let's not be naive
You should at least state who paid for that study.
VMware reckons its virtual machines (VMs) cost less than rivals because you can stuff more of them in a physical server. It cites a Taneja Group study, just released, to support this claim. Jeff Boles, a Tanaja analyst, says: "We’ve validated in a number of tests that VMware virtualised servers can run twice as many …
You should at least state who paid for that study.
Nice press release there.
You know someone could take that and use it as the basis of a well balanced and carefully thought-out article - we call them journalists.
Also, it is possible to just take that information and spread copies of it around un-criticized and with all the original spin intact - we call them photocopiers.
Would of been nice if they had actually shown exactly what made up the various line items in the cost comparison. What network gear are they using? Storage? What servers? How come the hardware costs on the MS side are $20k more, the doc says they are using the same servers, are they using different network and storage? If so why?
Are they running enterprise edition of ESXi or Standard or Foundation? What's included in the 'management infrastructure' ?
Hyper-V isn't even a consideration for me I'm more interested in comparing their costs to my own, just to see how it stacks up.
So, if you artificially limit yourself to 16GB of memory, VMware can perform better, and they're using that to justify the cost with memory prices how they are today?
Tell you what, go spend £600 on an extra 32GB of memory and run that test again VMware, then I might be interested.
...ah, here it is. Heh, big fat zero (personal use) and I can create new images; it's not just a player.
It may have issues (USB support is a shade wonky) but it is sooooo much cheaper that I'll deal with that (I don't really need USB support in my VMs anyhoo).
And as for working with them on a desktop...well hellooooooo seamless windows!
Strange that this Vmware-sponsored study makes no mention of Parallels Virtuozzo Containers, which actually does deliver greater virtual densities and which thrashes ESX and every other hypervisor on any ROI/TCO analysis... HP & Quest (partners of both VMware and Parallels) both have white papers floating around on this subject.
The prime candidates for virtualisation (around here at least) are things that naturally end up bandwidth limited (either disk or network) as the server load goes up. How exactly does VMware make any difference at all when this is the case?
Myxiipix unfortunatley you're wrong. If you put 32gb in the server, the value would be even better for VMware. Basically the hypervisor has some technology which enables better consolidation ratios. I.e more memory means better ratios. Managing memory in a virtual environment is more important some say than CPU. In fact VMware has had this built into the hypervisor since 2002. Way before MS Hype and Xen variants were even dreamt up in the developers heads.
Errm, I like virtualbox as much as anyone but we are talking virtualizing datacentres here, not running XP in a VM so you can continue to enjoy the dubious benefits of using Outlook as a mail client.
Paris, because I think about her box virtually all the time
Currently ESX 3.5 supports upto 256GB of ram in a physical server, and upto 64GB of ram per guest. ESX 4.0 release is just around the corner in the next 2-3 months will support even more.
As AC said, if you put more RAM into a server with VMware the ratio should only go up in favor of VMware (provided you aren't fully CPU bound). Additionally VMware with it's memory "deduplication" (if groups of ram are the same it points all servers to one copy rather than having 20 copies of the same dll, etc like a symlink or shortcut) can logically present more ram than is in the box physically.
Reg, you're getting lazy. Copy and paste of press releases is weak. Let eWeek do that. Next, the BS detector is ringing quite loudly. For example,
"Jeff Boles, a Tanaja analyst, says: 'We’ve validated in a number of tests that VMware virtualised servers can run twice as many applications than other hypervisors at equal or even greater performance levels.'"
Let me get this straight. VMware claims to be able to run twice as many vms at the same performance or FASTER than the competition. Let me guess, these VMs were idle at the desktop? Not buying this for a moment. Let me make a wild guess at who paid for this "analysis."
Clearly VMware is feeling the heat from Xen and Microsoft after price gouging for years.
They don't really do the same thing......................
You could also re-write your corporate applications to run like SET at home did on volunteer home PCs if you didn't care about response time.
as much as I love VMWare and it's functionality (having not tried HyperV yet), I still have a big gripe with this study...
"In March of 2009, VMware contacted Taneja Group to conduct an assessment of
how VMware ESXi 3.5 and Microsoft Hyper-V compare"
So long and thanks for all the fish.
You lot can debate the merits of different virtualization technologies all day, and it's true VMware has some cool memory management techniques, but my over-riding thought from this piece is that after the press release, I was expecting the journalist to say "so, we put this to the test..."
For server (eg DL580) memory? Bargain, where can I get some?
The Register giving VMWare a load of free publicity, uncritically just republishing VMWare publicity bumpf. Biting the hand that feeds IT but, funnily enough, not biting the hand that feeds the Register's advertising coffers
Yeah VMware can consolidate memory, so it can be shared IF there are common pages between Server builds. So this is only of use if all your VMs OS are up to the same patch level on the same ESX server. Sure you can over commit memory on a ESX server at the cost of performance. Is it better to do that than stack up the memory and pay a lower CitrixXenServer price for the software? I bet it is not.
Interesting that they give the dollar costs there in this press release- I would like to see the figures (especially for the VMWare licences) in GBP instead, the $ figures there look way too cheap to me.
@the spectacularly long named chap....sounds like you'd be better off upgrading your LAN/ buying some new storage devices first. Aslo, people usually recommend against virtualising any app which runs high CPU too, as the VM overhead (somewhere between 10% and 25% of processing power is used to run it, though those figures are very hard to confirm) will actually reduce the available CPU.
I have a feeling Zones on Solaris would win any virtual machine density competition.
Just don't try to patch the damn things.
Says pretty plainly there that the MS solution required 5 servers instead of 3. That accounts for their 2 extra VMs for the VM environment.
And I'm only a PHB working the help desk, not a network tech. But at least I can read.
. . . even on large scale discounted global accounts . . . its what happens with fiirms the VMWare that have dominant market shares and this drives customers to move towards more efficient alternates... it seems amazing that Microsoft haven't fully got their act together on this and have left the field largely open to VMWare ...
Wake Up! Negotiate and haggle.
. . . always hold out for a deep discount, especially at the end of a quarter ! : )
people should be weighting the cost of virtual servers against the cost of physical boxes...
on a cost per application basis.
there is no point in having the latest greatest machine as an intranet server in a company of 20.
and even if you don't have the latest greatest, you still have to power the older machines
that is to say if you have a small company
and you have a
a domain controller,
an email server
a web server (for external clients to see your business site)
a sharepoint server (for intranet)
a database server (for the sharepoint site)
a proxy server
a terminal services server
a central AV server
a backup server
and they are all separate machines, that are very underused, then you should be considering putting those onto one box.
you won't save anything on licensing. but you will save on disk costs, (as your VMware disks can be better spread around your storage). you will save on machine costs, (because you nly need buy one machine rather than nine) you will save on power and cooling costs -because again 1 machine takes less power than nine.
you'll also make a saving on space costs as well...
I mean, if you assume that you;re working in an office where those are your internal corporate servers, then you have development servers, test servers etc that are still needed, that could be the difference between needing one rack, or two rack in your server room.
How would VMWare perform in these scenarios?
I would love to see 64 OS's running on a 1U high single socket server.
SUN's LDOM's make VMWare look pretty terrible.
Or 256 OS's running on a 4U high quad socket Server.
fscked by SHA-1 collision? Not so fast, says Linus Torvalds