vSphere 5 licensing and Moore's Law

[officeglen]

I have not fully evaluated the impact of the new vRAM licensing on my environment, but I suspect that I will be in reasonably good shape for the next few months. My concern is what happens to vRAM over time?

One thing that we've all relied on is the impact of Moore's Law on DRAM capacity and CPU performance. Over my time as a VMware customer I've upgraded my environment in a fashion that very closely followed the popular IT constant. With this new licensing change is VMware doubling my licensing costs every 18 months? When will the rules change next? How can we plan for the future when the rules keep on changing?

I will be definitely evaluating my virtualization options over the coming months. While I still believe that VMware is best virtualization platform and I can't help but feel betrayed and disillusioned.

My hope is that some day future marketing students will study 'New Coke' and 'vRAM' in the same class.

[DSeaman]

New Coke, VMware, and NetFlix...would be a very interesting class indeed.

[wdroush1]

Am I not really surprised if we realize that VMWare hasn't really thought this though beyond "$$$".

[J1mbo]

I'm not sure if there are limits on what opinions I can post now with vE status... but it is very concerning to me indeed that there has been absolutely nothing from vmware on how the limits will change over time.

[kcucadmin]

what's even more concerning is the current entitlements don't even cover some of the very basic Hardware configs there hardware partners are rolling out.

the UCS platform: one of it's major features was pRAM density and vRAM overcommit.  wow.  that feature just got killed with licensing requirements that would cost more than the actual hardware.

48GB ram just isn't very much RAM these days. Win2k8R2, win7 are all 64 bit OS's, the whole point to move to 64bit os's was to move beyond the 2GB/2GB limits of 32 bit os's.  now more and more apps are coming online that support 64bit to leverage more RAM and now all of a sudden i'm restricted to 48GB per processor?

that is so counter to what has been happening.  What were they thinking?

96GB would make more since.

[weinstein5]

Hope fully VMware is paying attention to some these discussion on this board and will rethink the policy -

[FredPeterson]

Robert Samples wrote:

... now all of a sudden i'm restricted to 48GB per processor?

Nowhere does it say you are restricted to 48GB per processor.  You are entitled to 48GB of assigned vRAM per Enterprise Plus license.

Yes, a license can apply to a socket as well, but for the purposes of memory it has no connection to a host.

The licensing model will be two fold.  You input a license into the pool - it then adds whatever the vRAM entitlement is to the pool (based on license type) and makes the vSphere license available to be assigned to a host.  The license never actually has to be assigned to anything.

WIth that said, I am in complete agreement that this licensing change is terrible news many SMB's who optd for heavy consolidation due to the cost of memory and are now likely to be heavily marketed at by their local Microsoft VAR's or MS reps.

PS: I have no idea why it chose to say "Robert Samples wrote" instead of kcucadmin...

Message was edited by: FredPeterson

[J1mbo]

I don't know why this issue isn't gaining traction - only 500 views compared to 37,000 on the licensing thread?  Yet it builds in obsolescence for v5 in about 12-18 months if vRAM entitlements remain static.

Maybe the plan is to offer cloud services at prices around the vSphere 5 host license costs at that point (which may well be a realistic option for many businesses by then) through some system of deep-discounts for hosting providers.

It can be cost effective to buy hosts with enough RAM for 12-18 months and upgrade after that, for example 18-months ago 48GB in 8GB DIMMs was about £5k in the UK, now it's about £1k.  Of course with v5 licensing vmware would require (assume standard) licenses of about £1.5k to use it.

Perhaps that's extreme, but wind forward 18 months then £1k is likely to buy 96GB but licensing for it will be £3k.

3 years from now, the same £1k might buy 192GB but licensing for it would be £6k.

Very clearly this just doesn't work commercially - maybe point releases for v5 might steadily double vRAM entitlement?  Or perhaps the ordinary 3-year write-off period is simply too long and server hardware should be effectively expensed?

[wdroush1]

J1mbo wrote:

I don't know why this issue isn't gaining traction - only 500 views compared to 37,000 on the licensing thread?  Yet it builds in obsolescence for v5 in about 12-18 months if vRAM entitlements remain static.

Maybe the plan is to offer cloud services at prices around the vSphere 5 host license costs at that point (which may well be a realistic option for many businesses by then) through some system of deep-discounts for hosting providers.

It can be cost effective to buy hosts with enough RAM for 12-18 months and upgrade after that, for example 18-months ago 48GB in 8GB DIMMs was about £5k in the UK, now it's about £1k.  Of course with v5 licensing vmware would require (assume standard) licenses of about £1.5k to use it.

Perhaps that's extreme, but wind forward 18 months then £1k is likely to buy 96GB but licensing for it will be £3k.

3 years from now, the same £1k might buy 192GB but licensing for it would be £6k.

Very clearly this just doesn't work commercially - maybe point releases for v5 might steadily double vRAM entitlement?  Or perhaps the ordinary 3-year write-off period is simply too long and server hardware should be effectively expensed?

The issue is covered multiple times in the licensing thread, but to be honest, it's a minor part being as the new licensing wrecks a lot of our setups before Moore's law applies to us.

[DSTAVERT]

I think the instant companies decided that they could trade VMware licenses for pRAM purchases the licence change became critical. VMware has a fiduciary responsibility to protect shareholders investment and license revenue erosion hit at VMware's core asset. I think it is unfortunate that the vRAM entitlement change was not introduced with vSphere 4.1. Had it been introduced earlier before the economics of hardware VS licenses tipped in favor of hardware this wouldn't have had the same impact. It would also have given VMware time to fine tune entitlements and allow organizations to grow into the change.

What the right numbers are remains to be seen. Is there anything that will be fair to all sides? I think everyone will need to look deeply whatever their ultimate decision. VMware must maintain revenues otherwise it fails. If VMware does make changes to entitlement levels and or add license packs bless their hearts but continued loss of core revenue would not be good long term for VMware or their customers. It doesn't take a financial genius to see that it wouldn't take long for VMware Licensing revenues to be cut in halves every few months given the hardware trends. I do hope something takes place to easy the transition but I also hope people take a more balanced view of the situation.

[scottyyyc]

The reality is, quite simply, Moore's law kills per-processor licensing in a virtual world. This is not a small decision on the side of VMware; they've known about this for years, and I'm sure have bantied about the decision for just as long. As with what dstavert said, eventially VMware has to address this somehow. I don't think vRAM is the answer, though, or not at the new proposed costs. Perhaps a per-core strategy would have worked better, because at least then you're riding the multi-core bandwagon.

I dunno, but I'm strongly opposed to this licensing change. It just stopped my v5 upgrade plans dead in their tracks. Not to mention, Ent+ was just a little too expensive for my company, so we went with standard. With only 24GB vRAM / processor license, we're well below our current usage (128GB per 4 proc server). Now RAM upgrades have to coincide with VMware licensing upgrades, which aren't cheap.

As I said on the licensing thread, looks like I'll be getting know know 4.1 really well over the next couple years.

[wdroush1]

DSTAVERT wrote:

VMware must maintain revenues otherwise it fails.

Losing 6+ digit contracts is how you do that, right?

The problem if VMWare wants to protect revenues but it's very quickly having two competitors that are multitudes cheaper than it come up alongside it in features. Their response is to choke the customer out and hope that when they come to they wont be going over to their competitor because they don't have cute vRAM mascots (http://twitter.com/VMware_vRAM).

VMWare is pretty much going to lose revenue due to competitors anyway at this point (the competitors are taking VMWare's market share rapidly), the only way they'll keep it stable is by volume... and that requires they find some miraclous way to basically reclaim the marketshare that Microsoft and XenServer alreay has taken (even if vRAM didn't happen, it would be hard to get VMWare's licensing to a competitive price point). Jacking up pricing is just going to lose marketshare quicker (it's the only reason Hyper-V and XenServer have any traction to begin with).

A drop in revenues isn't death to VMWare (we're looking at north of $1bn in profits guys...), especially considering how much further out their market is spreading with desktop and application virtualization (and dabbling in mobile virtualization), the server hypervisor market is just starting to mature across more companies and VMWare is finally getting some competiton.

[kcucadmin]

if VMware's goal here was to protect license erosion then they should not of been pushing the market via their partners to scale up to 100'sGB of pRAM.

Cisco and HP both have really been pushing larger pRAM models using VMware to leverage the CPU and RAM capabilities of today's hardware.  for them to basically say LOOK what our software can enable you to do, and then allow that for ONLY 2 YEARS (ucs, high density pram blades).  then pull a HUGE bait and switch mid hardware cycle is poor strategic planning.

companies like VMware don't dry up that fast, if they wanted to switch to the new model, fine.  I understand.  but the way they have been leading their customers to THINK has been super dense high pRAM consolidations, now thier saying oh by the way to fully use that pRAM we would like you to pay considerable more in vRAM licensing.

all i can think is Bait and Switch.  for years they have been building this SCALE up approach.  and just like that, over night.  BAM!  uh sorry we really only wanted you to be able to allocate 48GB per license.

can they not see how this might appear to their customers, people who's jobs are to plan for 3-5 year buckets.  you have to give some flexibility for something like this. or at least a heads up 12 months before it happens to give people time to adjust projects/budgets etc.

i've heard rumors that i have 30 days to make the decision to switch my licenses to 5.0 or loose out forever?  again  strong arm bait and switch.

[wdroush1]

scottyyyc wrote:

The reality is, quite simply, Moore's law kills per-processor licensing in a virtual world. This is not a small decision on the side of VMware; they've known about this for years, and I'm sure have bantied about the decision for just as long. As with what dstavert said, eventially VMware has to address this somehow. I don't think vRAM is the answer, though, or not at the new proposed costs. Perhaps a per-core strategy would have worked better, because at least then you're riding the multi-core bandwagon.

I dunno, but I'm strongly opposed to this licensing change. It just stopped my v5 upgrade plans dead in their tracks. Not to mention, Ent+ was just a little too expensive for my company, so we went with standard. With only 24GB vRAM / processor license, we're well below our current usage (128GB per 4 proc server). Now RAM upgrades have to coincide with VMware licensing upgrades, which aren't cheap.

As I said on the licensing thread, looks like I'll be getting know know 4.1 really well over the next couple years.

Not really, as processing power increases, the cost of the same slice of work x amount of years ago should become cheaper.

[wdroush1]

Robert Samples wrote:

if VMware's goal here was to protect license erosion then they should not of been pushing the market via their partners to scale up to 100'sGB of pRAM.

Cisco and HP both have really been pushing larger pRAM models using VMware to leverage the CPU and RAM capabilities of today's hardware.  for them to basically say LOOK what our software can enable you to do, and then allow that for ONLY 2 YEARS (ucs, high density pram blades).  then pull a HUGE bait and switch mid hardware cycle is poor strategic planning.

companies like VMware don't dry up that fast, if they wanted to switch to the new model, fine.  I understand.  but the way they have been leading their customers to THINK has been super dense high pRAM consolidations, now thier saying oh by the way to fully use that pRAM we would like you to pay considerable more in vRAM licensing.

all i can think is Bait and Switch.  for years they have been building this SCALE up approach.  and just like that, over night.  BAM!  uh sorry we really only wanted you to be able to allocate 48GB per license.

can they not see how this might appear to their customers, people who's jobs are to plan for 3-5 year buckets.  you have to give some flexibility for something like this. or at least a heads up 12 months before it happens to give people time to adjust projects/budgets etc.

i've heard rumors that i have 30 days to make the decision to switch my licenses to 5.0 or loose out forever?  again  strong arm bait and switch.

To be honest, large pRAM boxes was the only way to make VMWare cost effective for a lot of places.

Also, I hate scale out, it really ruins a lot of the efficiency of consolidating in the first place, why buy a half of rack of fairly low-mid end servers when I can buy 3Us worth of high end servers and not 3-4 times as much hardware to power, swap parts on, etc.

[ats0401]

I'm sorry, but these arguments about the old licensing model killing them is crap.

Is this the "license revenue erosion" or the "continued loss of core revenue" you are talking about, that was announced yesterday?

VMware on Tuesday reported fiscal second quarter earnings that exceeded Wall Street's expectations, driven by strong enterprise licensing agreement sales and steady maintenance and professional services growth.

VMware's Q2 profit nearly tripled to $220 million, or 51 cents a share, while revenue rose 37 percent to $921 million. Excluding items, VMware's net income was $235 million, or 55 cents per share, easily surpassing the 47 cents per share that analysts polled by Thomson Reuters had been expecting.

VMware's Q2 licensing revenue rose 44 percent to $465 million,

while software revenue, which VMware defines as software maintenance and professional services, grew 30 percent year-over-year to $456 million. Operating income was $291 million, up 56 percent from last year's quarter, while operating cash flow grew 114 percent year-over-year to $463 million.

[J1mbo]

Exactly - core count growth has only really been a mechanism to sustain CPU processing speed growth at rates the market expects (and demands).  Look at the relative performance of 2x Intel 5680's vs. 2x Zeon 2GHz.... guess what this works out at c.60% annual compound growth rate between their respective launches.