Investigating the VMware Cost-Per-Application Calculator

re: Investigating the VMware Cost-Per-Application Calculator

Jairo Cetina — Sun, 22 Nov 2009 06:29:29 GMT

Hi Didier.

I desagree with you post, I'm employed of a Microsoft Gold Certified Partner. MCSE, MCITP, CCNA and VCP but VMware is most powerfull virtualization infrastructure than hyper-V.

I was making many labs comparing Hypervisors and VM in VMware have more performance than other virtualizations solutions.

1. VMware have more experience in virtualization solutions.

2. VMware need less disk, RAM and CPU for operating system.

3. VMware have a better manage of hardware resourse.

4. If you add more RAM in a Hyper-V host them need add more CPU in each additional VM; it mean that you have to buy more procesor.

5. VMware hypervisor it's more secure that hyper-v.

Best regars,

Jairo Cetina

re: Investigating the VMware Cost-Per-Application Calculator

MattMcSpirit — Mon, 02 Nov 2009 09:43:02 GMT

Hi Mihir

Thanks for the kind words.

Comparing SQL in VMs and Multiple SQL instances isn't something I would typically compare (I'm 100% not a SQL guy! ;-)) although I believe this would be helpful for you: http://technet.microsoft.com/en-us/library/dd557540.aspx

Thanks!

Matt

re: Investigating the VMware Cost-Per-Application Calculator

MIHIR — Thu, 29 Oct 2009 18:07:21 GMT

HI Matt,

Excellent article and give me more insight about the comparing both product. I’m doing a research on ROI for SQL Server 2008 with VM or multiple instance on Windows 2008 with resource governor.

Have you ever thought about the comparison of multiple SQL instances on one physical server vs SQL server on VM or Hyper-V.?

re: Investigating the VMware Cost-Per-Application Calculator

MattMcSpirit — Sun, 27 Sep 2009 10:40:19 GMT

Hi Phil,

Thanks for the comment. I'm glad you enjoyed the post, and found it useful. It would be harsh of me to position memory oversubscription as a myth as such, but, in most (but not all) cases, it can be overcome by adding more RAM, however, I acknowledge that this isn't always possible.

For the mid-market and small enterprise Customers I tend to speak to, they rarely need to hit the 'hardware limit' i.e. the max supported RAM of the server, so they always have that opportunity to increase their scalability as and when they need to. Combined with the low entry costs of a Microsoft solution, they tend to achieve what they set out to do, with a strong ROI.

There's no denying memory oversubscription, through the different techniques it utilises, is incredibly clever, and powerful, but not everyone will use it to gain greater levels of scalability - moreso as a safety net.

Thanks again!

Matt

re: Investigating the VMware Cost-Per-Application Calculator

Phil — Sat, 26 Sep 2009 21:38:33 GMT

This is a great post and good to see someone pointing out some of the myths surrounding memory overcommit. When we were evaulating Hyper-V R2 (beta at the time) and ESX it was immediately obvious that we could double the RAM in the Hyper-V hosts against paying 3 times more for VMWare overcommit.

re: Investigating the VMware Cost-Per-Application Calculator

Lord Melch — Tue, 22 Sep 2009 15:16:11 GMT

>as they are the market leader

Humm I wonder for how long..

re: Investigating the VMware Cost-Per-Application Calculator

MattMcSpirit — Sun, 20 Sep 2009 10:51:52 GMT

Hi Ahmed,

1) - Your first question - well, this would depend solely on what the customer's day to day activities are! I imagine that, out of the box, the System Center suite, with it's 4 key technologies, will provide the capabilities to perform nearly all of the typical day to day activities, however, I would expect an IT Admin within a customer to spend the majority of their time in SCOM, as this allows them to monitor all the other components of their System Center environment, but at the same time, SCOM provides customised views if you have a 'Database guy' or an 'Email guy', they can just see the bits that matter to them. It's an extremely extendable technology, which really can be tailored to the customer's infrastructure. I'd strongly suggest you have a watch of this video: http://www.youtube.com/user/duanebms - they are monitoring a custom web app called PetShop, and the video shows how the customer can monitor, and react to changes during the day - very powerful.

2) - for Network/Firewall management - it depends on what you're looking to do. Windows Servers, from 2008 onwards, have a very granular bidirectional firewall, which can be configured centrally via Group Policy, so you could gain a strong level of protection just from this, and if you combine this with another in-box feature of 2008/R2 - Network Access Protection (NAP), this will effectively help to quarantine non-compliant OS's from the network, so for instance, if you have an OS that is out of date with patches, and you've started it up, rather than expose it to the full network, you can bring it up, the NAP server will pick up that it's out of date, and it will be quarantined immediately. Again, this is centrally configured. There's obviously nothing stopping a customer using a firewall they may have already invested in too.

3) - Dynamic Datacenter - Hmm, great question, and you know what, I would say it's pretty close, but from different angles. VMware have a more feature-rich, tick for tick, virtualisation platform than anyone else, however, we have a much more comprehensive, overall management platform than VMware, with System Center. Does either orgnisation truly know what their Dynamic Datacenter would look like? Possibly not, but in my opinion, System Center gives Microsoft an edge, because it focuses on the 1 area that is most important in the datacenter, and that is the workloads, whereas VMware, as of yet, predominantly focus on what you can do with the VMs. Obviously there is plenty of room for debate on that question, and 2 people's Dynamic Datacenters could look completely different, but I think both organisations are in a strong place, but for me System Center has some distinct, relevant advantages, but that's just me!

Hope that helps!

Matt

re: Investigating the VMware Cost-Per-Application Calculator

Ahmed — Sun, 20 Sep 2009 08:41:28 GMT

this is a great article and definitly a reference for arguments ;)

just to make sure i get the full picture,i got a couple of questions

1-what is the hidden costs on customizing vSphere or system center products to actually fit the customer day-to-day activities ,my guts is telling me this is not free in both

2-what is the alternative for Hyper-V customers to use for Network/firewall management ,how costy is that?

in your opinion who is getting closer to the full dynamic datacenter vision ,how many years will it take to get there ?

Ahmed

re: Investigating the VMware Cost-Per-Application Calculator

MattMcSpirit — Thu, 17 Sep 2009 11:18:33 GMT

Hi Shawn,

Thank you for taking the time to comment, and also for the flattering feedback. I couldn't say what Jeff thinks of it - not sure he reads my blog!

With regards to the memory overcommit - you are absolutely right, and I apologise wholeheartedly for overlooking this factor. If you're using DRS, there a much lower chance of running into the degraded performance due to hypervisor swapping. I guess if you've got the Advanced SKU or lower, then this will apply more, but, as you rightly state, anyone with Enterprise/Enterprise+ will have DRS, so should be in a better position :-)

Thanks for the comment - this is the kind of thing I'm looking for!

Matt

re: Investigating the VMware Cost-Per-Application Calculator

Shawn — Thu, 17 Sep 2009 05:28:17 GMT

Matt,

Great article.

If only more Microsoft employees were like you.

Your article should be required reading every MS virtualization person. Curious as to what Jeff Woolsey thinks of it...

One point you make while recognizing the value of memory overcommitment is that overcommitment assumes all the VMs aren't going to want all of their RAM at the same time and break the benefits. Well that's the nice thing about DRS, VMs will be migrated off to get things under control again. I doubt all VMs across your whole datacenter will want all their RAM at the same time.