I’m not here to blow Microsoft’s trumpet. I’m not a MS employee, don’t own shares and, to be honest, they have a billion dollar marketing engine and Jerry Seinfeld to do that for themselves.
But when I read something like InformationWeek’s “9 Reasons Enterprises Shouldn’t Switch To Hyper-V” by Elias Khnaser then I have to say something. Let’s go through this point by point.
1) Breadth of OS Support
Referring to non-Microsoft OS’s “Hyper-V, however, supports only Windows and SuSE Linux”
Actually it also supports Red Hat Enterprise Linux. And the IC’s have been released under GPLv2 and are finding their way into other Linux distros as I type this. Hey Elias, seeing as you aren’t a fan of Microsoft, I guess you don’t Bing. Try Google and do some research next time.
2) Memory Management
“For starters, it recommends having a host in standby mode, which means, ‘Have a host that is not serving VMs running so that in the event of a host failure, the standby host can be used to cover for its martyred cousin’.”
This is known as host fault tolerance. If a host fails then you automatically fail over the VM to another host. It isn’t unique to Microsoft. Xen and VMware do this too. You allow for a spare host or two (depending on the size of the cluster and fault tolerance required) so that if a host fails you don’t lose your VM’s. It’s a GOOOOOOD thing.
“If you don’t have memory oversubscription, how exactly do you expect to power-on VMs when a host experiences hardware failure?”
Yes, Hyper-V does not have memory over subscription or RAM bursting. Memory oversubscription is not supported in production by VMware. If something goes wrong you are told to turn it off by VMware support as step #1.
“Hyper-V’s reliance on a general-purpose operating system, in this case Windows Server 2008, makes it a security vulnerability unto itself”.
Oh really? How many breakout attacks has Windows Hyper-V had? Or Microsoft virtualisation full stop? Zero. How many have VMware had on ESX in the last 2 years? 1.
How many Microsoft patches for Hyper-V or Service Packs have broken virtualization or lost VM’s in the last 2 years? None. How many updates have broken ESX hosts in the last two years? At least two, including some Update 1 for ESX 4 which was withdrawn last week.
How easy is it to patch a Windows Server 2008 R2 cluster? It’s a doddle with WSUS. ESX or Xen? I doubt it’s so easy. Heck, an ESX patch is a complete OS upgrade.
If you so choose you can run Hyper-V Server 2008 R2 or a Core installation of Windows Server 2008 R2 to reduce the patching footprint. I choose to go with a full OS for hardware troubleshooting reasons. If some hardware breaks on my hosts I’ll ID it in no time at all – heck the OpsMgr HP Insight Manager management packs make that easy but I can also use the HP on-server tools in a GUI.
4) Live Migration
“Considering Microsoft’s frequent weekly updates for Windows Server 2008, that would take an administrator double or triple the time it would an ESX admin just to move VMs from host to host in order to apply security patches and properly secure his deployment”.
Really? All VMware implementations can have multiple VM’s fail over at once on 1GB Ethernet? Seriously, I wasn’t aware that this was possible. Does the VMware solution require 100% host fault tolerance with VM shadowing for this?
Hyper-V allows you to initiate multiple Live Migrations at once but they are sequential.
Wait a second: “Considering Microsoft’s frequent weekly updates“. MS normally releases updates every second Tuesday (Wednesday for those of us in Europe and east of here). And if you use any sort of patch management then it’s up to you when you do that patch deployment. Using VMM 2008 R2 you can put a clustered host into “maintenance mode” and go do something else for a while. The VM’s will be moved automatically to hosts selected by Intelligent Placement.
5) VM Priority Restart
“If you intend on running all virtual—and you should—the ability to prioritize your VMs by importance is crucial, and the ability to recover from host failures based on VM importance is even more crucial”.
When I restart a standalone (un-clustered) Hyper-V host I actually do have VM start prioritisation. You can specify how long after the parent partition boots up that each VM should start up.
“In the event a host that is running 60 VMs fails, for example, I want to make very sure that my virtual infrastructure can restart my failed VMs on another host in a certain order”.
I’ll grant you that one. For a workaround I suspect you could do this if using VMM. Specify a priority value in the VM custom properties in VMM. Write a PowerShell script to gather the names of all non-running VM’s on that host. Query those VM’s for the custom value. Order the start-ups by that value.
“I don’t want Exchange, SQL, and IIS to come up before my domain controllers, DNS server, or DHCP servers”.
We who work in virtualisation call that chicken and egg. I think you’ll find that VMware recommend that there should be at least one physical domain controller. I certainly would advocate that, e.g. Hyper-V/VMM work best in a domain and why start up the hosts before the DC’s?
6) Fault Tolerance
“This feature takes system availability to highs that are truly unheard of, and to no one’s surprise, it is available only with vSphere. The ability to run a single VM in lockstep with a shadow VM simultaneously, executing on both primary and secondary VMs at the same time, provides for continuous high availability that we never had in the physical world with this much ease.”
OK, which is it for you Elias? All the fault tolerance in the world or not spending money on hosts. Go back to the quote on point 2 and you’ll see you are contradicting yourself.
Actually, you can do Live Migration between sites with Hyper-V. It leverages hardware solutions from the likes of Compellent or HP LeftHand to create a cross-WAN/campus CSV and then do Live Migration across that. EMC has something for their Clarion and HP have CLX for the EVA but they don’t support CSV yet.
7) Hot Adds
“In a virtual environment, however, there should be no reason why we cannot add more memory, disk, and peripherals on the fly to any powered-on VM. Except if you’re using Hyper-V”.
You can hot-add SCSI disks in Windows Server 2008 R2. You just need to have added the virtual SCSI controller. Hot add of RAM is not possible.
8) Third-Party Vendor Support
“However, when we examine the third-party tools that support Hyper-V and those that support vSphere, the gap is significant and swings heavily in VMware’s favour”.
That’s probably true. But quantity does not equate to quality. There are one or two partners I’d like to see supporting Hyper-V that aren’t. But I’ll tell you this much, Microsoft’s more open approach sure seems more appealing than the “don’t dare compete with us” approach displayed by VMware at their conferences toward their partners, e.g. Veeam daring to back up VM’s on ESXi.
And anyway, I have solutions available to me for anything I need to do on Hyper-V/VMM.
“When choosing a virtualization infrastructure, you are making a strategic decision about the basis upon which your organization’s critical systems are going to run. It is a decision that will have far-reaching consequences; this is not some piece of software that you can just decide to change half way through the project”.
Hyper-V has such a small footprint that maturity isn’t a big deal. Does Hyper-V/VMM have all the bells and whistles of VMware’s ESX/vSphere? Nope.
Here’s the facts about Hyper-V/VMM. They are rock solid. 18 months of usage and no stability or performance issues. Manageability is easy thanks to VMM and OpsMgr. My job is easier thanks to them. I’m excited about how well Hyper-V/VMM have worked out and about the future. The new stuff coming from the Opalis acquisition, where things are going with Azure integration, server application virtualisation (independent of the OS), and the potential for leveraging boot-from-VHD for V2P based on performance monitoring … it’s all going to be fun over the next two years.
I wish journalists like this guy would do some research before they write. These articles are misleading.