History

At WinHEC 2006 Microsoft presented in details its new operating systems, Windows Vista and codename Longhorn, backup capabilities.
A new feature called CompletePC Backup will be able to produce a block level image of the whole machine while it’s running, thanks to the Volume Shadow Service (VSS), and restore it from bare metal, thanks to the Windows Recovery Environment (WinRE), even on different hardware.

The most interesting news for virtualization administrators is that the backup format used is .vhd, the virtual disk format used by Virtual PC and Virtual Server.
This should mean, even if Microsoft didn’t state it clearly, that anyone will be able to backup a running Vista or Longhorn and restore it inside a virtual machine. Which is a way to do physical to virtual (P2V) migration.

The idea of using backup capabilities for P2V tasks isn’t new: few months ago Acronis launched last version of its flagship product, TrueImage, with a new feature called Universal Restore, capable of modifying on the fly the HAL and other drivers needed by operating system during restore on virtual hardware.

Look at the WinHEC 2006 presentation Backup And Restore In Windows Vista And Windows Server Longhorn.

Quoting from the Intel patent application:

A method, apparatus and system for transparently unifying virtual machines (“VMs”) is disclosed. An embodiment of the present invention enables a user to interact with various applications on a VM host while unaware of the VM structure on the VM host.
The user may be presented with a unified desktop interface representing a composite and/or unified view of the VM host. Via this unified desktop interface, the user may perform all necessary commands and/or receive output. Invisible to the user, the unified desktop interface represents a unification console. The unification console may be an independent component (e.g., an enhanced VM) and/or a subset of a virtual machine manager (“VMM”) component on the VM host.
In either situation, the unification console may, alone and/or in conjunction with the VMM, route and/or redirect and/or transform and/or filter the user’s commands to the appropriate applications and redirect and/or copy and/or transform and/or filter the output from the applications to be displayed in the unified desktop interface…

Macsimum News wrote an extensive analysis of the application: part 1 and part 2.

Who worked with seamless windows technology in thin computing world should have a clear idea of how the whole thing will appear.

Mike Laverick published a notable 108-pages whitepaper for advanced users on how to perform an upgrade from ESX Server 2.x / VirtualCenter 1.x to new Virtual Infrastructure 3:

This guide is designed for people who already know ESX 2.x and VC 1.x very well.

It based on the “delta” two-day training that can be optionally attended for those people with prior experience. It also contains additional information beyond that course. This guide is certainly NOT for novices or new-users!

Although I’ve chosen to call this an upgrade guide, it’s by no means a definitive statement on upgrading. For that you need to read VMware documentation. I also view this guide as upgrade of skills as well as software.
It is not a comprehensive guide to ALL the differences – just the primary ones. I hope to make this guide gradually more comprehensive, and cover all new features. As you might gather there’s a lot marginal GUI like changes which are not included…

Even if it’s based on beta and RC builds no books are yet available on the topic so it’s the best resource out there right now.

Download it here.

Update: The guide has been updated to reflect VMware product final releases and has been expanded in several areas.

Quoting from TechWorld:

VMware’s virtualisation software, which leads the market by a considerable margin and is fast becoming the platform of choice for server consolidation projects, is too resource hungry, too slow and too expensive, according to the boss of a much smaller virtualisation company.

Serguei Beloussov, CEO of SWsoft which sells virtualisation product Virtuozzo, has slammed the virtualisation industry’s standard-bearer, saying that Virtuozzo’s OS shim-style virtualisation is better than VMware’s all round, in that customers can fit more virtual machines (VMs) into a server, and that it’s easier to manage and so costs less to run.
…
On the other hand, VMware reckons its approach is more robust. With Virtuozzo, if the OS kernel falls over, all the VMs are wiped out. This not possible with VMware’s ESX Server because each OS is self-contained and runs above a VMware hypervisor.

“VMware is I/O intensive”, he said, claiming that its overhead could be 5-10x higher then that of his own product. “The problem is that you cannot find out because the VMware licence agreement means you can’t publish performance comparisons,” he said.

Read the full article at source.

This article’s affirmations should be read in pair with SWsoft reporting about Microsoft TechEd 2006 attendees interest in Virtuozzo.

Claims about performances are very interesting: while VMware licensing prevents public publishing of benchmark comparison (something could change soon), I would be glad to read any SWsoft internal analysis and testimony to virtualization.info readers what I found out.

After hearing some debatable claiming from a Microsoft salesman (something Microsoft should start to control much more carefully), Geert Baeke wrote a piece about how differently virtual machines high availability (achieved with host OS clustering) is treated in a Virtual Server 2005 R2 infrastructure and in a ESX Server 3.0 + VirtualCenter 2.0 + VMotion.

The article exposes true differences and it’s worth to read, but approach it carefully. Going beyond a mere technical point of view I would object the comparison should not even take place considering that VMware solution in the scenario starts from 5,750 dollars. Virtual Server offers weaker high availability capabilities but it costs zero.

The real problem is Microsoft insists to compare Virtual Server to VMware ESX Server + VirtualCenter (now called Virtual Infrastructure). It’s not the case. And everybody in the market should avoid to mimic such bad marketing approach.
Let’s compare Virtual Server 2005 R2 with upcoming VMware Server 1.0. That would be much more honest.

The SWsoft blog about Virtuozzo published a small but interesting report of customer interest at just concluded Microsoft TechEd 2006 conference:

• A lot more people are using virtualization compared to TechEd 2005. 80% are using virtualization and 100% are evaluating it for deployments – compared to 50% and 60-70% respectively at TechEd 2005
• VMware users were much more interested to alternative approaches. Last year VMware users may or may not have had the time to talk to us… this year they did
• Hardware virtualization is still used mostly for testing and development. Only about 25% or less of the people I talked to use it in production
• About 80% of the production hardware virtualization users mention “OS sprawl” either as a problem they’re dealing with now or something they’re anticipating in the near future

Read the whole article at source.

I would further investigate why customers could be start looking for alternatives. In this other article SWsoft suggests reasons.

Dunes published a Flash demo of its flagship product, Virtual Service Orchestrator (VS-O), used for semi-automating VMotion tasks in a VMware virtual infrastructure.

Check it here.

VMware seems to be restless this year. Immediately after the Virtual Infrastructure 3 release the company announces a new initiative in collaboration with Intel:

Intel Corporation and VMware, Inc. announced the launch of the Intel-VMware Virtualize ASAP, a global program to further accelerate IT customer deployment of applications in virtual environments using VMware Infrastructure on Intel Xeon processor-based platforms.
…
Virtualize ASAP benefits customers by enabling software vendors to develop best practices for optimized application deployment for VMware Infrastructure on Intel-based server platforms. Through this engagement with leading software vendors, Intel and VMware will jointly drive tuning, feedback and technical optimization for VMware virtualization software on Intel processor technologies. Over the long-term, Virtualize ASAP will help direct continued improvements in support for VMware software on Intel’s platforms and processors.
…
Software vendors who are initial participants in Virtualize ASAP include: Altiris, BEA, CA, Cassatt, Citrix, CommuniGate Systems, DataSynapse, Hyperion, Interwoven, LANDesk, McKesson, MySQL, NetIQ, Open Country, Platform Computing, SAP, SAS, SyAM, Sybase, Symantec, Tangosol and UGS…

Visit the official Virtualize ASAP site.

Quoting from the IBM official announcement:

…
The new IBM Virtualization Test Drive program is designed to help partners more easily sell IBM’s Virtualization Engine portfolio. To kick off the partner program, IBM is using selected Business Partner Innovation Centers throughout North America to demonstrate to SMB customers the benefits of virtualizing their server and storage infrastructures.

The program includes assistance in setting up a center of competence, technical and industry education, sales enablement and support and tools to help IBM Business Partners assist their clients to design and implement virtualization solutions around IBM’s Virtualization Engine portfolio. The IBM Business Partners can also provide a “real world” environment in their innovation centers to simulate client scenarios and test applications.
…
IBM is expanding the program to Europe and plans to roll out the “Virtualization Test Drive” partner program worldwide later this year. The partner program augments IBM’s broad internal sales force that focuses on selling IBM’s Virtualization Engine portfolio. Additionally, the program complements IBM’s software partner initiative — Ready for Virtualization — that extends support for applications on IBM servers and storage managed by IBM’s Virtualization Engine portfolio…

Network World published an article providing some interesting values on server consolidation trends:

…When companies first apply virtualization, the average ratio is around 15:1. After one year it drops to 8:1 and after three years stabilizes at an average of 3:1. In order to fit 15 to 20 virtual machines on a single server, the average CPU utilization of these virtual machines must be 5% to 8%. These represent servers with permanently low demand on CPU resources or with intermittent demand: testing and development servers, small Web servers with many static pages, rarely used applications, and so forth….

Read the whole article at source.

I would say this values could be misinterpreted.
One thing it’s true: when a company embraces server virtualization starts loading the virtual infrastructure with low resources-demanding services, typically web servers, DNS, DHCP and others. This approach easily leads to achieve 15:1 or 20:1 ratios. After that 2 kind of things can happen:

• the company grows and virtualized services start demanding more host resources. In this case, with the same physical hardware available, less virtual machines can be allowed to run concurrently.
• the company is convinced virtualization is cost-effective and reliable enough and start moving in the virtual infrastructure also more critical services like mail servers, databases, and other hungry-resources services. Also in this case the amount of virtual machines allowed per host has to decrease.

In both cases the initial VMs / host ratio decreases, even reaching bottom values of 3:1. But what is unsaid, and Network World is missing to report, is that companies experiencing such scenarios 99% of times buy more physical servers to redistribute all virtual machines running. So the final server consolidation ratio should be calculated among all available hosts.