VMware Inc.’s ESXi 3.5 platform consistently posted a 50 per cent VM density per physical server advantage over Microsoft Corp.’s Hyper-V, leading VMware users to lower costs per application and reduced guest OS licencing fees, according to a new report from the Taneja Group.
Jeff Boles, senior analyst and director of validation services at the Hopkinton, Mass.-based research firm, said while ESXi demonstrated a “1.5 to 1” density advantage over Hyper-V across a multitude of application workloads, in some cases ESXi could achieve a “2 to 1” advantage.
That means that if users can safely consolidate eight servers on Microsoft’s platform, they will be able to comfortably consolidate 12 to 16 servers with VMware, he said.
“The biggest surprise of the report was really about the cost differences between these two solutions,” Boles added. “You see quite a bit of messaging in the media about Microsoft being cheaper to buy, but the VM density factor is such a big influence in the total cost of ownership.” Boles found that the cost per application in a VMware infrastructure was between five to 30 per cent less expensive than in a Microsoft environment.
But this gap will not last long, according to other industry experts.
Tom Bittman, a virtualization analyst with Gartner Inc., said the density advantage can be attributed to VMware’s memory overcommit feature, which allows companies to use more virtual memory on a server than they would in the physical environment.
Because some applications and servers allocate memory in case of an emergency, memory overcommit allows you to bypass these precautions and successfully run more applications than would actually fit in your memory – so long as the applications don’t end up needing to use the space.
“On average, what we’ve seen is that VMware users have about 120 per cent of virtual memory configured to physical memory, so a 20 per cent higher density than they would have if they didn’t have memory overcommit,” Bittman said.
“So when (Taneja Group) says eight to 12 or eight to 16, they are giving ESX too much credit. I think it’s real, but it’s not that wide of a gap.”
In its report, the Taneja Group estimated that about 70 per cent of VMware customers use the platform’s memory overcommit feature.
Bittman said that once Microsoft follow through on its plans to incorporate memory overcommit – which is expected soon – the density gap will significantly close.
In addition to that, Microsoft is providing its answer to VMware’s VMotion functionality with a new feature called Live Migration, which involves the moving of running VM’s from one host to another without the interruption of services. The new feature is available as part of an update to Hyper-V in the Windows Server 2008 R2 beta.
“For early adopters of R2, who are now using live migration, it’s not just a nice feature that allows you to balance workloads, but it’s necessary to managing planned downtime in a Hyper-V environment,” Bittman said.
This was one of the reasons Hyper-V was being held back, he added.
According to London, Ont.-based Info-Tech Research Group Ltd., VMware’s VM density cost advantages disappear if the client only requires supporting a small number of VMs, which is the case for the SMB market.
“Hyper-V is not ready to compete against the enterprise version of VMware,” Habeel Gazi, research analyst at Info-Tech, said in an e-mail. Although it is significantly less expensive per VM, given the same number of VMs on each physical server, than the enterprise version of VMware, the latter is the better option for large customers.
Hyper-V is good enough for the SMB market, he added, and as the Redmond, Wash.-based giant continues to integrate new features to its virtualization platform, the challenge will be to maintain its price advantage in the process.
And with the vast majority of Fortune 1000 companies already on board with VMware, it would appear that Microsoft will have to settle for the smaller players at this point.
“It’s going to take something very compelling to move those big users off of VMware, but for the small businesses that haven’t virtualized yet, I believe they are the ones that are going to be heading toward Hyper-V in droves,” Bittman said. “The difference is large enterprises have been doing it for a while now. They understand it, they are investing in it, and they don’t need to be sold on it.”
In the small business market, companies often want to start cautiously and run test projects, he added, which has contributed to Hyper-V’s somewhat sluggish start.
Another big concern, Bittman said, is that Hyper-V relies on a parent operating system, Windows, which can create issues of its own.
“You are running on top of a very thin hypervisor, but on the side, you’ve got this big fat copy of Windows,” he said. “That’s where you get driver support, management support, and so on. But that copy of Windows is a huge single point of failure, so if it goes down, so do all the virtual machines.”
Many Hyper-V users will see this issue put into practice every Patch Tuesday, when administrators have to take down all of their virtual machines when updating their OS, Bittman added.
For Boles, in addition to VM density, the Microsoft platform is also currently saddled with too much complexity, especially when you look at OS and management infrastructure licensing.
“And who knows if Microsoft can fix that, because licensing has been the classic challenge for organizations using Microsoft infrastructure solutions,” he said.
Ultimately, Boles warned SMBs that haven’t gotten their feet wet with virtualization to weigh their options carefully, look beyond the upfront costs, and determine the TCO of the virtual platforms they are considering.