VMware Cloud Community
SuperSpike
Contributor
Contributor

vSphere 5 Licensing

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?

@Virtual_EZ
Reply
0 Kudos
1,980 Replies
hmtk1976
Contributor
Contributor

The crippled Essentials (Plus).  That hard limit on vRAM can be painful.

Reply
0 Kudos
IT_Architect
Enthusiast
Enthusiast

>What licensing was not cost effective and what was the change to make it cost effective?<
What I'm talking about is when XEN and Hyper-V made their big push.

>The  major changes I can think of were when SMP came included in all  licenses (3.0 I think) So when did the math start working out?  And can you give a  practical example of the benefits of setting cores per CPU?<
WOW!  Without the ability to assign vcores to a vcpu, each vcpu was seen as a physical processor.  A lot of software is licensed by physical cpu and they don't care about cores.  Just about everything Microsoft sells would apply.  How about SQL Server?  How about Microsoft Server?  When they allow 4 CPUs, would you like those CPUs to all be single core?

>Hardware just doesn't cost that much, there is no new learning curve, and you don't have all of your eggs in one basket.

So if a VMware server goes down, somehow the VMs in it stay running, all with current information.  I didn't know that.

>Who's floating who?<

Check into what the current concerns about EMC's future.

>They  are not using a "usage-based" model but an allocation-based model<
Exactly!  If they were already using a usage based model I would not have suggested it.


>>Making VMware cost about 25% of the benefit customers derive from it would maintain broad appeal.

>The  sales pitch should be the value of the added features which would be  difficult or expensive to add to existing infrastructure.  HA,  snapshots, vMotion, DRS/DPM, DR replication (req SRM5), thin  provisioning and the flexibility of VMs.
Those are the things that people get addicted to after the fact.

>>When  I was CTO, and a a salesman used that on me I was thinking, "That is  one stupid  bank.  They should have been on CITRIX or Terminal  Services."

>Not everything runs on Citrix or TS, which hopefully the "stupid" bank was aware of.
They weren't.  It was not that far from us.


>Tho  my first View client started off the project with "We will never do  Citrix or TS ever again" and has loved their 120 seat View install.<
I've done entire factories on CITRIX.  I like the simplicity of TS/MDS also.  I've never worked with VDI, but I'm glad you brought it up because I have a multi-location investment firm that I need to do very soon.

Reply
0 Kudos
aroudnev
Contributor
Contributor

This (about cost efefctivity of Vmware licensing) is not correct. Vmware was ALWAYS effective IF you use combination of Free version + Essential (or Infrastructure) and Standard / Enterprise. The problem is that Vmware sales never ever said a word about 3.5i, 4.x ESXi free versions, Essential and other cheap alternatives, so they disteracted a lot of users even before Vmware 5. If you even knew about ESXi3.5 and 4.0 free version, you was always able to prove cost effectiveness for Vmware (yes, maybe you will license 6 standard systems and keep 6 more on free licenses for a while for QA or development, but you could always fit the licensing into your budget).

In our case, we discovered free version and cheap versions just in time to be able to encourage all IT to switch to the Vmware. AS time show, most upgraded to different enterprise versions so Vmware did not lost money. But the very existance of cheap and free versions was a key in this success.

I had a few cases with my friends who came to Vmware, got a proposal with a huge price tag, and was ready to give up - and all stay wth Vmware when I show them alternatives (free version, and essential version). Most already upgraded to th emore exxpensive licenses. Again, free and cheak versions was a key element in this success.

It does not work with Vmware 5 (free and essentail limitations are too strict) and at the same time, just now more competition appear in this area. So I will not be surprised seen declined sales in Vmware.

Reply
0 Kudos
admin
Immortal
Immortal

>Hardware just doesn't cost that much, there is no new learning curve, and you don't have all of your eggs in one basket.

So if a VMware server goes down, somehow the VMs in it stay running, all with current information.  I didn't know that.

Research Fault Tolerance (circa 2009) and High Availability (circa 2006) then I'll dive into your "it wasn't cost effective until they raised prices" argument.

Reply
0 Kudos
IT_Architect
Enthusiast
Enthusiast

>>Hardware just doesn't cost that much, there is no new learning curve, and you don't have all of your eggs in one basket.

So if a VMware server goes down, somehow the VMs in it stay running, all with current information.  I didn't know that.

>Research  Fault Tolerance (circa 2009) and High Availability (circa 2006)

I attempted to find out what you were talking about.  Everything I found about it references the future tense.  That would be useful information for me if it is real  The only time I've had that type of environment available to me was Netware starting in the early 90s, which wasn't virtual and pretty resource intensive.  It should certainly be much easier to do in a virtual environment, and not matter so much what the platform is.  However, it would still not be as simplistic as it sounds.  Again, I didn't see evidence that they have pulled it off yet.  Restarting on another host with all sessions lost is the only one that I know about.

Reply
0 Kudos
admin
Immortal
Immortal

Fault Tolerance was released with vSphere 4 and works great with 1 vCPU.

High Availability is available starting with ESS+ and will restart a VM for Host, VM or application faults in ~16 seconds + OS boot time.

In my book that beats calling your server vendor and hoping the tech has the right part when he shows up in 4 hours.  The only way to duplicate it with hardware is boot-from-SAN and owning spare identical servers.  Good luck automating that.

>Restarting on another host with all sessions lost is the only one that I know about.

Ah, what technology are you comparing this too that will keep sessions?  One of those $70k parallel server setups?

Reply
0 Kudos
Dracolith
Enthusiast
Enthusiast

Joshua Andrews wrote:

>Hardware just doesn't cost that much, there is no new learning curve, and you don't have all of your eggs in one basket. 

Unless you are buying a hardware-clustered box your eggs are certainly in one basket with hardware - something VMware has *not* been since 2003 and the release of vMotion.

"Eggs not in one basket"  with physical servers is simple -- it means that if one server fails,  only the services dependant upon that particular server OS instance necessarily experience downtime.     With ESX;  all the services  that were living on that server experience downtime, which will be multiple servers.

That shouldn't be hard to see -- if you have 5 VMs on a physical server, and that server dies,  you have 5 OS instance that experience downtime.

Depending on the nature of the failure,  it may result in data corruption --  after the server failure,  you may have 5 VMs that are permanently damaged and cannot be powered back on,  even after you replace the faulty server component, or transfer hard drives to a new server.

If you have 5 physical servers, and a single physical server dies, you have  1   OS instance that experiences downtime.

Although in the aggregate, with 5x as many servers,  you can expect on average  that hardware failures will occur 5x as often statistically, than if you had one server,  assuming failures statistically follow a normal distribution.

In that manner you do have all eggs in one basket -- although with ESX, the total amount of downtime per OS instance may be less statistically;

it requires that you use HA  to achieve that;   without HA,  virtualization makes your environment 5x less fault tolerant due to 5x as many single points of failure  for 5  server instances.

With ESX, even with HA, your eggs are still in one basket.   They just call the basket something different, it used to be called "server" with DAS,  now it's called "SAN".   If your SAN box goes down, all your servers stored on it go down.

SAN total failures tend to be slightly less frequent than server failures;  more frequently, the issue that crops up with these kinds of systems are performance issues.

Hardware  shared-nothing architectures can easily be designed to be more fault tolerant than ESX or anything else relying upon shared components

between failover nodes,  they are just vastly more expensive  and hard to justify.

ESX clustering with a SAN  is actually a lot less expensive than high-availability solutions that avoid placing eggs in one basket;

the availability assured is much less,  but it is a good cost tradeoff for many situations.

Reply
0 Kudos
icontrolyourpow
Enthusiast
Enthusiast

I will be out of the office 12/6 and 12/7, however I will be checking email periodically and will be reachable on my mobile phone for emergencies.

If you need immediate assistance, please contact David Dailey.

Reply
0 Kudos
IT_Architect
Enthusiast
Enthusiast

>Ah, what technology are you comparing this too that will keep sessions?  One of those $70k parallel server setups?<

NetWare SFT-III was a Netware add-on in 1993.  By 1995 it was $1495 for 100 users and cost $3995 for more than 100 users.  It required two identical boxes.  Starting with Netware 4.2, it was included in the install, but you still had to buy the licenses.  The biggest drawback is performance on a busy server.  Today's hardware and virtualization are a perfect match.  However, full SMP and pre-emptive multi-tasking makes coordinating non-deterministic I/O an incredible feat.  I see the VMware guys working with it are talking about one CPU now, which I'd wager means one core at the VM.  Even if it's not so quick when they get the first version out, I'll have a use for it if/when they get it finished.  My current need can be met by CARP configured failover + load balancing in a web app environment where the routers themselves are virtual.  That requires ~1 to 3 seconds to cut over, depending if the failure happens with a vm or a vrouter where a physical switch needs time to understand what happened when all of the sudden the IP is associated with a different MAC on a different port.  However, I do lose my sessions.  (This conversation has been good for me because it generated some brain flatus on how I MIGHT be able to mitigate most of that.)  However, that's a much simpler scenario than SFT-III and what VMware is currently working on.  You got my heart palpitating when I thought you were suggesting that VMware currently has SFT-III type capabilities.  My "misteak".

Reply
0 Kudos
srnhpp
Enthusiast
Enthusiast

                               
Reply
0 Kudos
aroudnev
Contributor
Contributor

I can't agree. WIth VM, I can easily restore any system from the image backup in a few hours, I can run HA (using SAN to store VM-s) or can

use modern replicate system to the standby copy on the fly technology (fault tolerance?) and so on. With HW< I don;t have options at all.

Then, it is easier to have one good server with 5 VM-s vs have 5 not-so-good serevrs with HW systems (all different so all backup systems are different).

Of course, VM method requires better plannng, else you can end up with 10 VM failed because of bad hardware and no backups at all.

Reply
0 Kudos
pkuzma82
Contributor
Contributor

This is definitely a very large price increase and will reduce overall density.  If I can only have 64GB of ram per CPU according to enterprise licensing then I am able to lower costs on hardware to compensate, but this strategy seems will reduce density and this will stunt hardware evolution.  Why would anyone purchase a 32GB stick of memory?  V4 I would, V5 no need.  

Reply
0 Kudos
rapier55
Contributor
Contributor

I will be out of the office until February 12th.

Reply
0 Kudos
JAndrews42
Enthusiast
Enthusiast

>  If I can only have 64GB of ram per CPU according to enterprise licensing then I am able to lower costs on hardware to compensate, but this strategy seems will reduce density and this will stunt hardware evolution.  Why would anyone purchase a 32GB stick of memory?  V4 I would, V5 no need. 

The RAM limit is based on licensing and allocated memory, not physical hardware.  While you do need one license for each physical CPU in your server, physical memory is not licensed.  Assuming Enterprise licensing, a 2-CPU server with the minimum of 2 licenses can have up to 2TB of physical RAM and run VMs with up to 1056GB of allocated vRAM*.

If you want to run VMs with more vRAM, simply purchase additional licenses

*note this assumes you have a monster VM with 1TB of RAM, plus additional VMs with a total of 32GB allocated.  The monster VM will only count 96GB towards the cap giving you 32GB of additional vRAM to allocate.

Reply
0 Kudos
pkuzma82
Contributor
Contributor

So a single vm can use a maximum of 96gb of vram licensing even if it is at 1tb of ram? That is better. Although it will not help my situation. I have many hosts with 144gb of ram and 2 cpus that are used for view workstations. My initial thought was that this would prevent me from assigning the physical ram over 128gb to vms unless I purchased a third license. This is a fairly steep increase in price so I would look at lowering the cost of the hardware to match the licensing, but purchase more of them. This seems like the wrong direction because we have been consolidating and lowering energy costs, cooling, etc... but with the new pricing model this is the only way to achieve the same goal as v4 with roughly the same budget.

Sent from my Verizon Wireless Phone

Reply
0 Kudos
sakibpavel
Enthusiast
Enthusiast

While 192 GB vRAM is far less than 512 GB real RAM in 5.0 vs 4.1 Essentials (Plus) is still a viable product for small companies.
Technically, you were limited in theory to 768GB of RAM with ESS/ESS+ and could oversubscribe from there, up to 1.5TB total (supported).
So ESS+ degredation was 1.5TB -> 192GB
Sakibpavel
Reply
0 Kudos
Rumple
Virtuoso
Virtuoso

Quote - I have many hosts with 144gb of ram and 2 cpus that are used for view workstations

Hosts used for Workstations fall under the Desktop licensing I believe…not standard esx licensing and that’s a different licensing model (I think its unlimited)

You may want to check into that.

Reply
0 Kudos
JAndrews42
Enthusiast
Enthusiast

>Hosts used for Workstations fall under the Desktop licensing I believe…not standard esx licensing and that’s a different licensing model (I think its unlimited)

Only if you have the View Bundle or the "vSphere Desktop" licensing - both of which are Enterprise Plus with unlimited vRAM, but you cannot run anything that doesn't support the VDI environment on the hosts.

Reply
0 Kudos
pkuzma82
Contributor
Contributor

Does that vram count include esx overhead? .. 144gb of ram minus about 10% for overhead only leaves less than 2gb that would not be allocated. It may work out. Im going to check on the desktop licensing.

Sent from my Verizon Wireless Phone

Reply
0 Kudos