vSphere 5 Licensing

[SuperSpike]

I took a minute to read the licensing guide for vSphere 5 and I'm still trying to pull my jaw off the floor. VMware has completely screwed their customers this time. Why?

What I used to be able to do with 2 CPU licenses now takes 4. Incredible.

Today

BL460c G7 with 2 sockets and 192G of memory = 2 vSphere Enterprise Plus licenses
DL585 G7 with 4 sockets and 256G of memory = 4 vSphere Enterprise Plus licenses

Tomorrow

BL460c G7 with 2 sockets and 192G of memory = 4 vSphere Enterprise Plus licenses
BL585 G7 with 4 sockets and 256G of memory = 6 vSphere Enterprise Plus licenses

So it's almost as if VMware is putting a penalty on density and encouraging users to buy hardware with more sockets rather than less.

I get that the vRAM entitlements are for what you use, not necessarily what you have, but who buys memory and doesn't use it?

Forget the hoopla about a VM with 1 TB of memory. Who in their right mind would deploy that using the new license model? It would take 22 licenses to accommodate! You could go out and buy the physical box for way less than that today, from any hardware vendor.

Anyone else completely shocked by this move?

[kmcferrin]

taz722 wrote:

However, is it not true that as part of consolidation, customers are moving to more dense servers with more CPUs per chassis. If  I have 8 sockets on my servers and with the limit of 96 GB per CPU , my total entitlement for the server is 96 * 8 = 768 GB, which seems prett high. I think where this licensing hurts is when customers have single/dual socket servers. But,aren't we moving away from those to more powerful servers. What proportion of your servers are small vs. large ?

More dense servers?  Yes.  But not necessarily with 8 sockets.  If you look at it buying anything other than a rackmount server with 8 sockets isn't possible.  Many people are virtualizing on blades that have 2 or 4 sockets.  This makes sense because buying a single 8-socket server can be quite expensive, especially compared to buying four 2-socket blades, or two 4-socket blades, and blades tend to allow lower hardware costs and greater hardware density already.  Another issue is that depending on how big your environment is, using 8-socket servers is putting too many eggs (VMs) in one basket (host) for comfort.  Finally, as someone else pointed out, you can get processing power in a 2-socket blade today that eclipses what was available just 2-3 years ago in an 8-way box.  Going with high-density, high-capacity "commodity" hardware makes a lot more sense.

I can't speak for everyone, but in my experience (in SMB and enterprise space) an 8-way virtualization host would be the exception rather than the rule.

[JAndrews42]

>Customers who had Enterprise back then haven't seen much of an increase in value of their licenses.

Of course everyone wants all the new features and technologies for free, that's only natural.

However, since the original upgrade (from Ent to Ent+) was only $350/CPU I think VMware did a good job of giving customers an inexpensive way to keep the top license tier.

[hmtk1976]

JAndrews wrote:

>Customers who had Enterprise back then haven't seen much of an increase in value of their licenses.

Of course everyone wants all the new features and technologies for free, that's only natural.

However, since the original upgrade (from Ent to Ent+) was only $350/CPU I think VMware did a good job of giving customers an inexpensive way to keep the top license tier.

That and higher SnS.  I though we paid SnS to get upgrades, not to pay for research so VMware can create new SKU's so they can withhold new features from companies paying for support.

[GaryHertz]

tomaddox wrote:

We just deployed a bunch of dual-socket servers with 12-core CPUs and 256 GB of RAM. On a per-socket count, they're smaller, but they have waaaaaay more horsepower than our old four-socket dual-core systems (triple the CPU cores with CPUs four generations newer) at half the footprint. This is the root of VMware's issue: as enterprise customers hit their asset refresh cycle, they can substantially increase their system density without increasing their license count.

Only problem is it's not true.  With anything but advanced or enterprise + you would need to double your licenses to support those 12 core sockets under 4.1 licensing.  With either advanced or enterprise + you are capped at 12 and will need to buy more licenses when higher density cores become available.  Performance core to core is better when upgrading from an older platform to a newer platform but upgraded OS and software requires the increased performance.  Lets put a stop to the core density argument.  If you look at the facts it isn't true.

Look at this from another angle.  I've reduced 70 physical servers down to 4 with the potential of doubling that on my current hardware.  Thats a loss to Dell of 66 servers and a potential loss of 136.  Same would hold true if I purchased from HP, IBM or Cisco.  I don't see any of them going bankrupt.  Are they coming up with a licensing sceme that is based on vServers to recup those losses?

[tomaddox]

Gary H wrote:

tomaddox wrote:

This is the root of VMware's issue: as enterprise customers hit their asset refresh cycle, they can substantially increase their system density without increasing their license count.

Only problem is it's not true.  With anything but advanced or enterprise + you would need to double your licenses to support those 12 core sockets under 4.1 licensing.  With either advanced or enterprise + you are capped at 12 and will need to buy more licenses when higher density cores become available.  Performance core to core is better when upgrading from an older platform to a newer platform but upgraded OS and software requires the increased performance.  Lets put a stop to the core density argument.  If you look at the facts it isn't true.

Look at this from another angle.  I've reduced 70 physical servers down to 4 with the potential of doubling that on my current hardware.  Thats a loss to Dell of 66 servers and a potential loss of 136.  Same would hold true if I purchased from HP, IBM or Cisco.  I don't see any of them going bankrupt.  Are they coming up with a licensing sceme that is based on vServers to recup those losses?

Don't get me wrong, I agree with you. I suspect that the VMware and EMC execs are looking at these increasing server densities, though, and are anticipating a decrease in existing customers purchasing new licenses. What they've failed to realize, however, is that their new licensing model will have the same effect, with the added result of driving customers into the arms of their competitors.

[kcucadmin]

personaly it's becoming very clear to me that vmware see's it's growth model as extracting more and more cash from it's existing base, not growing the base.  Frankly, I dont see us staying with vmware much longer, as soon as some viable alternatives become available we will make the switch.

[GaryHertz]

tomaddox wrote:

Don't get me wrong, I agree with you. I suspect that the VMware and EMC execs are looking at these increasing server densities, though, and are anticipating a decrease in existing customers purchasing new licenses.

Once again increased core densities will not lead to a loss of revenue under the 4.1 model.  Cores/socket are capped therefore as core densities increase additional licenses are neede to support the additional cores.

I didn't disagree with the rest of your post.  I'm just tired of seeing the lie which is increased core densities will lead to a loss of revenue.

[Baddos]

I agree that if VMware is trying to push us into a consumption model, there shouldn't be a traditional capex cost and maintenance cost. Right now we've paid for vmware tells us to condense our vms in to fewers hosts, but then on the other hand they want more money for doing so.

People try to claim that VMware needs more money because future sales will hurt from increased hardware resources, yet they discount the fact the applications only continue to grow at a steady rate as well.

[kcucadmin]

Baddos wrote:

People try to claim that VMware needs more money because future sales will hurt from increased hardware resources, yet they discount the fact the applications only continue to grow at a steady rate as well.

I totally agree, we use to only have 8 physical servers to meet all of our server needs, in less than 2 years, (once I went virtual) that exploded to 70 servers.  Mostly for "Work Load" isolation. 

Once you detach a "Work Load" from being constrained to physical hardware, it's very tempting to allow vm bloat.  just to simplify troubleshooting, application isolation, security isolation.

honestly hyper visor licenses should be tied to hardware, perhaps they should explore a "CAL" based license model for FEATURE sets per VM. 

that way i can say these are non critical vm's, standard cal,  theses are tier 1 vm's, ent+ cal.

it will scale better LONG Term for them than having to "Adjust" licensing models every few years to account for moore's law.

i know allot of folks dont want to have to pay by the VM model, it's too dynamic, etc etc etc.  i would suggest using a POOLing system like vRAM and entitlements... i.e. a Ent+ Cal entitle 10 vm's to be powered on, etc.  who knows what the exact numbers would be, but i just see vRAM being a issue long term, i see socket count being an issue long term.

[sergeadam]

You ultimately don't want a licensing model on vAnything. While that model works well in a hosting environment, most of us do host or even contemplate chargebacks.

The licensing has to be tied to hardware. That is a fixed, known entity. VMWare also needs to get to the realization that software is moving the way of hardware, which now follows the razor model. You move software to get the customer to pay yearly maintenance. That is your stable revenue stream. I've been using ESX for over 7 years. I've called support 3 times so far. I pay SnS for 2 reasons. It's easy to budget. It funds their development and gets me the next version.

Look at the XenServer model. The software is cheap. Support, not so much. 

[tomaddox]

sergeadam wrote:

You ultimately don't want a licensing model on vAnything.

I've called support 3 times so far. I pay SnS for 2 reasons. It's easy to budget. It funds their development and gets me the next version.

I am in full agreement with the first statement. I am not interested in paying for consumption. I have used support a fair amount, however, so I definitely value VMware's support resources; I also expect that paying SnS will fund bug fixes and patches that keep me from needing to call support, improving my experience and reducing VMware's costs.

[sergeadam]

Don't get me wrong. I fully appreciate the support. In those 3 cases, they've pulled my gonads out of the fire. Point is, many people don't need the support resources much, leaving that revenue to fund software development.

[ph0bia]

Support isn't the issue here, I too am pleased with their quality of support and hope that it stays that way.  The bottom line is that enterprise plus customers currently paying for maintenance suddenly have to buy more licensing to properly cover their hardware for vSphere 5, or accept the fact that that can no longer fully utilize their hardware - and this is unacceptable.

How did VMware get this great customer base and all of their revenue to start with?  Simple, by creating the very best product in the market space, that's it - no other magic to it.  If they lose their focus and work on complicating their licensing models and anger their customer base in so doing, then they will fail, if they get back to their roots and continue to focus on making virtualization more and more awesome while keeping licensing simple and reasonable, they will continue to be the kings of virtualization.

[rjb2]

sergeadam wrote:

I pay SnS for 2 reasons. It's easy to budget. It funds their development and gets me the next version.

We view SnS exactly the same.

[skywalkergcm]

Where's the Green?

VMware needs to increase the vRAM allocations per licensing unit further to reflect what's capable in today's hardware space and above all make Enterprise Plus UNLIMITED on all counts period!  Take care of the Enterprise clients as well as the SMB clients on both ends of the scale.  Clients in the Enterprise + space are already paying a premium licensing fee!  While I can see paying a bit more per socket for the new / added features, I cannot see the substantial fee increases for larger scaling solutions.  The finance guys will laugh us right out of the room!

If VMware considered these figures w/ their largest clients in mind, then I do not see how VMware justified, given today's financial constraints / environments, that we as Enterprise clients can sell these added fees to our finance departments?   Perhaps VMware believes clients cannot switch to another hypervisor provider or that we may be _forced_ to switch from the powers-that-be on high?  This is a dangerous game.  Once a decision is made on-high to vacate VMware from an operational environment, it will happen no matter how painful.

The effect of the vRAM pricing will force many designers architects considering scaling UP to 1024GB - 1536GB per 4 socket host and subsequent large numbers of VMs with larger vRAM needs to reconsider VMware solutions.  Suddenly, using many smaller less efficient systems, all creating a more negative environmental impact, or alternative hypervisiors now becomes attractive.  Somehow I do not think this is the correct message or what VMware intended to encourage while preserving or increasing its revenue models.

For example, in our setups, we run fairly over committed and push our existing host to ~ 80% REAL pRAM utilization.

Doing some quick math the "vRAM TAX" for Enterprise Plus would cost about ~ $25K PER HOST in ADDITIONAL licensing fees when pushing the vRAM allocations to ~ 1024GB or a total of ~ $39,000 per HOST in just vSphere 5.x E+ licenses.

For an ultra aggressive large-scale client targeting 1536GB per host and assuming vRAM allocations near that, then that's an added $42,000 per HOST or a total of  $56K per HOST just in vSphere 5.x licensing fees!  Unaffordable, unjustifiable and Yikes!

I did not even add in the SnS support fees for this but they are also substantial.

The net effect of this vRAM fee on Enterprise clients is that it discourages us financially from deploying "super Green" high-density solutions which which are the most cost, power, space efficient and environmental friendly.  Certainly today's vRAM quotas are not designed to encourage scaling up solutions but to wildly increase revenues being extracted from VMware's most efficient high-scaling clients - that top 5%.. hmmmm that play is familiar.

  
I can only hope VMware sees the headlights on this train before the impact is felt.

[wdroush1]

Duncan wrote:

ph0bia wrote:

So it appears that I am looking at about a 400% increase in license costs for all of my new dual 12-core bl465c 256GB blades, we have dozens of blades altogether and I guess VMware wants me to now go to my bosses and explain to them why suddenly our license costs are so much higher.  This at a time when competitors such as Microsoft come in and visit the CxO's directly, offering them pretty much 'whatever it takes' for us to consider switching our core environment to their solutions.  I can only imagine that M$ is full of glee over this change, and I can say with great certainty that I am not going to approach my bosses with this price increase - to do so would be occupational suicide.  My recommendation will be that we end our maintenance agreement immediately.

Assuming you are using Enterprise+ how would that be a 400% increase? Or are you overcommitting 200%?

I've heard of 200%+ without overcommit using TPS and very-like boxes.

[wdroush1]

J1mbo wrote:

We just deployed a bunch of dual-socket servers with 12-core CPUs and  256 GB of RAM. On a per-socket count, they're smaller, but they have  waaaaaay more horsepower than our old four-socket dual-core systems (triple the CPU cores with CPUs four generations newer) at half the  footprint. This is the root of VMware's issue: as enterprise customers  hit their asset refresh cycle, they can substantially increase their  system density without increasing their license count.

How exactly is this any different to any server refresh cycle of the past 30 years?

Well VMWare has somehow convinced people that this multi-core technology somehow doesn't apply to Moore's Law. If it wasn't for multi-core technology I'd expect CPUs in the 20+GHZ range, but you start to run into limitations with that apparently IIRC.

[wdroush1]

Duncan wrote:

Bigi wrote:

The only problem with the licensing now is that vRAM is based on RAM assigned to VM and not actual RAM used. So much for the overcommit feature, makes it worthless.

vSphere 4.x runs great, people that have more RAM than currently will be licensed under the new model should stay with 4.x. Lets see how the numbers turn out in the next 18-24 months and maybee version 6 will be more RAM friendly.

overcommitment savings is more than "licenses". what about the cost of hardware? what about the cost of management? what about the cost of network / storage I/O ports? it is not worthless, far from!

Hardware is nearly as cheap as licensing alone (for RAM that is, at this point I'm buying RAM twice), network/storage infrastructure is negligible when dealing with 5 digit licensing fees. The amount of additional use I get out of overcommit isn't going to be some kind of monumental savings for me, and actually means that I need to buy hardware to sustain growth anyway.

[wdroush1]

Gary H wrote:

tomaddox wrote:

We just deployed a bunch of dual-socket servers with 12-core CPUs and 256 GB of RAM. On a per-socket count, they're smaller, but they have waaaaaay more horsepower than our old four-socket dual-core systems (triple the CPU cores with CPUs four generations newer) at half the footprint. This is the root of VMware's issue: as enterprise customers hit their asset refresh cycle, they can substantially increase their system density without increasing their license count.

Only problem is it's not true.  With anything but advanced or enterprise + you would need to double your licenses to support those 12 core sockets under 4.1 licensing.  With either advanced or enterprise + you are capped at 12 and will need to buy more licenses when higher density cores become available.  Performance core to core is better when upgrading from an older platform to a newer platform but upgraded OS and software requires the increased performance.  Lets put a stop to the core density argument.  If you look at the facts it isn't true.

Look at this from another angle.  I've reduced 70 physical servers down to 4 with the potential of doubling that on my current hardware.  Thats a loss to Dell of 66 servers and a potential loss of 136.  Same would hold true if I purchased from HP, IBM or Cisco.  I don't see any of them going bankrupt.  Are they coming up with a licensing sceme that is based on vServers to recup those losses?

Err, that is still 128GB/license, for not bleeding-edge hardware (you can actually get more RAM a socket than that), with TPS, overallocation and compression (so I can push ~160GB easily, 200+GB if intensively packing them). Where as now you get 96GB/license, no TPS, no overallocation, no compression.

[sergeadam]

VMWare got where it is because it’s product was head and shoulders above everybody else THEN. I had no problem standing on front of a row of CxOs and justify spending money on VMWare because it was that much better.

They’re still the best. But not by much anymore. With 4.1 I could still justify the additional costs. I can’t justify v5. I might stay with v4.1 until EOL. By then I’m hoping one of 2 things happen. VMWare comes up with a real licensing model. Or the competition passes them.