Today was a big day for VMware. I’m going to provide some summary coverage of the products launched today, but only a quick recap; I’ll have more in-depth analysis and information on the products and their key features and improvements in future blog posts. No doubt there is going to be plenty of other coverage on the launch as well, and I’ll likely produce a special “Short Takes” episode with a summary of some of the related links, so look for that as well.
Now, on to the product announcements!
vSphere 5
As fully expected, VMware today announced VMware vSphere 5, the next generation of their virtualization suite. VMware continues to drive virtualization “higher” in the data center as they target even the most mission critical applications, so vSphere 5 offers support for massive VMs (up to 32 vCPUs and 1 TB of RAM per VM). With vSphere 4, there were only a few instances where a mission critical application couldn’t be supported because of resource constraints. That already-slim window shrinks even more now with vSphere 5.
Also in the vSphere 5 release, VMware has added a lot of features to help simplify and automate the virtualization layer. This is fully expected and a natural part of vSphere’s continued maturation. Some of the features that VMware packed into this release for improved administration and management include:
- vSphere Auto Deploy: VMware now offers a fully supported PXE boot solution that offers completely stateless ESXi hosts. Need to deploy a new ESXi host? No problem, with Auto Deploy it can be done in minutes. Need to deploy a new ESXi image? Change a few rules in the Auto Deploy engine and reboot your host—and you’re done. It’s pretty powerful stuff, in my opinion.
- Storage DRS: vSphere DRS is the darling of many data centers, transparently moving VMs around to keep cluster workloads balanced. vSphere 5 introduces the same concept for storage, called Storage DRS. (Just as a side note, I’m not clear if the “DRS” in “Storage DRS” still stands for “Distributed Resource Scheduler,” since that’s not really applicable to storage. Anyone know?) Using information on storage capacity usage and (optionally) I/O response times, Storage DRS can shift virtual disks for VMs from datastore to datastore—using the concept of a datastore cluster—to keep storage utilization balanced. Like vSphere DRS, Storage DRS also performs initial placement, simplifying the VM storage provisioning process. This is something that has been in the works for years (I first heard about it from VMware in 2008), and it’s great to see it finally make it’s appearance.
- Profile-Driven Storage: This is another killer feature. Building on the vSphere Storage APIs for Storage Awareness (more popularly known as VASA), profile-driven storage allows administrators to define VM storage profiles that describe the features or attributes that storage must possess in order to satisfy the requirements of the VM (RAID type, disk type, capacity, protection level, replication, snapshots, etc.). Then, based on VM storage profiles, when you create a new VM, perform a Storage vMotion, or clone a VM or template, vSphere will use the VM storage profile to show you which datastores are compatible (compliant with the profile) or incompatible (noncompliant with the profile). This provides a huge benefit to the vSphere administrator in ensuring that VMs are stored on the right storage with the right support.
Of course, that’s not all that vSphere 5 has to offer; there’s also a laundry list of other new features:
- VAAI v2, which includes hardware offloads for NFS and new thin provisioning awareness
- All-new framework for vSphere HA, which eliminates the primary/secondary model and provides significant new features
- A new vSphere Storage Appliance, to turn local (DAS) storage into shared storage in environments where a dedicated SAN isn’t possible and performance is not the key consideration
- A new version of VMFS that offers datastores up to 64TB in size without the use of extents
- Significant performance enhancements for Storage vMotion, and the ability to relocate snapshots
- Improvements in NFS to support scale-out NAS
- Software FCoE initiator (only supported on Intel X520 NICs at initial release). I have a couple of the Intel X520 NICs that I’ll be doing some additional testing with against vSphere 5, so look for those results on this site soon.
As you can see, it’s quite a significant release. But wait, there’s more…
vCloud Director 1.5
VMware also announced vCloud Director 1.5, which offers a number of new features:
- New APIs: vCloud Director will offer broadened “southbound” APIs, so that solutions like vCO, UIM, and others can provide further automation in highly virtualized environments
- Linked Clone support: vCD 1.5 will support linked clones, so that deploying new workloads will happen faster and with less storage consumption.
Site Recovery Manager 5
In addition to vSphere 5 and vCD 1.5, VMware also unveiled SRM 5, with new features like:
- Built-in automated failback: While vendors such as EMC provided failback plugins, those plugins didn’t provide the full functionality of SRM when performing a failback. SRM 5 now provides full failback support, a key feature that many organizations have been requesting.
- Workload mobility workflows: SRM starts adding support for workload mobility workflows, to move workloads cold between sites. (Hmmm…think about VPLEX plus SRM workload mobility workflows…give you any ideas?)
- vSphere host-based replication: vSphere 5 can now offer replication on a host-based level, for environments where array-based replication isn’t possible due to constraints (budgetary or otherwise). Naturally, there’s a trade-off in using host-based replication versus array-based replication, but it’s a nice feature to add for lower-end customers.
vShield 5
Last, but not least, VMware announced vShield 5. vShield 5 brings improved management to the table as well as some new features:
- Static routing functionality in vShield Edge: This provides vSphere and vCloud Director administrators greater flexibility in modeling network topologies.
- New product offering in the form of vShield Data Security: This is an integration of technology from RSA DLP (Data Loss Prevention), that offers administrators the ability to discover and report sensitive data in virtual machines.
All in all, VMware unveiled a lot of new functionality today that is targeted at driving the further adoption of virtualization and addressing concerns over virtualizing mission critical applications.
As I mentioned earlier, look for more in-depth articles on some of the new features and functionality in the coming days and weeks. Thanks!
Tags: vCloud, Virtualization, VMware, vSphere
-
I’m afraid the entire event was eclipsed by the horrible new licensing model. It’s all anyone is really talking about since probably most of the entire customer base will now see a doubling of costs for their current environments. The only buzz I see is how do I move to Hyper-V or Xen3
-
It is great that vSphere 5 can support a guest with 32 vCPU and 1TB of memory but …. If I understand vRAM licensing you would be required to purchase 22 sockets of Enterprise plus to have a guest with 1TB of vRAM.
Here is a summary of how VMware’s new licensing model works:
vSphere 5 grants you 48 GB of vRAM per physical processor with Enterprise Plus (24 GB per processor with Standard/ 32 GB per processor with Enterprise) – this provides 96 GB of vRAM for each two-socket host and 192 GB for each four-socket host in an aggregate pool. The pool spans all linked vCenter servers.
“At any given point in time, the vRAM capacity consumed by all powered-on virtual machines within a pool must be equal or lower than the pooled vRAM capacity.”
This really changes the ROI model for vSphere, its architecture and where it fits into the enterprise.
Consider the economics of consolidating onto 2 socket servers with 256GB in vSphere 4 vs vSphere 5. vSphere 4 requires 2 sockets of enterprise plus where vSphere 5 now requires 6 “sockets” of enterprise plus for the same server to accomodate the 256GB of vRAM.
I realize that the need for n+1 and other factors mean that you don’t have to license 100% of physical memory but many of the mission critical applications that are now candidates for virtualization technically because of what vSphere 5 brings to the table now face a significant economic challenge because of the vRAM requirements. Overall, the enhancements in vSphere 5 are a huge step forward but the new vRAM licensing model will not help me accelerate the virtualization of mission critical applications with high memory requirements.
-
Gchapman, jhep,
I hear you guys, but I’m not going to comment until I’ve personally had the chance to run the numbers and see what things look like. I want to be able to provide a fair and balanced review of the new licensing model, even if that review isn’t good for VMware. Give me some time to go through the numbers.
-
I had held off my vmware to Hyper V migration, thinking VMware would actually lower the cost of its product so I could justify it. Oh well.
Now its on like Donky Kong…..Hyper V FTW!!!
-
I initially had the WTF? response to the new licensing model, especially since we just completed a huge ELA last week with VMware for a lot of Ent+ licenses. Now we find that our licenses don’t cover unlimited memory but are restricted to the total amount of ALLOCATED memory, not CONSUMED memory.
VMware considers this “fair” saying that you’re charging users for what they use. In fact, you’re charging users for the possibility of having the memory, not actually what they’re using. Over-provisioning could now have a financial penalty on your environment. It’s like paying interest on a line of credit before you withdraw a dime.
That said, I had a quick look at our environment and at this time I believe that we are not currently adversely impacted by the new model. That doesn’t mean we won’t be as we grow but right now I don’t think we exceed the vRAM limits on the licenses.
I think we’re getting lucky because we have 2-socket remote site servers with as little as 16GB or 32GB of memory. With Standard licenses, those now allow for a 48GB vRAM total (2x24GB) and in effect they’ll be subsidizing our larger datacenters that are in the same Virtual Center environment.
As memory requirements increase with new releases of operating systems and applications, we’re going to be in the same fight as we are today with number of cores. We might not be able to install, say, Windows Server 2012 because we cant afford the ESXi licensing and that just doesn’t make any sense. VMware will be forced to change the licensing model either by coming up with something new (a true chargeback model) or by simply increasing the licenses from something as tiny as 48GB per license to 64GB or larger.
Overall, I hate the new model but I don’t think it’s going to drastically hurt us in the short term. Longer term, my thoughts of deploying larger SQL servers just got shot in the head. A 256GB guest will now cost over 5 licenses at a list price of $3,495 per server – i.e. over $20K – whether it ever uses that memory or not.
-
Your next book can be Mastering vSphere 5 Licensing.
More seriously Citrix and Microsoft have to be pretty happy at this own-goal by VMWare. -
I wrote a blog about the licensing changes, and ran a few numbers for a VDI scenario. Not pretty! Depending on the edition you license, v5.0 could cost you 25% to 240% more per VDI VM.
http://derek858.blogspot.com/2011/07/vsphere-50-licensing-changes.html
-
Larry, don’t migrate just because of licensing—make an informed decision based on all the factors.
Ed, what you’re doing is what everyone needs to do: take it calmly and evaluate your own environment.
Andy, based on the feedback I’m getting I’m not so sure I want to tackle that!
Derek, thanks for the link—I’ll give it a look!
-
Does anybody know if (and how) VSPP licensing will change with VSphere 5?
-
Norberto, I believe that the new enterprise licensing model is closer to VSPP than the previous version. I don’t know how much, if any, VSPP changes with vSphere 5.
-
One of my customers is actually using VMware enterprise for their QA department. Each VM has 16GB of RAM and averages ~95% utilization of that RAM. They have ~1000 of these.
Before they were paying ~700k for 248 sockets of Enterprise. Now they’ll need to cover a difference of ~700k to upgrade to 5. This is a customer that was already on the fence about VMware. In this instance I’d probably gun for an alternative solution since they’re really more about consolidation than the other features out there.
A LOT of people will eventually use the cheaper/free solution because its “good enough”; not necessarily because its better.
The idea that VMware is proud to announce I can virtualize ANYTHING but then couples it with a licensing model that encourages me not to do so is a little absurd. If we look at an 8 way DB server with 256GB of RAM thats going to cost ~17k with Enterprise and ~19k with Enterprise Plus.
I suppose I COULD come out ahead assuming I have as much RAM per socket as my license supports (i.e. a 2 socket blade should have 96GB of RAM with Enterprise plus.) In that instance with HA you will see a tremendous savings (your failover blades will actually be essentially free.)
In my own environment I’d probably shift all my low hanging fruit over to Xen or Hyper-V and push on both VMware and my alternate vendor to give me better pricing and better features.
-
Scott,
We were already on the fence. We have 24 hosts in total, all Dell R610′s, with dual 6 core/96gigs of RAM, spread out amongst 2 sites. Today we have Enterprise licenses. (not Enterprise +).
These hosts virtualize Windows Servers/VDI only, for which we purchase Datacenter licenses where we can, because it saves money on Windows licensesing.
We were already on the fence about VMware because we essentially own everything for Hyper V minus the SCVMM piece which is cheap. Add to that for what we use feature wise on VMware, Hyper V R2 is “good enough” for us.
Now I must move to Enterprise + just to use the RAM we already own if we renew in Feb of next year.
I honestly thought VMware was going to more competitive on the pricing since I know/here of some movement towards Hyper V R2 mainly because of pricing.
This change will only speed it up.
-
I was designing a greenfield virtualization platform for a rapidly growing company safe in the knowledge that I can discount all but vmware from the running… Now however I need to take a broader approach and seriously look at the alternatives, if they provide 90% of the needed features for the company then I’m going to recommend them instead. We do not have cash to throw around like confetti especially when in increase in cost is so sevre from one version to the next.
I’m a lowly VCP3 and have always been happy with all the vmware products but vmware have just given MS & Citrix the break they need to start growing their installer base.Sorry guys massive massive own goal here!
-
Great News ! I believe that the support offered by the new vSphere 5 for massive VMs (ie. 32 vCPUs and 1 TB of RAM per VM will explore positive aspects of a cluster. Since, the vSphere 5 includes a new set of rich features to automate the virtualization layers, it is going to be a boom in the virtualization market. Thanks for keeping us updated about the latest technologies.
-
Thanks Scott, but honestly we don’t care. I don’t see a single feature that’s worth a 300%+ increase in licensing. I say this based on the fact that we reviewed the new VSphere Licensing documentation and did the calculations. Frankly, we’re disgusted. Disgusted because VMware claims this new licensing to be more fair. Disgusted because they claim this licensing to be more simple. Disgusted because this as an incremental release with the biggest “screw you” to customers ever.
Just as bad have been the VMware websites (shills) that have been defending this extortion as fair. Now we know which sites are getting payola from VMware. Looks like quite a few.
-
Rob A, I don’t know anything about your environment, so I can’t comment on how vSphere 5′s licensing will affect you. I’m sorry to hear that it looks like there will be a substantial increase in licensing costs. Just for my own benefit–and please understand I am neither agreeing or disagreeing with VMware’s licensing changes at this point–would you mind walking me through your calculations? I’m curious to know why there will be such a substantial increase.
-
I think a big part of charging on allocation is that it makes overcommit a lesser value for the customer. The value only gets worse as your environment gets larger.
-
The way I see the licensing at this point isn’t as big an issue as I feel people are making it out to be. Say you have a host with 4 Sockets and 128G of RAM. With Enterprise Plus licensing you are entitled to 192GB of vRAM, a 50% over-commitment of the Physical RAM which is great. I doubt anyone tends to utilize 100% of the Physical RAM on the host which means even more room on the overcommit side (i.e. say you utilize 96GB of Physical you are still allowed 192GB committed.) With that calculation, if you had 96GB of Physical RAM utilized and maxed out the allowed commit level of 192GB than all your VM’s would only be at 50% utilization, which I believe isn’t the true point of virtualization.
Now let’s look at this with a cluster of 6 of these hosts. You now have 24 Enterprise Plus Licenses with the ability to run 1152GB of RAM in the vRAM pool. I am sure I am not the only one who sets a threshold (mine is 85% utilization) so with the 768GB of actual RAM to use you end up using a maximum of 652.8GB of RAM in this cluster. Keep in mind the vRAM pool is everything in a single vCenter Server, so if you have 3 clusters of 6 hosts in one vCenter your available vRAM pool contains 3456GB RAM that can be allocated anywhere within that vCenter server not just the host.
All this information fits with exactly what you were paying for Enterprise Plus to begin with, so the costs don’t go up unless you are really sloppy with your resources. What I don’t like is that when I look at this scenario I see that in theory you could drop down to just Enterprise licensing to save some money, but you lose features that you may use, and I don’t like that idea, so here you are back at the same battle you had before where you are licensing for features not resources only now you see a ceiling that you most likely won’t hit.
So what I think is that it is likely we won’t see anyone trying to build a VM to the Maximums of vSphere 5, like virtualizing an Oracle rack or several large clustered SQL servers, but I still think that for the most part that once people sit down and look at their individual environments that the licensing storm that is brewing will subside and we can focus on the other 35 minutes of the presentation from July 12th.
Now let’s look at this with a cluster of 6 hosts. I now have 24 Enterprise Plus Licenses with the ability to run 1152GB of RAM in the pool. Again I only have 768GB of actual RAM to use and like to have a little bit of buffer (about 15%) which means I only intend on using a maximum of 652.8GB of RAM in this cluster.
All this information fits with exactly what I was paying to begin with, so my costs don’t go up unless I am really sloppy with my resources. What I don’t like is that when I look at this I see that in theory I could drop down to just Enterprise licensing to save some money, but I lose features that I use, and I don’t like that idea, so here I am back at the same battle I had before where I am licensing for features not resources only now I see a ceiling that I most likely wont hit.
-
If you’re building a net-new environment and you’re not using dense blades and you’re using enterprise plus then yes; it’s great.
If I run through VMware’s basic ROI comparison against Hyper-V now; Hyper-V goes from losing by 15-20% to now winning by 2%. Where is the value?
-
Touch of good news though folks, and I heard it from the VMware SE’s mouth @ the COVMUG this week…
“…we aren’t going to prevent you from starting up or running your VM’s, that would be stupid. We are just going to nag you about the memory/licensing issue…”
So if you can handle the “acknowledge” or “ok” button popping up every 10 seconds or so, this could really be a non-issue!
Wonder if they are opening up that API for someone to be able to “press the button” programmatically?
Site Sponsors
20 comments
Comments feed for this article
Trackback link: http://blog.scottlowe.org/2011/07/12/vmwares-launch-event/trackback/