Update 8/3/2011: VMware announced updated licensing terms (link is to my post on the matter).
As part of the vSphere 5 & Cloud Infrastructure Suite announcements today VMware announced a new licensing model. And, as expected, people are having a fit. A few of us were briefed on this new model last week, and I’ve got a four-day head start on the denial, anger, bargaining, depression, and acceptance that seems to follow this change. Let me work through it with numbers from my environment, as an IT professional, in a professional way. Hopefully this will let some people pass from the anger stage to bargaining (perhaps with their VMware sales representatives) and on to acceptance.
Before I start, I do want to say that for some people this will be a negative change. For some it will be a positive change. And for a lot of people it’s just a change. Every environment is different, and this change will definitely alter the way we size VMs and hosts. The change from Advanced licensing to Enterprise will also play into it, as will the different licensing levels in general. I urge you all to do the math for yourselves and find the truth of the situation, and if it’s bad tell VMware about it.
Personally, I don’t like that the change penalizes those using the “fewer, bigger machines” model, which is a giant time & money saver. I do like the new licensing because I can now assign direct values for chargeback for VMs, based on size. If you don’t do chargeback this is neither here nor there.
My environment:
- 429 virtual machines, with 1533 GB of RAM allocated to them.
- 19 licensed hosts, in 4 discrete clusters managed with a single vCenter instance:
- Dell PowerEdge R900, 128 GB RAM, 4 sockets – Quantity 6 (24 sockets total, 768 GB)
- Dell PowerEdge R610, 48 GB RAM, 2 sockets – Quantity 3 (6 sockets total, 144 GB)
- Dell PowerEdge R710, 96 GB RAM, 2 sockets – Quantity 4 (8 sockets total, 384 GB)
- Dell PowerEdge R810, 256 GB RAM, 2 sockets – Quantity 6 (12 sockets total, 1536 GB)
- 50 sockets of Enterprise Plus licensing.
- Clusters built on N+1 sizing, which changes the counts to:
- Dell PowerEdge R900, 128 GB RAM, 4 sockets – Quantity 6 (768 GB)
- Dell PowerEdge R610, 48 GB RAM, 2 sockets – Quantity 2 (96 GB)
- Dell PowerEdge R710, 96 GB RAM, 2 sockets – Quantity 3 (288 GB)
- Dell PowerEdge R810, 256 GB RAM, 2 sockets – Quantity 4 (1024 GB)
Things to note, based on the vSphere 5 Pricing & Licensing Guide:
- This is NOT a per-host license, it’s a pool. That will even things out a bit between hosts that are not full and hosts that are.
- This is NOT based on physical RAM. It’s on allocated RAM.
- In an N+1 situation you will not pay for the +1, because there are no VMs allocated on your spare capacity.
So, I have 2832 GB of physical RAM in total.
However, if I fill all of that I’m hosed, because I will have used all the +1 in my N+1 sizing. Not cool. So if I remove the +1 hosts from the counts, focusing just on the “N,” I have 2176 GB of capacity.
With my Enterprise Plus licenses I get 48 GB per socket.
48 GB * 50 Sockets = 2400 GB of licensed capacity in my pool.
2400 GB of licensed capacity minus 1533 GB allocated to VMs = 867 GB surplus in the pool.
When I fill my cluster I’ll still have 224 GB of surplus.
So I’m okay for the vSphere 5 upgrade, and I will have to figure out how I’m going to size my next set of hosts because the math is different for that now. I’ll cover that in another post, because I haven’t run the numbers yet.
Please keep comments professional and informative. Thank you!
Update: I’ve continued thinking about this topic over at “A Look at VMware Licensing & Environment Growth.“
Comments on this entry are closed.
Right-size the VMs…! :)
That will help a) your costs, b) your consolidation ratio, c) make individual consumers better ‘cluster citizens’.
And you’re point about [charge|show]back is spot on. That, along with the new licensing, will actually help people in the efficient consumption of IT resources… which can only be a good thing for all.
Thanks for sharing your numbers, Bob. I think it’s very important for everyone to take the time to run their own numbers and come to an informed decision. Yes, for some people the change may be bad. For some, it might be good. Your post will go a long way toward starting a calm, rational analysis of the change and what it means for each person’s organization.
Good post, Bob. My current numbers are similar (just on a vastly smaller scale).
Unfortunately for us it will negatively affect our soon to be purchased/implemented DR site (was going to go fewer, large RAM boxes).
One on my beefs is that I have to monitor vRAM usage now not out of performance and capacity concerns, but out of budgetary.
I agree with the ‘usage based’ charging model because that’s the model that the cloud is based on. I’m just not convinced this was the way to go.
Thanks for sharing the numbers. It definitely helps demonstrate how it plays out. I did not see anything mentioned about memory overhead. How is that factored in?
I’m in total agreement about running the numbers. I am surprised VMware hasn’t released a handy little calculator for determining the # of licenses needed.
Calculator will be out soon is what I have been told.
I think the new pricing is pretty screwed for most people.
The idea of consolidation is to reduce the number of servers by buying larger servers to handle more load *per* server instead of many smaller servers. At 48 GB per CPU socket – well, I just don’t buy this as a great way to further virtualization projects.
My new cluster is built upon 4 2 socket, 96 GB RAM systems, glad I sized them this way but i was already looking at doubling the RAM because in our final rounds of testing, we still see RAM as being the top consumer percentage wise over CPU utilization in normal operations for our customers.
And now cloud.com being bought by Citrix? For Realz? :(
I’m unfortunately on the poor side of the change – my Enterprise licenses no longer cover my cluster’s physical RAM but I desperately need the improvements in VMFS that are available with vSphere 5. My cluster is small (4 dual-socket hosts, two with 96GB and two with 48GB) and runs at about 60-65% RAM utilization right now. The problem I face is twofold – I was planning on upgrading my 48GB hosts to 96GB and I’m bringing an Exchange server online, in FT mode, as well as a View installation which will push my utilization on the current configuration to the breaking point. (That’s why I need the RAM upgrade.) While I had been toying with an upgrade to Enterprise Plus I certainly don’t need the additional pressure on my budget as an already-cash-starved higher education division.
> 429 virtual machines, with 1533 GB of RAM allocated to them
That’s actually quite impressive. We currently have 84 machines, with 493324 allocated. Stealing and updating your layout…
7 licensed hosts, in 1cluster managed with a single vCenter instance:
Dell PowerEdge 2950, 64 GB RAM, 2 sockets – Quantity 4 (8 sockets total, 256 GB)
Dell PowerEdge R810, 256 GB RAM, 2 sockets – Quantity 3 (6 sockets total, 768 GB)
14 sockets of Enterprise Plus licensing.
So, I have 1024 GB of physical RAM in total.
With my Enterprise Plus licenses I get 48 GB per socket.
48 GB * 12 Sockets = 576 GB of licensed capacity in my pool.
576 GB of licensed capacity minus 494 GB allocated to VMs = 82 GB surplus in the pool.
Now of course, with vSphere4, I had a licenced memory capacity for all 1024GB in our servers.
We also have/had in the works plans for probably another 4 R810s with 256GB so we can give more memory to our SQL Development servers, and also P2V one of our live SQL boxes. That was going to cost us 6 CPU licences – with vSphere 4, that’s now going to cost us 16 licences – 4 times as much.
I can see vmware doing a lot of volume in vSphere4 licences between now an 12th August.
My problem with the new licensing model is that it seems it was created for a different era of servers. The blades we deploy are 144GB with 2 sockets. There is no licensing model that accurately reflects the VRAM of this single server. Now I understand that VRAM is a pool of servers approach, but even purchasing 4 sockets for a N+1 cluster would waste RAM.
We are called to deploy large VM with loads of memory. You can say right-size your VMs, but when you are dealing with custom apps running on Java, 20 GB VMs become common place.
Potentially a usage method would work, but if you want to implement it, don’t base upon CPU – Memory ratio from 1-2 years ago.
Here you go folks. I did some math with the scenario that VMware provided in their vSphere Pricing document and what happens as the ‘customer’ increases their VM count beyond the “ideal” scenario. I also investigated how licensing cost changes to max out boxes and vRAM capacity. It truly comes down to price/vm. Hope you find it helpful.
http://www.vtesseract.com/post/7557543715/vsphere-licensing-doing-some-math
The great thing about the new licensing model is that you don’t have to buy all your licensed capacity up front if you think you’re going to be running a large amount of vRAM eventually. You can license the minimum and then purchase licenses just when required to be in compliance. The value that vSphere 5 brings is still going to far exceed the costs even when considering environments with very large VM’s. I think one impact will be that there is no point having memory overcommitment, as you’re paying for all the vRAM. So you might as well fully reserve 100% of allocated RAM to every VM. What is the point allowing ballooning and swapping when you’ve already paid the licenses for the vRAM entitlement. The answer I got from someone was that RAM might be expensive and then you wouldn’t want to purchase the full capacity of physical RAM to match your entitlement. To be honest this is probably quite a naive statement, considering RAM isn’t that expensive, especially consider the cost of performance problems if you don’t have the physical RAM to back the allocated RAM.
Michael, you make a good point about cost of RAM being low. Looking at hardware configs prices, though, it seems that cost of RAM doesn’t scale linearly with capacity, but it actually becomes much more expensive once you go above certain capacity sizes. So the optimal configuration from a RAM cost stand point maybe actually lower than the maximum that the server could reach, possibly making overcommit still very valuable.
The other key use case cases for which the value of overcommit remains intact are HA and host maintenance mode, i.e. temporary spikes of usage. To the best of my knowledge these use cases were actually considered the most valuable by most customers.
Alberto, absolutely spot on, especially when comparing the 16GB RDIMMS to 8GB, as there is more than a 2x cost in some cases. On the HA front, most clusters I’ve been seeing are designed so that even with maintenance mode the VM’s will not be forced to share/balloon/swap their memory, as most clusters are not 80 – 90% utilised. So even with the HA scenario it’s unlikely in most cases having the 100% reservations would be a problem, especially when you don’t want performance impacts during maintenance operations. Also the more varied the workloads the less likely you will be to go over a vRAM entitlement, especially when the HA failure capacity is taken into account. The new licensing model also encourages greater efficiency in allocation of resources. But all of this still is a small component of the overall cost of a solution when you consider the operational, labour and applications costs.
Is there a tool available from VMware which I can run on my environment and see what the effect will be on vSphere 5.0 licensing?
This is what we desparately need.
Tool is coming – will be ready by GA (this quarter)
Why I need a Tool ??? To understand how much VMWare screw me up.
It´s enought I will move from VMWare to a more reliable company ! ! !
We unfortuantely end up on the short end of the stick with this one. We just bought new production servers (5 server x 256GB RAM). We were looking to seriously ramp up our VM utilization. We currently have 96GB of RAM in each of our 5 hosts. They are all sitting around 60% utilization. We want to start virtualizing a few of our tier 1 services. With the new pricing model, I’m going to burn a “CPU” license with one or two large VMs.
We finally got our boss convinced that VMware was still the way to go (as opposed to HyperV) and that the renewal was worth the cost. Now that this is out, our renewal in 2-3 years is going to be SIGNIFICANTLY more than it was going to be this year. As an education customer, our licensing cost is pretty good. Unfortunately, VMware doesn’t discount the SnS very well, so SnS ends up costing us as much as the license itsefl.
Ridiculous.
32GB per enteprise licensing – ridiculous
Assumption that the average server is purchased with < 100GB – ridiculous
Its pure and simply a cash grab. VMware has realised that we're getting more bang for our back CPU wise, and therefore continue to license via CPU will limit revenue. Switch to licensing via Memory – problem solved.
Memory is the limiting component which is why we've been cramming as much RAM into our box as possible.
Everyone I've spoken to are faced with paying more – which directly relates to the majority who've posted above who come out worse off, despite the OP's stats.
We have a fleet of dual core 128 GB hosts, averging 50-60 GB allocated per CPU, thats ignoring Memory Overcommitt – a much toubted VMWare feature. We'd be up for almost 2x the licensing costs in V5 – simply not gonna happen.
VAR's, resellers are all getting hammered and answering the phone with "Are you calling about the new licensing? Yeah, we're pissed off to, and are informing VMWare that they are about to lose customers – its going to cost us money too since we don't resell Hyper-V!"
As if we needed additional incentive to begin considering HyperV – and I'm a VMWare Junkie from years go. They're not that far behind..
Andrew, I fully agree with you. These calculations and figures just prove my first thought about new vSphere 5 licensing model – it is beneficial and flexible for owners of big farms and Enterprise Plus edition licenses. It is aimed to big players, and I am afraid now smaller virutalization customers can swing a bit to free MS Hyper-V solutions.
Also the value of main overcommitment technology TPS is significantly decreased with new licensing. The overall rate of memory overcommitment will decrease, but my assumption was that resource overcommitment was one of the main drivers towards virtualization.
Another bad impact is that vsphere admins will try to size their VMs to fit vRAM pools and this can affect VMs performance, thus, undermining trust and confidentce in virtualization technologies. I have had some of such situations before.
They point is… every version we get a new change… always to worst…
I fell like VMWare putting its hand inside my pocket…
Hey all, thanks for the great comments. This is a tricky subject, and it’s hard when a change like this happens that is inconvenient, at best, to everybody.
VMware is going to release a tool to help assess the licensing situation for environments. I’m working on getting a copy for myself and more information. I’ll post it when I know more.
Oh, and with regard to right-sizing, this whole licensing change vindicates my being a RAM Nazi for the last six years…
Here’s another question..
In high-memory-consolidation environments, we are likely to end up with more ‘CPU’ licenses than there are cores..
What about the opposite? Say we have a quad-core box that is CPU heavy, requiring minimal allocated memory to VM’s. Can we get away with only 1 or 2 CPU licenses?
How can VMWARE have it both ways? Its either a memory license or a CPU license..
Re: Ram Nazi
Amen. After P2V’ing dozens of servers, I’ve learned to never tell them I’m cutting their vRAM allocation in half. New servers get nothing extra unless they can justify it in performance after installation. Obviously there are some (Exchange, SQL) that get more – but following VMware’s best practices.
People on my View pools were practically rioting when they first logged in and found 512MB of RAM – before they even launched their first app. Seriously, I actually did test these in dev before I rolled out the pool! :)
Manual calculations? That’s sooooo 2010 ;)
Here’s a script:
http://www.peetersonline.nl/index.php/vmware/calculate-vsphere-5-licenses-with-powershell/
Hugo
I support a rather large and diverse environment that doesn’t have the diversity in VM chargeback that we’ll need to immediately reconcile with this licensing change. We haven’t limited vRAM use and have been bringing in 2-socket servers with 144GB for almost the last two years to run vSphere 4 Ent/Ent+ for our mainstream consolidation. I was already overdue to attempt to calculate our platform costs given our current deployment standards; this will add a fine new wrinkle. I knew that at some point the socket-based licensing model would be adjusted to reconcile with the increasing amount cores per socket, but with this change it feels like the rug has been pulled out from underneath me. At this point I assume that VMware is willing to sacrifice some SMB and less-critical usage for more large-scale/large enterprise profits (think EMC, not so ironically). Even prior to this announcement I’ve been looking at the possibililty of incorporating Hyper-V for some of our lower-tier requirements. This vRAM tax is probably going to accelerate that.
Thanks for calming me down a little bit before I try to reconcile with our current deployments to see how vSphere-5-compatible we’re going to be.
VMWare has simply given Microsoft the entire SMB market with this pricing model. Not everyone is a cloud service provider or can use chargeback. For years we have been essentially taught that memory overcommit was the way to go and now thousands of existing deployed installations are going to essentially be shut out from upgrading to VS5 due simply to costs.
VS4 is good till 2014, and as nice as the new feature set is, I can’t justify a doubling in costs to deploy it. Maybe when the MBA clowns move past the “Cloud” things will move back to normal. Unitl then its VS4 for the forseeable future.
Your setup includes a lot of legacy systems, like your Dell R610 with 24 GB/Socket. No one would spec a new host like this, but those systems defacto “donate” vRAM entitlements to your newer systems.
New(er) installations might not have that luxury.
” I do like the new licensing because I can now assign direct values for chargeback for VMs, based on size.”
Even under the old model there was nothing preventing a chargeback based on VM RAM usage. The new licensing model merely adds another factor into the calculation.
It also really screws setups with hot-standby VMs to guard against things Guest-OS errors: Those systems will sit idle 90% of the time consuming a minimum amount of RAM, but you’d still need to license for the full amount of vRAM you need to configure in case the box goes from hot-standby to active.
“The change from Advanced licensing to Enterprise will also play into it”
This also means that next time your support contract comes up for renewal you’ll need to pay for your Enterprise license, which will most likely be more than for the “old” Advanced license
After seeing the huge storm of complaints I started wondering what on earth I was missing, because it didn’t seem like that big a deal to me. So I sat down with the licensing guide and looked over my environment. I’m an SMB running 4 clusters of VMware in 3 locations, all running Enterprise licenses, with 1-4 servers in each cluster and 24-32GB of ram per CPU. This serves our needs very well(truth be told its probably overkill but when you’re at the extreme tail end of the logistics chain there is comfort in N+1 or better), and the only cluster where I’m even at limits is the ERP cluster.
I guess I’m the one they were talking about where its just a change and nothing more.
I do see a huge change for many, but the obvious question is – why would you upgrade to vSphere 5 if you didn’t need to?
Now this question is coming from someone who is almost always pushing to get the latest updates or versions into production, but this licensing change is putting me into a holding pattern to run the numbers and look over the new features to see if there is a cost benefit to the upgrade.
Nobody is holding a gun to my head saying “5 is out – install it now!”
The second I have to build new vSphere capacity with vSphere 5 licenses because they won’t sell me vSphere 4 licenses is the point at which this becomes a problem for me. I’m not even considering upgrade scenarios yet.
Thanks Bob for the post. This is informative – but this is going to hit SMB’s most and those who don’t care a dime about the cloud and its chargebacks.
I suppose this is VMWare’s licensing conundrum (remember 3.xx to 4.xx anyone ? ) all over again. I personally would have preferred licensing based on Physical CPU / RAM. This method of licensing is a sensible business decision for VMWare by netting more customers where it hurts most (RAM) but not necessarily a customer friendly one.
I am in the small pool of SMB folks to whom this will be “just a change” as my hosts just don’t have that high a RAM/socket ratio. However, I’m running on 3-year old hosts. That tells me that VMWare based their vRAM tax on 3-year old hardware, not current hardware, and certainly not on the hardware that is likely to be available, say, 3 years from now. The ratio of RAM/socket is only going to increase over time, making this pricing scheme more and more outdated. Now, if VMWare had based their vRAM tax on current hardware (say, 256GB of vRAM per socket for E+), few would be complaining. Had they based it on hardware that might be anticipated in 2 years or so (say, 512GB or 1TB of vRAM per socket for E+) they might have been lauded. As it is, they’re being excoriated, and rightly so.
I’m in a strange spot where I’ll be receiving everything I need to build our first cluster shortly. I don’t even have my environment in place yet and it already looks like I’m getting the shaft two years down the road or whenever I’m forced to upgrade to keep up support. I work for a K-12 environment and getting approval for this extremely modest project was a small feat. I’m going to receive two Dell R610s in a couple days with dual cpu/64GB of RAM. We’re opting for the essentials plus bundle. From what I’m seeing, upgrading will mean I’m losing the use of 16GB of physical RAM. How does this help me at all?
I really hope they adjust the EDU/Academic pricing model.
Having read through the licensing white paper, it appears that you can exceed your vRAM entitlement if required, as long as you then add more licenses to bring you into compliance. What if this situation is only temporary? eg. in a disaster recovery situation, I may be creating a large VM as a replacement for a physical server that died, and will be removing that VM once the new physical hardware arrives in a day or so. Am I going to be stung for the extra licenses, even if I (as seen over the course of the whole year) never need them?
After reviewing all the latest features, I would say it should be called vSphere 4.5 instead of vSphere 5.0 as there isn’t much improvements feature wise over the previous 4.1 version.
To my great surprise, VMware launched it’s latest flagship product vSphere in such a hurry, it was originally planed to be released in Q3, 2011 or later. Why is this?
As people say “the devil always lies in the details”, after half reading the latest pricing guide, I quickly figured out the answer to the above question.
It’s all about $$$, VMware tells you the latest vSphere 5.0 doesn’t have any more restriction in CPU/RAM on an ESX host, that sounds so fabulous isn’t it? Or IS IT?
Let’s make a simple example:
Say you have the simplest cluster with two ESX hosts with 2 CPU and 128GB RAM each, you Enterprise Plus edition for these two is USD13,980.
With the previous vSphere 4.1, you have UNLMITED vRAM entitlement and up to 48 cores.
With the brand new vSphere 5.0 pricing model, for the same amount of license (ie, USD13,980), you can only have 192GB entitled vRAM, so in order to have the original 256GB vRAM entitlement, you need to pay extra 2 more Enterprise Plus license, which is USD6,990.
The more RAM your server has, the more you are going to pay with the new licensing model.
So my conclusion is VMware is discouraging people going into cloud in reality. Think about this, why would you buy a Dell Poweredge R710 (2 sockets) with only 96GB RAM installed? The maximum RAM Powerdge R710 is capable of 288GB RAM but you need to pay EXTRA (288GB-96GB) / 48GB = 4 more Enterprise Plus license.
In reality, CPU is always the last resource to run out, but RAM IS! Future server will have much more powerful CPU for sure, but RAM is still the number 1 factor deciding your cloud capacity, IOPS is the 2nd, Network is the 3rd and just to remind you once more, CPU is the last!
Very clever VMware, but will potential customer buy this concept is another story.
Hum…may be it’s a strong sign that I can finally sell my VMW after all these years.
* Please note the above is my own personal interpretation as a user, it doesn’t represent my current employer or related affiliates.
I keep seeing people say this upgrade could be “good” for some customers? How? Best case, you just move to vSphere 5 with no costs. I wouldn’t consider this good, I would consider this the expected result of an upgrade while under paid support! For a majority of customers, this will be a cost added upgrade when we’ve been paying support for the last two years.
I’m very disapointed in vmWare for the spot they’ve put me in since I’ve been the major virtualization pusher in my company. Now I need to go to my boss and let him know our budgets are all wrong, and try to explain what the support/maint that we’ve been paying for is good for?
Mike
I think “good” is relative, Mike. It’s at least not “bad,” which is what a lot of the initial reaction was saying. I’m writing up a new post right now with my last couple days of poking around with futures numbers, to see if I can get growth in my environment to be budget-neutral again! I think I can, but only because of some of the other advances they’ve made (autodeploy, etc.).
Good news is that this new vRAM licensing won’t have an impact on VMware View or desktop environments (even with other solutions). So check out the new VMware Desktop licensing, good cost savings to be had.
http://blogs.vmware.com/euc/2011/07/vsphere-desktop-licensing-overview.html
Also Brian Maddon’s blog, which includes a handy spreadsheet to help with the calculations.
http://www.brianmadden.com/blogs/brianmadden/archive/2011/07/14/vmware-vsphere-5-for-vdi-desktops-licensing-cost-calculator.aspx
Michael, I don’t think that’s quite true. My understanding of their View solution is that you need to purchase the Desktop View server license (or whatever it’s named) which gives you unlimited vRAM. However, this implies that all your desktops are then run on this new, dedicated host (because you can ONLY have View desktops running on it). But I have a client with a 2 host cluster where the server and workstation vm’s are not split out by host. So now i have to either dedicate one of the hosts strictly for workstations (which will break my cluster) or I have to license all 25 of the workstations at the server vRam rate. not good, however you slice it.
Biggest question to me is how the importance of what virtualization starter out to be about seems to forgotten in this scheme:
It looks to me that the higher you consolidate on fewer boxes, the more you end up paying? and the use of Transparent Page Sharing will make it even worse – this used to be a really good argument against Hyper-V on HW cost savings, but if these then goes directly into more licenses.. whats the argument ?
Am i getting this wrong or what..?
Unfortunately, this is going to hammer us. We built out a UCS platform with a lot of 2-proc/192 or 256GB RAM. We also have a VBlock on the way with B230s — 2-proc/256GB RAM. We’re looking at our licensing costs more than doubling, and this was after convincing c-level mgrs to go VMWare. I’m not worried where I am, but I wouldn’t be surprised if some folks who pushed for VMWare get fired over the direction they set.
What really gets me about this, outside of whether or not it costs money, is that the licensing directly contradicts the usage of their product. Virtualization, at it’s core, is about maximizing the utilization of hardware to get the most value from dollars spent on hardware.
I could understand licensing based on physical ram, similar to the physical cpu, however basing it off of the virtual ram used is basically the same as licensing based on virtual cpus. Thus the direct contradiction in ROI. The more I consolidate, the more it costs me.
It’s also frustrating that (if my math is correct), they’ve reduced the amount of ram you can use per cpu license for existing subscribers. We’re use advanced with 6 sockets across 3 hosts, but now can only use 192 GB total across all three hosts instead of 768 GB (256 GB per host). That’s also being extremely generous in assuming 0 over-provisioning of ram (which is never the case).
If we were capped out, it effectively increases costs by 4x or we can stay with vSphere 4. So much for the “free” upgrades if your sns is current.
This also make virtualization of environments weigh extremely heavy in favor of low ram-high cpu applications as opposed to low cpu-high ram applications. If they are going to cap vRam, they should at minimum have a 64 GB per socket for standard and 96 GB per socket for Enterprise. Maybe leave 32 GB for Essentials. Thats much more realistic for nearly all of the environments I’ve seen.
As another commenter stated, it looks like their usage data is a couple of years old, and every year in virtualization is a long time as fast as it is progressing.
To compare apples to lemonade… This is little different then if Microsoft started licensing it’s Windows Server Standard product solely based on the number of cores in the server, and for the same price as the previous version, only allowed 2 core per license. If you want to upgrade a 2 cpu x 4 cores each server, you’d pay 4x as much. It’s insane. But at least with this example, they aren’t contradicting the purpose of the software product.