No More vRAM Licensing?

I was just about to post. I'm completely frustrated. I just bought licenses a month ago and now this. If it turns out I overbought I'm not going to be happy which most likely I would have if they flip it back right?
 
Excellent news.

I wonder if they plan to compensate customers who had to buy additional licenses because of the vRAM limits?
 
The market has spoken.
I talked to our VAR yesterday and they have been selling a whole bunch of hardware to the SMB market segment where the buyers were specifically requesting Hyper-V rather than VMware.

VMware has more features for sure, but it's sort of like Photoshop, why buy that when all you need is to crop your holiday images and resize them for web use. Many SMB shops are perfectly fine with Hyper-V and that's where the grow is now imho.

It would be interesting to hear from other folks who work for VARs, but my guess is that the enterprise market is saturated and while the vRAM tax doesn't really affect SMB it created a psychological influence that was meh from a marketing perspective. Combine that with VMware's arrogance when it comes to pricing in general and I can see why a bunch of SMB shops choose a different product (be it Hyper-V, Xen, or KVM).
 
The vRAM risk has always been a future risk..but the future is coming very quickly. That's the problem. I've had VERY few customers actually hit their vRAM number but many were planning to hit it in a year or two and it was causing uncomfortable conversations. It also put pressure in other places, like server vendors. Every conversation around UCS I've had where we talk max RAM in a blade has been punctuated with a jab at vRAM by customers.

If true that it's going away that's GREAT. I just wish they had never done it. It was an unnecessary cultivation of bad will.
 
I'm also interested to see how vcops will be integrated into the various licensing editions.
 
W00t. Great news. Although I never ever reached the potential limit of any of the bundels I sold. 192 gb of ram is absurd amount.
 
I'm still a little worried... I'm more okay with the vRAM licensing then I would be of a per VM licensing model. But I also say this because like most, I wasn't hitting my vRAM limit right now but would of in another year or two.

The per VM model seems to be the direction they are going with everything else, I'm sure its only a matter of time before they move to that model with vSphere.
 
I just upgraded our biggest client's hosts by 512GB each and was really not looking forward to the ram licensing conversation with the director which would have definitely happened by early next year.

They should have never done it.
 
Last edited:
I'm still a little worried... I'm more okay with the vRAM licensing then I would be of a per VM licensing model. But I also say this because like most, I wasn't hitting my vRAM limit right now but would of in another year or two.

The per VM model seems to be the direction they are going with everything else, I'm sure its only a matter of time before they move to that model with vSphere.

If they go per-vm that would be the worst of both worlds. I would expect them to lay off of VMware View as they somewhat did with the vRAM tax. I'm pushing upwards of 1,000 users in View now which from a cost perspective is challenging enough already to find savings. Yes, there are other advantages, but security and manageability isn't the easiest thing to sell when you're takling money with an executive.
 
Awesome. hated the vram licensing. Much more simplified again.
 
Great news. Even if it didn't affect a company now, the odds are they would be affected by it soon which caused hesitation to go with VMware.
 
So are they going to license by CPU? How does that work, by socket or core?
 
VMware has always done socket, not core. I assume this will be the same.

Well that is only partly true, right? Wasn't 4.x licensed by socket with max core counts depending on the version you paid for? With 5.x they added the vRam entitlements and unrestricted the core counts... if I'm remembering correctly.
 
VMware has always done socket, not core. I assume this will be the same.

This kills me, I had had a customer today argue with me that he originally purchased core licensing..lol.
 
Well that is only partly true, right? Wasn't 4.x licensed by socket with max core counts depending on the version you paid for? With 5.x they added the vRam entitlements and unrestricted the core counts... if I'm remembering correctly.

Yeah..that's true. But I don't think we really had 8 and 10 core CPUs then so no one was really hitting it. By the time they did 5 came out and dropped it.
 
This kills me, I had had a customer today argue with me that he originally purchased core licensing..lol.

There was, but the limits were bonkers for the tech at the time so no one ever hit it - it was really a socket license :)

Unless you had unisys systems.
 
Yeah..I realize that..but they didn't buy core licenses...they bought socket..but he insisted and I dropped it since we upgraded to vSphere 5 even though his licenses on MYVMWARE site showed 8 CPU sockets.
 
There was, but the limits were bonkers for the tech at the time so no one ever hit it - it was really a socket license :)

Unless you had unisys systems.

I disagree. Cisco, HP, IBM, and all of them have blades/servers that can fit 512 GB, 1 TB, even 2 TB. With most workloads, the primary constraint preventing me from cramming more VMs into a system is memory. 128 GB versus 256 GB server? I can put twice as many VMs usually, and buy half as many servers (which is more important for power/cooling than server cost often times).

Many sites weren't hit with the limit... yet. But we all knew we were going to hit it. The 32 GB limit was pretty high for the free version, but you can get a decent desktop CPU for a whitebox that can have 32 GB of RAM. Next year? 32 GB as a limit for the free version is going to be a significant constraint (it was initially 8 GB limited, but they relented after the broohaha).
 
He was talking about core limits...not memory limits. There were core limits on the vSphere license levels before the vRAM stuff. When vRAM hit in vSphere 5 those core limits went away. Believe it was 6 cores per socket for Enterprise and 12 for Ent+.
 
He was talking about core limits...not memory limits. There were core limits on the vSphere license levels before the vRAM stuff. When vRAM hit in vSphere 5 those core limits went away. Believe it was 6 cores per socket for Enterprise and 12 for Ent+.

Oops, yeah you're right, it was 6 cores until you went to E+.
 
For those of you running free versions -- What does esxi do that xensever (free version) or xcp (oss/free versoin) or kvm don't do? None of those have anywhere near 32G memory limits IIRC. The memory limits just seem silly to me. I'm asking honestly as I haven't used esxi. The last free vmware server version I used was vmware server 1.0 IIRC.
 
For those of you running free versions -- What does esxi do that xensever (free version) or xcp (oss/free versoin) or kvm don't do? None of those have anywhere near 32G memory limits IIRC. The memory limits just seem silly to me. I'm asking honestly as I haven't used esxi. The last free vmware server version I used was vmware server 1.0 IIRC.

Install it and find out.
 
I will be annoyed if they raise the limits on the free version substantially. I just bought essentials for my home system last month to use my 48 GB (with plans to go higher). If 64 is suddenly the new 32 I'll have spent the $560 for nothing (today, anyway).

Viper GTS
 
Secrets don't last too long before they hit twitter and the blogosphere. It's sort of sad actually. The minute the email was sent out I saw a few people immediately jump to twitter saying that they know some super secret NDA information, then it leaks to a blog a few days later.... big surprise.

Either way, I'm glad they reverted back. I didn't have a problem with the vRAM consumption-pricing model but they would have had to done completely away with the license per-socket and went strictly on vRAM.
 
I will be annoyed if they raise the limits on the free version substantially. I just bought essentials for my home system last month to use my 48 GB (with plans to go higher). If 64 is suddenly the new 32 I'll have spent the $560 for nothing (today, anyway).

Viper GTS
Yes but you at least get storage api license which you don't in free. So you can use backups.
 
I will be annoyed if they raise the limits on the free version substantially. I just bought essentials for my home system last month to use my 48 GB (with plans to go higher). If 64 is suddenly the new 32 I'll have spent the $560 for nothing (today, anyway).


Nope. According to this it's still 32GB.
 
Secrets don't last too long before they hit twitter and the blogosphere. It's sort of sad actually. The minute the email was sent out I saw a few people immediately jump to twitter saying that they know some super secret NDA information, then it leaks to a blog a few days later.... big surprise.

Either way, I'm glad they reverted back. I didn't have a problem with the vRAM consumption-pricing model but they would have had to done completely away with the license per-socket and went strictly on vRAM.

I was amazed this stayed a secret as long as it did. I knew about it almost a month before the CRN article and that was the first real leak.

We told them at PTAB that the only way for these new vCloud Bundles to work was to get rid of vRAM. You can't ask a customer to pay that much (another whole vCloud Bundle socket license) just to add some vRAM. It would never fly and it was causing people to stay on vSphere 4.1 as they looked at other options.

It's a smart move.
 
I was amazed this stayed a secret as long as it did. I knew about it almost a month before the CRN article and that was the first real leak.

We told them at PTAB that the only way for these new vCloud Bundles to work was to get rid of vRAM. You can't ask a customer to pay that much (another whole vCloud Bundle socket license) just to add some vRAM. It would never fly and it was causing people to stay on vSphere 4.1 as they looked at other options.

It's a smart move.

Granted it was nearly a month, but hours after Carl's email in July went out I saw a few "ohhhh I know a secret" tweets. Although, nothing gravely damaging to the company if it leaked completely, but leaking of information has been a huge issue recently. It's just disappointing to see stuff like that.

It was a bad idea from the start and they did fight to hold on to it, glad they finally did away with it though. At least it shows that they listen to the customers.
 
Eh. We get excited about this stuff. Is hard not to make some comments about [REDACTED]. :)
 
For those of you running free versions -- What does esxi do that xensever (free version) or xcp (oss/free versoin) or kvm don't do? None of those have anywhere near 32G memory limits IIRC. The memory limits just seem silly to me. I'm asking honestly as I haven't used esxi. The last free vmware server version I used was vmware server 1.0 IIRC.

It's a convenience thing and a market share thing.

Right now at least, VMware the largest hypervisor footprint by a long shot. After vTax, Hyper-V and KVM/RHEV have made some inroads, but virtually (get it?) every data center that does x86 virtualization does ESXi/vSphere.

So on a pure percentage basis, VMware basis is the skill one probably has, so it's more familiar.

Also, virtually (getting old) every virtual machine that's offered up as a pre-made VM image is released on OVA or some other VMware. Sure, there are tools to convert, but it's just stupid easy and quick to load up an OVA.

The fact that ESXi is a very small contained hypervisor is also nice. Standing it up is simple (usually on a USB flash drive) and you can be up and running in 15 minutes.

And 32G for most of us isn't too much of a limit for the free version.

Technically speaking, I can't think of anything off the top of my head that the free version can do that the other hypervisor can't. It's more a matter of familiarity and convenience. Which as much as we don't like to admit it in the tech world, are important factors.

Now, 32 GB is fine for most right now, I can get cheap whitebox motherboards that can take 32 GB now. A year from now? Probably not, so I'd reconsider ESXi unless they up'd it some more.
 
It's a convenience thing and a market share thing.

Right now at least, VMware the largest hypervisor footprint by a long shot. After vTax, Hyper-V and KVM/RHEV have made some inroads, but virtually (get it?) every data center that does x86 virtualization does ESXi/vSphere.

So on a pure percentage basis, VMware basis is the skill one probably has, so it's more familiar.

Also, virtually (getting old) every virtual machine that's offered up as a pre-made VM image is released on OVA or some other VMware. Sure, there are tools to convert, but it's just stupid easy and quick to load up an OVA.

The fact that ESXi is a very small contained hypervisor is also nice. Standing it up is simple (usually on a USB flash drive) and you can be up and running in 15 minutes.

And 32G for most of us isn't too much of a limit for the free version.

Technically speaking, I can't think of anything off the top of my head that the free version can do that the other hypervisor can't. It's more a matter of familiarity and convenience. Which as much as we don't like to admit it in the tech world, are important factors.

Now, 32 GB is fine for most right now, I can get cheap whitebox motherboards that can take 32 GB now. A year from now? Probably not, so I'd reconsider ESXi unless they up'd it some more.

Thanks for a real answer. I appreciate it. I use KVM now but am always willing to try something new if I know it's going to buy me something neat/cool.....

I've thought of trying XCP but im not sure it would buy me much over the functionality I already have. KVM runs on linux which I have 5 years of experience and runs on ubuntu/redhat/etc and seems to work fairly well for my usage pattern. I'm already familiar with linux, though, so I guess that "like what is familiar" comes into play.
 
The other thing about it is what some of the labs guys in R&D are doing.

The keynote talked about grid computing, big data, hadoop, and HPC. We're actually running beowulf and large-scale hadoop clustering with SRP/RDMA workloads, and able to get almost exactly bare-metal performance with them, while maintaining some level of consolidation, single pane management, etc.

Think about being able to build a massive super-cluster where you don't have to worry about hardware choices... it's amazing shit. No one else can do that yet.
 
Back
Top