virtualboy blog

Matt McSpirit on Virtualisation, Management and Core Infrastructure

Responding to the VMware: Virtual Reality blog post…

Responding to the VMware: Virtual Reality blog post…

  • Comments 25
  • Likes

A couple of days back, I wrote a war-and-peace-length (OK, not quite!) post on the investigation of the VMware Cost-Per-Application Calculator.  If you haven’t read it, you can read it here.  To summarise, it showed:

  • On like for like hardware, a Microsoft Virtualisation solution encompassing Hyper-V and System Center, plus Windows licenses, would have a 3% cost-per-application over a vSphere Enterprise Plus and Windows Licenses combination.  Here’s a few caveats on that summary though:
    • 3% is very different to 20-30% that VMware tout as their cost-per-app saving
    • Hyper-V required 9 hosts, versus 6 on VMware, therefore Microsoft required 3 hosts worth of infrastructure costs, and software costs on top.
    • The whole argument relies on Memory Oversubscription, which can be countered by adding more RAM to the physical servers.
    • On actual hardware, from Dell’s website, the difference in cost was actually less than 1%.
    • However, when you doubled the RAM in the Hyper-V hosts, to run more VMs, for just $900 per host, we actually dropped to 5 hosts, thus bringing a saving of 31% in a cost-per-application model, in Microsoft’s favour.

So, to today.  I read this: http://blogs.vmware.com/virtualreality/2009/09/did-microsoft-just-agree-with-us-that-hyper-v-is-not-16th-the-cost-of-vsphere.html

The answer, plain and simple is, no, Microsoft, nor I, did not just agree with you that Hyper-V is not 1/6th of the cost of vSphere.  What I proved in this post, is, when comparing a Microsoft and System Center solution, with Windows licenses, with a vSphere Enterprise Plus with Windows Licenses solution, providing you provide more RAM to compensate for the use of memory oversubscription (which not everyone, believe it or not, uses), Microsoft’s Cost-per-App is 31% cheaper than on vSphere.

Where Microsoft quote 1/3rd of the cost, 1/5th of the cost, or even 1/6th of the cost, it isn’t about Cost-per-Application.  Cost-per-Application is a new method, conceived by VMware, to try to get away from the fact that the upfront licensing costs, of which the 1/3rd, 1/5th, or 1/6th refer to, is in fact, on a number of occasions, true.

Let’s have a look shall we, forgetting about the cost-per-application side of things, which I've shown is 31% cheaper on the Microsoft side, when, assuming common sense, you’d put more RAM instead of buying more hosts, and instead focus on the upfront software costs of a solution.

I’ll base this on an example I received through on email a while back, which was a quote for 15 Servers worth of Virtualisation and Management.  Each of these servers had 4 Quad Cores.  So, off I went to VMware’s website, added it to my basket – 60 CPU’s worth of vSphere Enterprise  (not Plus) and a vCenter Server, both with 2 year Platinum S&S.  What was the total software costs, regardless of infrastructure, hardware, power, cooling etc?  $260,058.54.

Now let’s look at the Microsoft side – well, 60 CPUs of Hyper-V Server 2008 R2, is $0.  Good start.  60 CPU’s worth of Server Management Suite Datacenter (unlimited agents for SCOM, SCCM, SCDPM, SCVMM, and the SCVMM Server bits too) come in at $744 per CPU, and include 2 years SA.  Add on to that, a Server license for SCCM SVR, SCOM SVR and SCDPM SVR, and you’re looking at about $2605 total including 2 years SA.  Grand total?  $47,425.

Both of these calculations exclude SQL, and Windows Licenses.

Now, let’s focus on the 1/3rd, 1/5th and the 1/6th cost differences that the Virtual reality post so kindly points out.

$260,058.54 / $47,425 = 5.483574907749077, which we’ll call 5.48 for simplicity.

So, to summarise with this simple example, it’s clear to see that in fact, a Hyper-V and System Center solution is between 1/5th and 1/6th of the VMware costs, and that was versus Enterprise.  If you use Enterprise Plus, the total for the VMware software is $314,556.54 with 2 year Platinum S&S, which means that the the cost of Hyper-V and System Center is actually between 1/6th and 1/7th (6.63 times) of vSphere.

If I halve the number of CPUs, down to 30, we’re now at $160,937.34 for vSphere Enterprise Plus with vCenter and 2 years Platinum S&S, versus $24,925 on the Microsoft side, which, again, means that a Microsoft solution comes in at between 1/6th and 1/7th of the cost of the vSphere solution.

So, to wrap up – hopefully readers of this blog will share my frustration with the Virtual Reality blog post, and how they have taken a post that had a clear objective of looking at the cost-per-application calculator, and instead, turned it round to focus on something it clearly wasn’t aiming to address.

The final part of the Virtual Reality post reads as follows:

“The bottom line is that Microsoft’s blog doesn’t uncover anything new about the VMware Cost Per Application Calculator. Quite the opposite, it confirms it. Try our calculator for yourself and create a customized report. You will find that it includes a sensitivity analysis showing vSphere’s cost per application at different consolidation ratios. The analysis clearly demonstrates that even at equal consolidation ratios (worst case scenario for vSphere), Hyper-V’s total acquisition cost is, at best, only marginally lower. Once you factor in vSphere’s tremendous consolidation ratio advantage over Hyper-V and vSphere’s ability to scale up to 2X more VMs than Hyper-V (check-out the “Evaluating the ESX 4 Hypervisor and VM Density Advantage” report), vSphere delivers the lowest cost per application by up to 20-30%. In fact, often vSphere becomes a less expensive solution than Hyper-V with just 1-2 more VM’s per ESX host – in addition to being a much more functional, more scalable, more proven product.

So you can either believe us when we say that Microsoft Hyper-V is actually about the same cost as VMware products or you can believe Microsoft when they say that VMware solutions cost about as much as Hyper-V – take your pick!”

My blog post does not actually confirm the cost-per-application calculator – did you stop reading?  It confirms that, with an ability to spec hardware effectively, you can save 31% on a cost-per-application calculation, using Hyper-V and System Center.  This is very different from the 20-30% cheaper in a cost-per-app calculation that VMware claims they have.

Sensitivity Analysis for different consolidation ratios?  The analysis clearly demonstrates that even at equal consolidation ratios (worst case scenario for vSphere), Hyper-V’s total acquisition cost is, at best, only marginally lower?  Let’s have a look at this table shall we?

vSphere Cost-Per-App Table

At equal consolidation ratios, we’re only marginally lower?  Come again?  $2366 or $3257? That says to me that a Microsoft solution is 27% cheaper, cost-per-application than vSphere – don’t tell me you were actually comparing Hyper-V and System Center to vSphere Standard Edition…?

“Tremendous consolidation ratio advantage over Hyper-V and vSphere’s ability to scale up to 2X more VMs than Hyper-V” – or, you could buy more RAM?  Yes, that’s a simplistic way of looking it it, but you’ve wound me up on this, so I tend to make things clear then!

“vSphere delivers the lowest cost per application by up to 20-30%” – Right, OK.  Assuming you don’t upgrade the RAM, as detailed in the previous post.

“Believe Microsoft when they say that VMware solutions cost about as much as Hyper-V” – I’m not going to rise to this one, if you’ve read this far, and if you’ve read the previous post, if you think that it what I was trying to say, I’d suggest you read them again.  Maybe I should just turn it round and say that VMware vSphere Cost-Per-Application Calculator Sensitivity Analysis proves Microsoft solution is 27% cheaper than a vSphere Enterprise Plus solution!  How about that one?

Then we get on to the comments section of the Virtual Reality post, which starts off promisingly, discussing the deep management offering that the full System Center Suite can provide, to which the response discusses an add-on technology for vSphere, known as AppSpeed.  On the surface, you could compare AppSpeed with SCOM, but then, we’d be in the reverse situation that we are in when we talk about Hypervisors, in the sense that you’re comparing AppSpeed, with a technology that Microsoft acquired in 2000, and has developed ever since.  This is a well adopted, mature monitoring technology, and contrary to the comment in the post, now, in R2, comprehensively monitor heterogeneous environments, including hardware, storage, network I/O and more, across servers, and desktops, and in terms of SCOM only being useful in pure-Microsoft environments, again, check out the cross-platform monitoring, the extensive Management Pack catalog to monitor Microsoft, and non-Microsoft platforms (heard or Veeam, or Bridgeways, for VMware monitoring?) or what about the new announcements from HP, around integration with Insight Control, or IBM’s Integrated PRO Management Pack?

Final thing to add about AppSpeed – it’s on at $1832.18 per CPU with 2 Year Platinum S&S.  The reason that I chose 2 year S&S, is to add-on to our example before:

“(If you use Enterprise Plus, the total for the VMware software is $314,556.54 with 2 year Platinum S&S, which means that the the cost of Hyper-V and System Center is actually between 1/6th and 1/7th (6.63 times) of vSphere.)”

So now we’ll add on 60 x $1832.18 = $109930.80, so add to that, our previous Enterprise Plus total of $314,556.54, and we’re at a total of $424487.34

So, $424,487.34 versus $47,425, which means we’re looking at 1/9th of the cost of vSphere and AppSpeed versus a Microsoft solution.

I’ll leave you with that.



Comments
  • It doesn't matter. VMware is a better, more mature product, especially when a "High Availability" scenario on Hyper-V is limited to 32 VM's per host. That alone deters everyone I talk to away from Hyper-V.

  • Hi Homer,

    Thanks for the comment.  It's actually 64 VMs per clustered Host, or up to 384 per non-clustered host.  64 per clustered host leaves us with 960 VMs per 16 node cluster (assuming 1 node empty for failover - optional)

    Hope that clarifies for you...

    Matt

  • what about network and disk i/o performance? are hyper-v latencies and throughput on par with vSphere near-native speeds? I think some real world application tests would be helpful - eq maximum number of exchange or sharepoint users supported by aforementioned 60 cpu configuration - to give perspective to dollar figures.

  • Hi Martin,

    Great feedback, and you are spot on - I too would like to see those type of tests performed, and those benchmarks published - it would be useful for customers and partners alike.  I think <personal opinion> it would have to be performed by a 3rd Party if we were doing a comparison </personal opinion>, however if you just wanted a pure Hyper-V performance test, this is something that the Exchange/SharePoint team would provide, similar to the Performance Tuning documents that have been produced here: http://technet.microsoft.com/en-us/library/dd277865.aspx (SharePoint as an example) - I would love to see more benchmark tests, like the ones that Scott Drummands has on his blog, but that's for those teams to produce those benchmarks.

    Not a great answer I'm afraid - I want to see them as much as you, trust me! :-)

    Maybe as we get closer to GA on October 22nd!

    Matt

  • I'm sorry, this really just seems like you are very angry and emotionally invested here.  What is all this red text and bold lettering?

    Don't get me wrong - Microsoft is great, without you, i would be out of work :). Obviously Hyper-V is great too, and a step in the right direction.  I just think that as far as usability VMWare's products win, hands down.  Enterprises don't want your new uppity Hypervisor, they want the industry recommended standard. They want VMWare. System Center and Hyper-V are great products, but VMWare is miles ahead.

    Also this post seems like a sales pitch from a fanboy angle who is mad at the VMWare folks.  

    Just sayin'

  • Hi Joe,

    Firstly, thanks for taking the time to comment - I appreciate it.  I apologise if the post sounded too much like a rant - in all honesty, the reason I wrote the first post, on investigating the cost-per-app calculator, wasn't to bad-mouth VMware, and even reading it back, I don't believe it is critical of their technologies, and in most cases, provides the credit it deserves.  The first point of the post was to provide Microsoft Partners, and their customers, with enough of a balanced view to say, 'OK, VMware have technologies that Microsoft don't like Memory Oversubscription, however Microsoft have some technologies VMware don't particularly in the System Center family'  We'll have to take a look at both, evaluate them, and decide from there.  The second point of the post was to say, look, VMware are out there in the marketplace, talking about how they are 20-30% cheaper, cost-per-app, than Microsoft or Citrix.  If I were a customer, and I heard that, I would instantly be interested!!  I'd be mad not to be, but what I showed, I believe in a fair manner, was that when you looked under the covers a little, the numbers didn't always add up.

    The reason this post seemed like a rant, was in pure frustration over the fact that VMware took that original post, and took exactly what they wanted from it, and spun it to their advantage.  Hence the reason for this post, for clarity.

    I agree with you, Microsoft is great, but VMware also have some fantastic technologies, as I acknowledged in the first post.  Being a VCP, I've seen first hand how powerful these technologies are, and with vSphere, VMware have raised the bar again in terms of their capabilities, but you would expect this - they are a Virtualisation company!  However, if I take my Microsoft hat off (it does come off occasionally!), as a neutral, VMware are going to have to work harder now, than say, 3 years ago, to justify the costs associated with their platform.  If a customer needs a specific piece of functionality, that only VMware have, and that will solve a specific problem for that customer, that customer would definately look at the VMware technologies - they would be mad not to!  However, if they need a piece of functionality that VMware, Microsoft and Citrix (as examples), all now offer, even if there is subtle differences in how it works etc, would that customer not be mad to at least evaluate the other technologies to see if the technology met their needs, and their budget?

    In terms of usability, that's a subjective argument, in the sense that to me, Microsoft's platform is more usable, but that's a familiarity aspect too.

    I agree to an extent that Enterprises have a strong preference for VMware, but remember, Enterprises are typically (but not always) the businesses that have had virtualisation for years, and have seen the benefits a long time ago, when there was little choice in the marketplace, but remember, the world isn't made up of Enterprises.  There are a number of 'analysts' that believe the battle won't be won in the Enterprise space, but in the Mid-Market, where, especially now in this economy, businesses are more price-sensitive than ever.  Even in the Enterprise space however, if I went in to the CIO, and the CFO of an Enterprise customer and said, "I can cut your costs using virtualisation, and instead of it being $315k on VMware (based on the example in this post), I can do it for $47k on Microsoft technologies", regardless of whether VMware is the industry standard or not, the customer is instantly intrigued.  Would Microsoft win this deal?  This is the key point - only if the technology provided what the customer needed it to provide.  If the killer feature above everything else was the Distributed Network Switch framework, and that was the most important feature for this organisation, then, simply put, they'd have to go for the VMware solution, however, if they were looking for a comprehensive deployment/patch/OS lifecycle capability, that would be provided by something like SCCM, then a Microsoft solution would fit the bill.  Choice is a great thing in our space now, as it forces both organisations to innovate, and it gives customers the chance to make up decisions not based on 'defaults' but on testing, evaluation, and budget.

    So, to say VMware is miles ahead - to you, maybe, to the next customer, maybe not - time will tell I'm afraid.

    Finally - fanboy I ain't - do I think our technology is perfect?  No!!  Do i appreciate and respect the competitive technologies?  Yes!  Will I admit if something is better than something we have?  Absolutely!  Will I defend my opinions when they have been misconstrued?  Definitely, and if that makes me a fanboy, then so be it.  Sales pitch though - it's all factually accurate, so, maybe it is, but that's not what it set out to be, so apologies for that!

    Thanks again Joe,

    Matt

  • Oh and Joe - You're right, I'm not a big fan of the bold/red lettering mid-paragraph, and on reflection, it does look a bit naff, so i'll remember that for future.

    I just wanted to ensure that the chaps at Virtual Reality didn't miss my emphasised points :-)

  • Matt, thanks for the response and explanation.

    I agree it isn't fair of the VMWare folks to mis-represent themselves or the cost of their solutions.  Also, I agree that Hyper-V can be better in some instances. Definitely agree that you have a right to defend your point of view, and in fact, I am quite glad you took the time to evaluate my comments and respond.

    I've enjoyed your mini-seminars on Microsoft

    's virtualization technologies very much.  Keep 'em coming!

  • Hi Joe,

    Thanks for the comment.  One thing to take away from all of this, is the customer wins - they have their budget, and they make their choice! :-)

    Hopefully my Microsoft overview videos can help influence them our way, but who knows! :-)  More videos coming soon on App-V!

    Matt

  • Matt,

    You claim you can make up for lower consolidation ratios by adding more RAM to the server to make up for it, and as you noted in your first post you could do so for an ESX box as well and get less hosts. Bottom line is, ESX Page Sharing and Ballooning will almost always result in an infrstructure requiring less memory. (and less hosts) The 'just add more memory' argument is your way of countering the fact that if more hosts are added to support the workloads, the cost gap very quickly closes. BTW price out adding more memory to a host if it is maxed on 4GB modules are you are forced to buy 8GB modules.

    What happens in a Hyper-V environment when host failures exceeds the host reserve? (Guest memory exceeds available cluster memory) That's right, some VMs stay off.

    What happens in Vmware? ALL VMs are still up and chances are Page Sharing and Ballooning have freed up enough memory that there is no performance degradation.

    This is not a simple feature check that can be glossed over - it has real world impacts.

    Let me ask you this, is it better to have less efficient software that you have to thow more hardware at to have equal performance? Nevermind that even with the extra hardware costs, the MS solution is still cheaper - this is not very "Green"

    Shouldn't we be encouraging support of products that are technically superior, in order to advance innovation, not ones that are cheaper but 'good enough,' have large marketing resources and leverage an existing monopoly to drive market penetration?

  • Hi Shawn,

    Thanks for taking the time to provide a detailed comment - I do value your opinion and your input.

    You are indeed correct, for a given host of XX GB RAM, ESX will consolidate a greater level of VMs than Hyper-V/XenServer, thanks to the Memory Oversubscription techniques, discussed in the first post.  Credit where credit's due - they are powerful, and are becoming more of a 'feature request' for Hyper-V in the future from customers (but not to the level of Live Migration!).  I appreciate the 'just add more RAM' argument, on the surface, appears quite thin, however, when you speak to Customers, and Partners, especially those in the Mid-Market and Small Enterprise, they are looking to save as much money as possible, yet they are rarely getting to the state where they max out the Server's RAM.

    When I speak to Partners who are embracing Hyper-V, it seems that, at least initially, they are using the same kind of server sizing, and the same kind of consolidation ratios as on ESX, so, as an example, customer has 40 servers, and after the capacity planning exercise, Partner X decides that they'll comfortably consolidate that down to 5 hosts, at 8:1.  Partner X states that we'll use a typical 2 x Quad Core CPU, along with 64GB RAM.  That's their standard type deployment, and this is quite common.  Sure, for that same 64GB RAM, you could get more VMs on an ESX host, but that's not to say that everyone will actually go through with it.  For those 5 hosts, agnostic Partner X positions 2 solutions for the customer (based on the requirements the customer has) of Hyper-V and System Center, versus vSphere Enterprise Plus and vCenter, and, based on the figures in the post above, the Microsoft solution will cost considerably less, yet still give the same results in terms of power savings, cooling, infrastructure savings etc.

    I completely agree with you about the 8GB DIMMS - boy-o-boy would I not like to spec up with those bad-boys!  However, I would instantly look at a 4U with 32 slots versus using a 2U with 16 slots, but then you could say, this means they would use more rack space in a datacenter etc, and costs would go up reflective of this.  To that I say, yes!  Assuming all customers are running in a datacenter, which again, not all customers have!  A large number of customers have their own 'datacenter' in a room somewhere, were floorspace isn't charged at a premium, thus the choice between 4U and 2U is less of a factor, if it means they are going to get the levels of RAM required.

    You're absolutely correct about if host failures exceed the host reserve, but you're talking about a scenario here that hasn't been accurately capacity planned to leave enough space, including, perhaps, an empty node, to account for this.  Yes, an empty node would be requiring power, cooling etc, but it's there for an emergency, and is optional.  This is the kind of capacity planning and accurate assessment that the skilled Microsoft Partner community would provide.  You could argue on the VMware side, if I lost 4 nodes of a 5 node cluster, would that remaining node be able to support ALL the VMs?  Even with Memory Oversubscription, it would be a tall order, but not without possibility.  On the Microsoft side, accurate sizing, especially with RAM, will help prevent getting into the issues you describe.

    In terms of having 'less efficient software' - what is efficiency? Efficiency to you, in this example, is obviously around memory, but to me, efficiency is around Management.  How efficient an infrastructure is, isn't related to memory, even in the physical world, pre-virtualisation, customers can have efficient, well-oiled infrastructures, and those that do, typically have management technologies, Microsoft or otherwise, providing that capability, and changes in that infrastructure are typically policy-driven, and well managed.  This should be no different in a virtual environment, so yes, Hyper-V might not be as clever with the memory, but it's performant in the way it does provide VMs with memory, and with System Center on top, provides the opportunity to have a very efficient infrastructure.  Microsoft has some strong messaging going forward around efficiency: http://www.microsoft.com/business/thenewefficiency/keynote/en/gb/ - So, in direct response, 'is it better to have less efficient software that you have to throw more hardware as to have equal performance' - if we treat, as you state, efficiency as memory utilisation, and hardware to mean RAM, rather than servers themselves, then while RAM is cheap, yes, I would say that is a viable compromise for a customer, and something that is then owned by them, and not paid for well into the future in the form of an S&S type agreement, which for them, may be more suitable.  Also, the RAM is always going to be useful, in the sense that if Hyper-V does being a memory oversubscription technology in the future, customers will have even more growth capability / headroom should they want it.

    In terms of your last point around 'Shouldn't we be encouraging support of products that are technically superior, in order to advance innovation, not ones that are cheaper but 'good enough,' have large marketing resources and leverage an existing monopoly to drive market penetration?' - I absolutely agree, we should be pushing support of technically superior products, but what is technically superior to you, may not be what is technically superior to another customer.  It's all about meeting customer requirements and doing it in such a way that meets expectations around features, performance, and budget. Would I expect VMware to have the 'best' hypervisor?  Sure!  It's the only one you have to pay for, so it would have to be the best, right?  Not necessarily.  Microsoft technologies, particularly Hyper-V and System Center, are driven by customer requests.  Take Live Migration as one example.  If Microsoft bake these features into the platform, and make them available at a lower cost, it's a win-win for Microsoft, Customer and Partner alike.  Customers get the features they want, Partners get the opportunity to work with customers and help them improve their whole infrastructure, for a lower cost, and Microsoft benefits as more customers use their technologies, and tus can help to feedback and improve the products for future versions.  However, that said, Microsoft could offer the cheapest technologies in the world, but if there was a key feature, say, Distributed Network Switch, that only VMware had, and the customer needed that to solve a key business requirement, would they go for VMware?  I would expect so!  But not without considering other platforms, and even the willingness to compromise on that feature for now, with the hope it will come in future versions of alternative platforms.  Coming back to 'technically superior' - on a Hypervisor for Hypervisor comparison, yes, ESX is a more mature, tick-box-feature filled platform, however, I wouldn't hesitate for a minute to state that System Center, as a suite, provides a much greater level of capability to manage the workloads than a pure VMware solution, at a much reduced cost too.  So, again, superior is a subjective opinion, and what we're finding in the UK, is that System Center is proving a fantastic platform for Partners and Customers.  Partners can help a Customer over a longer consultancy period, to implement a more comprehensive, infrastructure wide optimisation project, whilst, the customer saves on licenses, saves with virtualisation, and improves efficiency of their infrastructure too.  This kind of approach is resonating very well with Partners and Customers alike here in the UK, and further afield too.

    I hope that gives you a balanced side to my thinking - you may not agree with it, but I'm glad you took the time to comment.

    Cheers Shawn,

    Matt

  • Been reading your blogs and while I found first one more honest and clear, this one is treating the path set by Jeff Woosley (a marketing whizkid in MS churning irrelevant info). He likes to compare all the marketing aspects like VMware ESXi to Hyper-V & Hyper-V Server both (None Technical) and here you are making a comparison between Hyper-V Server 2008 R2 and ESX Server 4 / vSphere4 instead of Full Windows 2008 Server.

    Would really like to know how much a Windows 2008 R2 DC edition will cost for 60 CPUs (SA/No SA). This is for your statement "60 CPUs of Hyper-V Server 2008 R2, is $0.  Good start.", because Windows 2008 DC will cost for licensing and will be at par with vSphere4, and not Hyper-V Server 2008 R2.

    What Say?

  • Hi Animesh.  With the greatest respect, you're missing the point here - Hyper-V Server 2008 R2 is EXACTLY the same, functionality wise, as Windows Server 2008 R2 Hyper-V.  So, comparing Windows Server 2008 R2 Datacenter (which includes Hyper-V), this would be a cost also on top of the VMware pricing, hence why we compare Hyper-V Server 2008 R2.

    The advantage of buying WS2008 R2 Datacenter, is for the GUEST Operating System licenses, which you would need on both Hyper-V and VMware.  So, for your example, of 60 CPUS:

    Microsoft

    60 x Hyper-V Server 2008 R2 = Free

    60 x SMSd (Management) = $44640

    Server license for SCCM SVR, SCOM SVR and SCDPM SVR, and you’re looking at about $2605 total including 2 years SA.

    60 x Windows Server 2008 R2 Datacenter Edition (for the Guest OS's = $179,940

    Total cost = $227185

    VMware

    60 x vSphere Enterprise Plus inc 2 yr Plat S&S = $307,238

    1 x vCenter Standard inc 2 yr Plat S&S = $7,318

    60 x Windows Server 2008 R2 Datacenter Edition (for the Guest OS's = $179,940

    Total cost = $494496

    So, If a customer needs Windows licenses, they will buy them regardless of the Hypervisor.  If they don't, they won't!  Simple as that!

    $494,496 - $227,185 = $267,311 saving through Microsoft Virtualisation.

    I'll add to this, that Jeff is a Technical Program Manager, not a marketing whizkid, and I'd ask you to ensure you have the facts in place before trying to compare apples with apples.

    Matt

  • Hi Matt. I attended the Hyper-V seminar in Reading last week. Matt I guess my main point is you are still comparing ESX Enterprise Plus or Enterprise with Hyper-V and also using the VMWare Platinum support. Is it Microsofts view that Hyper-V is pitched directly against ESX Enterprise as a like for like product? I would also be interested if the Microsoft support agreement you use in your comparison is identical or better that an VMWare Platinum support agreement. I also wonder about the comparison where you add memory to the physical host because Hyper-V has no memory oversubscription. If you add memory to the host it also allows you to run more ESX VM's.

    If I understand it correctly the architecture of the hypervisors is also very different. Although Hyper-V is called a bare metal hypervisor it still requires the Server 2008 core as a root partition through which all hardware communications must pass or is this now different?

    The reason I ask is because from experience (admitedly not by counting how many patches are available on the website) we patch Windows servers much more frequently (often with reboots) than we do ESX servers.

    I will take away the data above and hopefully put together what we feel is a more balanced cost comparison. We have both technologies in house and our core business apps are Microsoft based so there is no biased here I just want to get through to the actual facts.

    Thanks for the presentation

    Regards

    Barry

  • Hi Barry,

    Thanks for getting in touch, and sorry for not responding earlier - I usually get an email telling me when someone has left a comment, but I didn't get one for yours and one other, hence the delay.

    I think what we need to be clear about, is I'm comparing hypervisor and management, with hypervisor and management.  When you say ESX Enterprise Plus or Enterprise with Hyper-V - this isn't what I'm comparing.  ESX, just like Hyper-V, is the Hypervisor, and that's it.  On the VMware side, you need vCenter to unlock capabilities like HA and VMotion, along with the more advanced capabilities like DRS, DPM et al.  On the Hyper-V side, you can unlock HA and Live Migration without needing an equivalent to vCenter, however granular management on top of the virtualisation platform is always recommended, I'm sure by both Microsoft and VMware!

    So, the reason I compare Enterprise/Enterprise Plus with a Hyper-V and System Center Suite solution, is because that's what's comparable.  If you wanted to achieve what Hyper-V and System Center Suite provide for Microsoft environments, you'd have to position Enterprise Plus, assuming you wanted apples for apples comparisons.

    Don't forget also, what you'd typically, in a larger environment, add on to a VMware project - you'd typically purchase an extra backup technology, and possibly have another technology to orchestrate patching, especially out to the desktop.  If you wanted to a technology to help you monitor workloads inside VMs, you'd be looking at another technology, AppSpeed of 3rd Party I guess, so all these aspects add up, yet System Center Suite does this, and more.  So, feel free to compare a vSphere Advanced project, with a Hyper-V and System Center Suite project, and I predict the costings would still be considerably cheaper on the Microsoft side, and the features, tick for tick, would outweigh a vSphere rollout, which lacked DRS, DPM, Storage VMotion, Host Profiles, et al.

    As for Microsoft support, my previous post, the cost comparisons I make in the session exclude S&S and SA, so the Partner can choose which to apply for each customer, however, with VMware, you have to buy S&S, whereas with Microsoft, it's optional.  So, a customer may not choose to buy S&S, and rely on the Partner providing support in the event of a failure.  As for Microsoft support, it's a tricky one.  I sometimes wish it was as black and white as 'choose gold or platinum, and for how long' as it would simplify things, plus, Software Assurance is not equal to support, so that doesn't help!  I do need to clarify exactly what support options are in place, however I know, in the simplest case, there is a per incident support, which is 24/7 etc.

    In terms of the Memory Oversubscription - yes, you could add it to ESX hosts too, but generally, a lot of our Partners aren't seeing this as a blocker in the slightest.  They run the capacity planner, the agree a 'comfortable' consolidation ratio (a lot of customers are still wary of lots of eggs, and few baskets, even on VMware!) and spec up the hardware acordingly.  Regardless of hypervisors, a number of Partners will deploy on 2 x Quad-Core, 48GB RAM as standard.  Sure, on that same amount of RAM, you'd get more VMs on ESX than on Hyper-V, but whilst the cost of RAM is less than vSphere, and you're getting greater capability within the management tools, customers are willing to have clarity around RAM usage, i.e. this 2GB RAM VM, always has 2GB regardless, vs. having more intelligence around the RAM, and paying for it.  Remember, if and when memory oversubscription does come in, it's quids in for those Hyper-V customers.

    The architecture of the hypervisors is very different.  It's different concerning drivers, it's different concerning hardware access.  But that's just it, it's different.  Different doesn't exactly mean bad - I would check out here: http://www.microsoft.com/video/en/us/details/3c31b7d8-7502-4505-84ab-9999717d8692 and here: http://msdn.microsoft.com/en-us/library/cc768520(BTS.10).aspx.  Remember, for performance, we provide enlightenment/paravirtualisation.  This is something VMware dismissed to a degree, yet are now implementing.  Also, if you're running on Second Level Address Translation (SLAT) hardware, memory usage by VMs is further optimised, as it's handled more in hardware.

    As for patches, I'd read Jeff's posts that I reference in here: http://blogs.technet.com/mattmcspirit/archive/2009/09/11/investigating-the-vmware-cost-per-application-calculator.aspx - all you need to do is look here: https://www.vmware.com/mysupport/s3portal.portal?_nfpb=true&_windowLabel=SearchPatch&SearchPatch_actionOverride=%2Fportlets%2Fpatchupdate%2FfindPatch&_pageLabel=s3portal_pages_downloadPatch_page and you'll see that there are a lot of updates for 3.5, with a lot requiring reboots.  Is that an issue?  I wouldn't say so, as long as you manage and patch your systems through policy, I don't see it is a problem.  Windows has a fair number of patches too, but Hyper-V itself, has much less.

    Hope that helps, and any questions, you've got my email :-)

    Matt

Your comment has been posted.   Close
Thank you, your comment requires moderation so it may take a while to appear.   Close
Leave a Comment