
They were arrogant from the get go but it unbearable lately. Their marketing keeps writing checks their products cannot cash.
Sure, the smooth upgrade and management were cool at first but everyone does that now and Nutanix’ first mover product is showing cracks. Sucks to be the first one to try things.
But now that the competitors are equal or even better in many aspects, Nutanix was pushing into more and more spaces. Software Defined Networking! Application Blueprint! Application Performance Management!
Honestly, all those products look like shareware. It’s laughable to demand almost as much as New Relic for APM and Morpheus for CMP but offering a puny subset of the functionality.
And now you have to start paying license per core and per storage capacity. WTF? Forcing customers into this new SW license model including audits and enforcement. They went from expensive to Oracle-esque but keep harping about the “vTax”.
And still, only 13% growth in a 70% growth market and $120m losses with another $100m coming (it will be more, you’ll see).
The convertible note for $500m (give us a loan, stock is collateral) they issued last May or so is due soon and almost depleted. They need more cash to keep the lights on and they need it quick. Problem is, their stock is worth shit now and might go lower.
They’ve entered the death spiral: big losses and no cash, more pressure, more discounts, more losses, smart talent sees the signs and bails, all leads to worse quarter, cuts, more pressure, more losses, ...
Six to nine more months until they get bought by IBM.
Very interesting is that Nutanix leadership vehemently denied that this is because of increased competition. It will be very interesting to see IDC numbers for VxRail, SimpliVity, and the dark horse HyperFlex.
Strong showing by either HPE and/or Cisco would show that their products are now mature enough to competently defend their high margin installed base. I think that all those VSAN and Nutanix installs are overwhelmingly in small and medium business with cut throat margins. Nutanix sold some large deals into mostly government and education which kept the show going but large sweeping enterprise wide deals are few and far in between. The real prize is Enterprise and with a lacking ecosystem and integration into existing environments your solution is a non-starter. VxRail doesn't play there, it's more for niche silos like VDI, as EMC/Dell will immediately upsell Unity and Powermax or go VxRack.
I am hearing HyperFlex creeping into that space. Interestingly enough, they do it in cooperation with NetApp (AFF) and Pure. Cisco UCS division pressuring ecosystem partners to pair up their storage offerings with HX?
If this holds true, and it's a big IF, then we are witnessing a sort of repeat of the server market around 2010: Dell and HP battling it out in SMB and small commercial, HPE and Cisco battling it out in high margin Enterprise.
Stay tuned, boys and girls. Next few quarters might be writing history!
A lot of folks bailed after shenanigans with territories, commissions, and goals. Complete shitshow and people talk, esp sales people. Once word gets out that you can’t make money, why would you go work there? All this was driven by a grow grow grow mentality. Trying to somehow lock in enough customers so you can extract monopolistic rent at some point.
So they weren’t able to hire talented people in sufficient numbers and now we see the results.
...analysts grilled the HCI vendor over inadequate marketing spend and sales hires.
Just because analysts and sales droids want to sell things doesn't mean that people want to buy things.
Too much voodoo, too little use of hardware offload and too narrow a hardware compatibility list to ever make it onto any my proposals.
Oh and be open about that 8 node limit that we all know you have but you won't talk about. Let people do performance testing of your platform rather than restricting their ability to do so -- unless of course you have something to hide.
Google for it.. plenty of articles out there..
Then wonder why the sales people never specify clusters larger than 8 in any designs.
Then look at the minimum configuration of replication factor of RF3, but recommendation of RF5, so new clusters need 5 nodes minimum for acceptable resiliency.
So, the sweet spot is 5-8 nodes a cluster, which is quite a cost step when you need to increase capacity. Anything less than that and you accept failure modes that will adversely affect resiliency. Anything more than 8 and you need to roll a whole new cluster again.
John from Nutanix here
@baspax
Our larger clusters are not a patchwork of smaller clusters. We scale from 3 to the number of nodes that you need in a single cluster. Nodes can be retired/repurposed as and when it suits. (There is another HCI vendor who 'federates' to get to higher node counts, we don't need to do this to scale.)
Our Prism Central management tool handles 1 or multiple clusters.
People tend to have more than 1 cluster because they have two sites and/or want DR.
At a single site people may want to use ESX for some workloads and AHV for others (for example Citrix VDI). Prism Central makes managing clusters with different hypervisors very easy. As a company we believe choice/openness is a good thing.
At a certain scale it can make sense to have more than one cluster at a specific site. For some customers that might be in the 16-20 node range, for others it might be at 50 nodes. I've had a customer who for compliance reasons had two 3 node clusters per site - without the compliance needs they could have deployed a 4 or 5 node cluster per site for all their workloads. Like so much in IT the answer is 'it depends'.
John from Nutanix here.
There is no 8 node limit or 5-8 node sweet spot. We have a lot of clusters larger than 8 nodes. We can mix different workloads in the same cluster, for example VDI and databases. We can also mix hybrid and all flash nodes, and different CPU generations in the same cluster if that suits. When it makes sense to have two clusters rather than one depends on a range of factors, many not to do with Nutanix. The right number of nodes depends on the particular customers requirement. People really can start a cluster at 3 nodes and grow it a node at a time to the size they need. We have options for 1 and 2 node clusters if that better fits. We are genuinely a very flexible platform capable of handling the vast majority of organisations computer infrastructure needs, now and evolving non-disruptive as the requirements evolve. Anyone doubting Nutanix can scale beyond 8 nodes should speak to some Nutanix customers who have done so.
We are introducing a set of additional capabilities beyond our core HCI platform, for multi-cloud environments and for workloads such as file sharing and object storage. IMO It's a very interesting set of ambitions building on a really solid core HCI product. (I have declared my own interest.)
[caveat: I work for a company that sells multiple server, storage and HCI platforms]
I have several customers with nutanix clusters > 8 nodes. They are specific use cases and we aren't typically recommending this unless the use case makes sense.
For high performance workloads, we are recommending different clusters. We are mixing workloads in some cases without issues.
We have some customers that love it. Other customers who have purchased it have already yanked it and moved back to a traditional architecture or are getting more out of VSAN at a lower cost.
Your core HCI is anything but solid. Dheeraj went to market with an obsolete data platform that is unsuited for the workloads it is being positioned. We all know this and see it falling over left and right. Mohit bailed and reimplemented this GFS spin off as Cohesity and that's what it does well, tier 2.
For f*ck's sake, Andrew Fikes literally said on stage six months after your guys spun off to commercialize GFS that GFS was crap and unsuited for non-batch jobs, hence the development of Colossus and Spanner.
Dude, object storage? Why would anyone do that? I can spin up a S3 bucket in a minute at 1% of your pricing.
@John from Nutanix Here.
So, no sweet spot, but a reference to the nutanixbible site (which is an interesting read). The 5 node minimum is real - to get acceptable levels of resilience of the data (inc. metadata) in the cluster, even says so on the above web site.
Then you go and say "We have options for 1 and 2 node clusters if that better fits.". A 1 node system isn't a cluster and it offers no resilience at all, so not enterprise ready,
Secondly 2 nodes is below the RF3 you talk about in the nutanixbible web site.
Then you go on to imply that cluster size is a function of workload. Well, if it scales properly, then you should be able to scale out linearly, unless there is some bottleneck that prevents this - otherwise why wouldn't customers do this in all cases ?.
I challenge your "no 8 node limit" statement. I've had several designs come across my desk whilst at different customer sites and they never go above 8 nodes. When I ask what happens next when we expand - its always a new cluster, now why would that be ?
Could it be that the age old mantra that the interconnect traffic increases as the node count goes up, so the bang for your buck decreases as you scale out making it less and less viable. Is this also the reason why you have to have storage heavy nodes to offset the disk performance issues. How many of those do we need for performance AND resilience for RF5 ???
You also say about broad hardware compatibility, but again, first hand experience, incompatible hardware derailed one site build as some standard built-in mainstream vendor card wasn't compatible with the software and it refused to install with a cryptic error message.
So, my position on hyper converged hardware - plenty of hype all round.
@Dwarf
John from Nutanix here
I referenced the nutanixbible.com site because it details for anyone interested how we work. It is a judgement call as to if RF3 is needed. There is some bad maths out on the internet on that topic.
I am sorry someone gave you bad advice about an 8 node limit, it really doesn’t exist. We have a lot reference customer running >8 node clusters. Until today I had never encountered this mistake. We don’t have a problem with interconnect bandwidth as we scale. I appreciate some architectures do. Mostly we do it all on two 10GbE per node, but we can have more ports and/or 25/40GbE for really bandwidth heavy environments. That is to do with the application bandwidth needs, not scaling the cluster.
Storage heavy nodes are to allow clusters that just need more storage capacity to scale in that dimention (when more CPU and RAM are not needed). It’s quiet a nice answer to one criticism of HCI solutions - that you have to add more CPU/RAM if you only need more storage, and these nodes don’t need an ESX license, a nice saving for ESX customers.
Again I am sorry you were trying to install our software on something not on the HCL. We provide free training to our partners on these topics. The HCL is there to avoid such mishaps. I was double checking an HPE build today to be sure it was all on the list. I suggest get whoever had the problem to contact their Nutanix Channel SE.
1 node (or even 2) is not, by some definitions, a cluster. That is true. It is not true to say they have no resilience. They are local redundancies and can replicate to another cluster. They do have their use cases, though they lack the elegant flexibility of 3+ node Nutanix clusters.
@John from Nutanix here.
Bad maths for RF3 - hmm, sounds like sales BS to me, the concept is very simple, 3 copies of each block as opposed to RF5 with 5 copies of each block, loose a node with RF3 and you only have 2 copies of the critical data, doing maintenance on another node at the same time - now only a single copy of the precious customer data. Further the response on storage heavy missed the point - if your resilience is to not use RAID and to do data resilience manually across the nodes in a cluster Then nodes that have far larger percentages of storage represent far higher risk to data integrity when they fail, this is just basic maths.
The same applies to small node count clusters where unbelievably you claim that ‘there are local redundancies’ perhaps you can explain this claim given that there is no hardware offload (which often provides resilience) and it’s a very well known design principle that single systems are not resilient - for any of a large number or failure modes, this is where the concept of n+1 comes from, even for one node, that results in 2, again basic maths.
I was touched by you thought that I would be installing systems - nope, one of your SI’s was doing that - but thanks for the blame the customer tactic.
I also note that you completely ignored a number of questions / observations over previous posts - perhaps those are the hard questions, but then on reflection and re-reading the whole thread, I note that the main thrust of your posts is just to claim that the views are wrong / uninformed whilst not provoding any referenceable information sources or answers. For those of us who have worked with the product, reflect that that might be why the sales are down. Sales claims only work for a while, but when the workload hits the metal, customers get to see the warts in the designs that makes them consider their options when outages and performance issues hit the planned scalability.
This post has been deleted by its author
John from Nutanix here
@Dwarf
There is a difference between how Nutanix protects metadata and the actual blocks written. We don't RF5 data. Not trying to duck anything but it is covered much better than I have space for at nutanixbible.com. I'm not going to point to a third party's completely wrong account of how RF2 and RF3 work (the bad maths), not bs, me not spreading bs.
1 node cluster mirror data and meta data locally, 2 node clusters mirror between nodes. you don't get non-disruptive upgrades on a 1 node cluster. It's a reason they aren't by some definitions a true cluster. Nutanix is completely open about this - no smoke and mirrors here.
I got the impression with your 'first hand experience' you had been involved in the install. My suggestion still stands, whoever was doing the install should contact their local Nutanix Channel SE. We train our partners, free of charge, how to do this things correctly. Occasionally someone (not blaming any specific person, certainly not the customer) drops the ball. When this happens we try and make sure it is not repeated. Please do pass on the suggestion, Nutanix will act on the request.
I think the big storage node thing is a bit of a straw man. It doesn't make sense to add a single massive storage node to a cluster made of up of small general purpose nodes. Equally 3 nodes of 80TB is probably a sub-optimal design, five nodes of 40TB would do a better job (and provide more useable capacity because of the smaller n+1). File storage is the most common place big nodes make sense and to get a better price point for unstructured data we now have an unstructured only license (Files now, Object (Buckets) very soon), which makes unstructured only clusters much more cost effective. I accept we are not always the lowest price option and for file workloads we've tried to address this. It is a new thing so we are adjusting aspects as we go in response to customer feedback.
I do not believe I have stated anything is incorrect without providing evidence/an explanation. You say sales are down - that is absolutely and publicly demonstrably not true. Software and support revenue was up 42% compared to the same quarter last year and 47% (to $1.1bn) for a rolling 12 months.
Hardware offload, sounds like Dwarf is connected to SimpliVity. I read Gartner's critical capabilities report on HCI ... SimpliVity isn't going to beat vSAN or VxRail or Nutanix or Cisco. Might manage 5th in the market because HPE does have a big customer base, or 4th if Cisco aren't pushing HyperFlex hard. People should read it for themselves rather than trusting spats like this on news sites. ... Much as I love a good spat.
Hardware offload, sounds like Dwarf is connected to SimpliVity.
Nope, Dwarf is not affiliated or connected to any vendor, I make up my own mind and I'm good at digging into products.
The hardware offload question is a very simple one. One clock tick of a microprocessor does one or less operations depending on the complexity of the operation being performed, Dedicated hardware to do specific tasks is well established - RAID controllers, TCP offload engines for NIC's, encryption cards that handle the complex encryption calculations all increase performance as hardware can always outperform software - because software runs on hardware. Think what other vendors are doing - MS stuffing FPGA's into their cloudy servers, now why would they bother to do that ???
There's also the little point that any CPU cycles doing things that are not directly related to the user experience are effectively wasted, so minimising this wastage by getting a specialist piece of hardware to do it results in better user performance.
The concept goes far further back - all processor architectures have DMA capabilities - taking the CPU out of the equation to move large blocks of data around. offload cards DMA the data into them, doing a CPU based linear read is going to be slower from the outset.
So, when I see a vendor pushing their product and saying its as fast as / faster than offload cards, I smell BS. pulling storage blocks from other nodes across the wire without using dedicated storage connectivity means congestion for other user facing IO, so again performance can't be better than when that traffic is not there, yet were told its fine, nothing to see here. Then there was the lack of public benchmarks - now why would vendors do that unless they had something to hide ?
Which is worse? The flawed archtecture, that is over marketed, or the blaming of sales and marketing for lack of growth? Nutanix is trying to go it alone with almost no eco system partners, in this day and age, customers are not interested in lock in hardware with a lock in cloud to go with it. They will continue to fall until some big company, that lacks any real HCI play, will buy them.
John from Nutanix again.
Nutanix is very public about its eco system partnerships. They can be checked out here: https://www.nutanix.com/partners/technology-alliance-program/
Definitely not trying to go it alone.
Flawed architecture? I joined Nutanix precisely because of its superior architecture - perhaps a debate for another time but we are extremely open about how we have built our product. Details available here: https://nutanixbible.com/
"Locked in hardware" - this is so 3 years ago. We have just announced yet more hardware platforms certified and customers can separate their hardware and software purchases. They include HPE, Cisco, Dell, Intel, Lenovo. As a company we believe in choice, not just for hardware but also for hypervisor.
Our cloud efforts work with Amazon,Azure and GCP as well as our own cloud services. We are building the tools to help customers avoid cloud lock-in. Worth checking out our Beam solution, which can help cost and security optimisation of public cloud use - entirely independent of Nutanix HCL.
This post has been deleted by its author
Hi there "John from Netapp" here .. which features do you think would be important .. ability to scale well beyond 8 nodes without worrying about the impact of bully workloads ? Support for vVOLs ? Support for VMware private cloud ? Support for RehHat OpenShift? Best in class integration with Kubernettes (going well beyond, and driving CSI standards) ? Storage I/O performance you can rely on with real quality of service guarantees (not just limits) ? Deduplication and storage efficiencies with no performance impact that actually save significant space (generally 5:1 - 10:1 before snaps vs Nutanix at about 1.5: 1 - 3:1 without needing yet _more_ DRAM and CPU overheads for the CVM ? Worlds best NFS and SMB File services ? Ability to replicate data to AWS, Azure, GCP, IBM today ? Lower TCO ? ... I agree a features comparison would be a hoot - though I don't think it would be entirely fair to compare a product built for All Flash, Containers and high speed networking and multi-cloud integration to something which was built to optimise the performance of a distributed filesystem in a single datacenter running on spinning rust over 1Gbit networks (must keep all data local because networks are soooo slow)
John from Nutanix here
AC - I am sorry your tender experience left you feeling Nutanix wasn't the right choice for you. It may have been that for your requirements Nutanix was not the right answer. It sounds like the reseller and Nutanix did not do a good enough job for you. Without details it is hard to say more (and I wouldn't in a public forum). Even when we don't win a prospect's business our aim is always to give a good account of why we are proposing what we are, and if there is a price premium justifying it. I hope if you deal with us at some point in the future you demand that from us, it is the least we should be doing.
Sounds like the reseller was boxing out Nutanix because they did not have deal registration or an alliance with another vendor. It happens.. we never lose on price and we are hardware agnostic so hardware is a push/non-factor. If you did an ROI it would have been a no brainer. No VTAX. AHV is free...
Dude, how about you shut up with your Nutanix propaganda. vTax, AHV is free ...
Nobody is buying your bullshit. You guys started licensing PER CORE AND PER TERABYTEforcing people into this licensing schema to extract more money and you have the audacity you call other people's licensing a tax? Get out of here!
Love the fact that in order to read these posts you're peppered by the No VTAX banner ads...and you are assuming I am a guy so your wrong multiple times....
No charge for AHV. Do Amazon customers have any idea which hypervisor they are on? no, because it does not matter.
We don't force, anyone, to do anything. We add ~1000 new customers per quarter: who else can claim that?
Nutanix gives you a choice for a platform that actually does what it says it will do. Extract more money is exactly what all good sales teams do; it is called charging a premium for a product that provides value.
The ROI's do not lie.
Two words: " Number 1" in the Gartner HC quadrant. Oh, and by the way, they made the quadrant for Nutanix; it never existed before.
Nutanix makes infrastructure invisible. The paradigms have shifted and Nutanix will do to its competition what Netflix did to Blockbuster. Subscription based licensing is here today, not the future.
*you're
The amount of Kool Aid bullshit oozing out this one short post is unbearable.
Buddy, adding 1000 customers per quarter to buy a dollar for 50 cents is not hard. Try raising prices and posting a profit because where we stand right now it looks like you are fucking sinking like the Titanic.
And here is some free advice: you see the Nutanix ads because you work for Nutanix and visited some Nutanix stuff. Now the affiliate network is sending you more Nutanix stuff. All I see is TheRegister Security Rapid! workshop.
Your lack of vocabulary is showing, no need for profanity...
Land and expand has a cost as does R+D. If you want to stay in the lead you have to reinvest. We are not breaking new ground here: the best and the biggest have all done the same and have survived.
Done responding; go back to your mediocre career and live in your ignorance..Its getting dark troll better get back under your bridge...
On another note, IT sales is generally pretty cut-throat, but in my experience, Nutanix sales teams take the cake when it comes to unprofessional behaviour. Even in situations when we've led with Nutanix, their sales (and management) teams have bent over backwards to break their own rules of conduct by cancelling DRs (because they don't want to sell Nutanix OEM hardware), refusing to price solutions, tanking RFPs 1 day before submission and attacking clients and competition with FUD after they've already lost deals. Never seen that with any other vendor, especially not any established/major vendor.
DISCLAIMER: I understand this could just be the local vendor team, but company culture can be pretty telling and is often unrelated to actual product quality - I may think the majority of IBM products are below-average and of little value to a client, but I could never fault the professionalism and attitude of their sales guys.
Anon for obvious reasons.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
