virtualization.info

Last week Microsoft and Red Hat announced the certification of their operating systems, Windows and Red Hat Enterprise Linux (RHEL), on each other virtualization platforms, Hyper-V and KVM.

It is a major announcement in many ways.

First of all, customers that have Windows/Linux mixed environments finally have a decent choice. 
Side by side with Novell SUSE Enterprise Linux, now Hyper-V (both R1 and R2) supports RHEL 5.2, 5.3 and the new 5.4.

More importantly, Microsoft and Red Hat validated the use of Windows Server 2003, 2008 and 2008 R2 as guest operating system on the KVM implementation that comes with RHEL 5.4.

On top of that Microsoft has even accepted to provide support to Red Hat users that run most of its enterprise applications inside KVM virtual machines.

Now, and only now, Red Hat has something concrete to tell to the customers.
With the large majority of virtual machines running Windows worldwide, without this mandatory step the new Red Hat offering couldn’t be considered anything more than an interesting future platform.

Thanks to the Server Virtualization Validation Program (SVVP) instead, KVM, or at least the Red Hat implementation of KVM, is at the same level of VMware ESX, Citrix XenServer, Novell Xen and Oracle VM Server in terms of support for Microsoft technologies.

Now Red Hat has to hurry up and show the serious stuff.

Labels: Alliances, KVM, Microsoft, Red Hat

In early September while most of the virtualization community was busy in San Francisco for the VMworld 2009, Red Hat was finally releasing the first piece of its new virtualization offering in Chicago at its Summit 2009.

The market expected the company to launch the new Enterprise Virtualization Hypervisor (RHEVH, a minimal version of RHEL plus KVM that could compete against VMware ESXi or Microsoft Hyper-V Server), and the new Enterprise Virtualization Managers (EVMs) for servers and desktops. But Red Hat only released RHEL 4.5.

In March the company announced that these new products would be released sequentially, starting mid 2009 and for next 18 months, but for now the general public knows nothing but a few technical details unofficially published by some a beta tester.

The ones that attended the Red Hat Summit in Chicago (or visited the Red Hat booth at VMworld) knows more. Luckily, Red Hat published some breakout sessions’ videos of the event, so we all can watch the ones related to virtualization:

Linux-KVM.com published an extensive synopsis of the first one above.
Here’s some points that are worth a highlight:

• Red Hat will support ISV software certified on RHEL whether it’s running on bare metal or running on the RHEL kvm or standalone kvm since it’s the same codebase.
• RHEV standalone kvm has a very small footprint of < 100mb in size which makes it easy to do things like pxe boot it.
• Hosts can scale host to 96 cores and 1TB RAM.
• Guests can scale up to 16vcpus and 256GB RAM.
• Supported Linux guests includes RHEL 3,4,5. Supported Windows drivers available for Windows XP, 2003 and 2008.
• NUMA, power management, memory page sharing (ksm) are some other important features. KSM important for density, very important and will be in product from day 1.
  Light workload VMs on a 48 core machine: 256 GB RAM could run more than 600 VMs.
• Testing results from internal and customers showed SAP workloads: 85-95% performance, Oracle OLTP: 80-92% bare metal. LAMP stack showed better than bare metal performance. Java achieved up to 94% bare metal.
• The management tools will be released in later half of 2009.
  Supports high availability by allowing VMs to automatically restart on other host when host having problems. Supports system scheduler at cluster level, live migration and power saver mode. There’s a maintenance manager that will automatically live migrate vms off servers during scheduled maintenance. Also includes monitoring and reporting tools.
  Support image management including templates and thin provisioning
  
  Note from virtualization.info: Red Hat published a video of Red Hat Enterprise Virtualization Manager on YouTube that we are featuring in the website sidebar and on virtualization.tv.

Labels: KVM, Red Hat, Releases

The formal launch of the new Red Hat virtualization offering based on KVM is just a few days away.

Excluding the products names, so far most Red Hat didn’t disclose any detail about the platform that will replace its previous implementation of Xen.

For the impatient ones, Mark Wilson published some concrete information about this product that are worth a check (our emphasis):

…It’s a standalone hypervisor, based on a RHEL kernel with KVM, and is expected to be less than 100MB in size.

Bootable from PXE, flash, local disk or SAN it will support up to 96 processing cores and 1TB of RAM, with VMs up to 16 vCPUs and 256GB of RAM.

Red Hat is claiming that its high-performance virtual input/output drivers and PCI-pass through direct I/O will allow RHEV to offer 98% of the performance of a physical (bare metal) solution.

In addition, RHEV includes the dynamic memory page sharing technology that only Microsoft is unable to offer on it’s hypervisor right now; SELinux for isolation; live migration; snapshots; and thin provisioning.

…

supporting guests from RHEL3 to 5, and from Windows 2000 to Vista and Server 2008 (presumably soon to include Windows 7 and Server 2008 R2).

RHEV is an x64 only solution and makes extensive use of hardware assisted virtualisation, with directed I/O (Intel VT-d/AMD IOMMU) used for secure PCI passthrough together with PCI single root I/O virtualisation.

…

The real story is with management and Red Hat is also introducing an RHEV Manager product.

I was impressed with (that I don’t remember seeing in System Center Virtual Machine Manager, although I may be mistaken) is a search-driven user interface. Whilst many virtual machine management products have the ability to tag virtual machines for grouping, etc., RHEV Manager can return results based on queries such as, show me all the virtualisation hosts running above 85% utilisation.

…

The third part of Red Hat’s virtualisation portfolio is RHEV Manager for desktops - a virtual desktop infrastructure offering using the simple protocol for independent computing environments (SPICE) adaptive remote rendering technology to connect to Red Hat’s own connection broker service from within a web browser client using ActiveX or .XPI extensions.

Red hat claim that their VDI experience is indistinguishable from a physical desktop including 32-bit colour, high quality streaming video, multi-monitor support (up to 4 monitors), bi-directional audio and video (for VoIP and video conferencing), USB device redirection and WAN optimisation compression…

Thanks to DABCC for the news.

Labels: KVM, Qumranet, Red Hat

In May, during its main conference Synergy, Citrix announced the existence of an open source virtual switch that may compete with the Nexus 1000V that Cisco made available for VMware vSphere.

In early June, the Citrix CTO Simon Crosby shared a very few details about it, but so far most of the virtualization community doesn’t know much about it. But the official website about the project quietly appeared online now: the product is called Open vSwitch and is released under the Apache 2 open source license.

The first release (which is almost complete and available online as well) is designed to support distributed networking (like the Cisco Nexus 1000V) and includes the following features:

• Visibility into inter-VM communication via NetFlow, SPAN, and RSPAN
• Standard 802.1Q VLAN model with trunking
• Per VM policing
• NIC bonding with source-MAC load balancing
• Kernel-based forwarding
• Support for OpenFlow
• Compatibility layer for the Linux bridging code
  (The Open vSwitch can be even used inside a plain Linux distribution in place of operating system bridge)

On top of that the following features are part of the roadmap:

• User-space forwarding engine
• sFlow
• Compatibility layer for VDE
• Ethernet over GRE (for ERSPAN and virtual private network creation)
• Full L3 support + NAT
• Priority-based QoS
• More management interfaces (IOS-like CLI, SNMP, NetFlow)
• 802.1x/RADIUS
• Support for hardware acceleration (VMDQ, switching chips on SR-IOV NICs)

The version available online is near the 1.0 (0.90.4), but it’s only available as source code.
The online documentation already explains how to use it with a XenServer 5.5 host.

Labels: Citrix, KVM, Xen

For a long time a number of contributors sponsored by Red Hat worked on a virtualization interface that could standardize the way hypervisors are managed, getting rid of the differences between vendors’ implementations.

The API is called libvirt and it’s around since early 2006.

Red Hat has a strong commitment on it, at the point that its imminent KVM-based virtualization offering is based on its, as announced in June 2008.
This is why the API is released under the GNU Lesser General Public License (LGPL) which allows the inclusion in any commercial product.

Through libvirt, a management platform running on Linux, Solaris, Mac OS or even Windows can already control both Xen, KVM, Sun VirtualBox, Parallels OpenVZ, QEMU, LXC and User Mode Linux (UML). But the best has yet to come.

The just released version 0.7.0 includes a number of remarkable new features, including support for the IBM POWER hypervisor and what seems a first attempt to support VMware ESX.

Of course this doesn’t mean that VMware will allow a product using libvirt to manage its flagship hypervisor without buying vCenter Server.
But for sure it means that in a near future Red Hat may be able to offer what Microsoft already offer with System Center Virtual Machine Manager (SCVMM): the capability to control multiple hypervisors through a single management console. And this may be extremely appealing for some of those customers that already purchased vCenter.

Labels: KVM, Red Hat, VMware, Xen

Red Hat is definitively preparing for the imminent launch of its new enterprise virtualization offering based on KVM and the Qumranet technologies acquired in September 2008.

A very important piece of the puzzle is how Windows guest operating systems will perform on the Red Hat implementation of KVM.
Most virtual machines on the planet runs Windows, so if Red Hat doesn’t shine here it will have nothing concrete to compete against VMware, Citrix and Microsoft hypervisors.

In mid-July the company gave a hint on how it plans to manage this aspect of its strategy: it released version 1.0 of its KVM para-virtualization drivers for Windows guest OSes under the open source GPLv2 license.

The set includes the network driver (kvmnet) and the device block driver (viostor) and both already are certified against the Microsoft Windows Hardware Quality Labs (WHQL).

Thanks to Linux-KVM.com for the news.

Labels: KVM, Red Hat, Releases

By now most virtualization.info readers should know that Red Hat plans to (finally) unveil its KVM-based virtualization offering on Sep. 1, at the Red Hat Summit 2009 in Chicago (and maybe at VMware VMworld 2009 as well).

The new product portfolio will include not one but two management solutions:

• Enterprise Virtualization Manager for Servers featuring Live Migration, High Availability, System Scheduler, Power Manager, Image manager, Snapshots, thin provisioning, monitoring and reporting.
• Enterprise Virtualization Manager for Desktops (the connection broker and management console SolidICE acquired from Qumranet in September 2008)

While the public knows how SolidICE looks like, nobody really saw the first management solution above, except the few lucky beta testers that Red Hat secretly selected before June.

The only public management solution for virtual infrastructures that Red Hat is working on is called virt-Manager.
The product is promising (supports Xen, KVM and QEMU virtual machines) but it’s in the work since September 2006, it still is at version 0.80 (released at the end of July) and doesn’t seem enterprise-ready at all:

Redhat’s virtual machine manager has come a long way and is starting to show some real usability. I’ve gone from using command line exclusively to now only using command line for testing purposes. There’s still a lot to do but it’s still only version 0.8 and it’s developing at a nice pace. Apart from the bug with creating vms using existing storage, I’ve had no real usability problems with it. I am, however, looking forward to a nicer main viewer.

Hopefully virt-manager is not the product that Red Hat plans to use to compete against VMware vCenter, Citrix XenCenter and Essential, Microsoft System Center Virtual Machine Manager and Oracle/Virtual Iron management platforms.

Labels: KVM, Qumranet, Red Hat

In December 2008 Virtual Bridges closed a major deal with IBM to bundle a Linux-friendly version of its Win4VDI connection broker (called VERDE) with Canonical Ubuntu Linux and the IBM Open Collaboration Client Solution (OCCS), which includes Lotus Symphony, Notes and other IBM products.

The deal was especially relevant because this bundle was designed to deliver a VDI solution based on the KVM virtualization platform that Ubuntu embeds. And IBM was the first major ISV to support its enterprise products inside KVM virtual machines.

Eight months later Virtual Bridges, IBM and Canonical are back with VERDE 2.0.

The first new thing in this release is the product strategy: Virtual Bridges completely replaces Win4VDI with VERDE, avoiding to market and sell two different versions of the same connection broker.

The second and most important news is related to a new key component of the package: a client-side virtualization platform.

The press announcement mentions the term client hypervisor, but in this case we are talking about a lightweight Linux distribution with KVM (which is not a hypervisor architecture).
Like over client hypervisors, this one requires Intel VT enabled so it won’t work on some laptops (courtesy of Sony and Intel).

The virtual desktop can be checked out and copied on the local KVM platform, allowing the mobile user to work in a so-called offline VDI mode.
At that point VERDE 2.0 uses a new Self-Managing Auto Replicating Technology (SMART) protocol to synchronize the local virtual desktop image with the primary one that resides inside the corporate virtual infrastructure.

Labels: IBM, KVM, Releases, VDI, Virtual Bridges

Oracle continues to stay mum about its integration plan for Oracle VM, Sun xVM Server and Virtual Iron hypervisors, but don’t hold anything when it’s time to talk about the new competitors.

Just two months ago the company dismissed the VMware virtual appliance initiative and its Marketplace, saying that it doesn’t contain anything but toy appliances.
One month later Oracle decided to clarify how the word co-opetition is not in its vocabulary, modifying the support policy to exclude every virtualization vendor that offer a hypervisor for x86/x64 architectures.

Today it’s time to hit Red Hat (and by some degrees Novell).
On its corporate blog last week Oracle highlighted its commitment to Xen and the open source:

…Oracle's Linux commitment began in 1998 with the first commercial database on Linux. Not only does Oracle run the whole business on Linux, but also run the base development on Linux for all our products. Today Oracle has over 9,000 developers working on Linux and provides Global Linux Support in over 100 countries…

The key point of this apparently candid post is about the quality of support that only Oracle can offer.
To support the statement Oracle points to another article about the reasons behind the launch of Oracle Unbreakable Linux:

Oracle Unbreakable Linux launched two years ago as a support program for existing Red Hat Enterprise Linux (RHEL) implementations or for new Oracle Enterprise Linux implementations. Oracle Unbreakable Linux program is about enterprise-class support that customers can't get (or is not available) from Red Hat.

…

Oracle brings the highest support quality, more value, and proven business practices to Linux support, including the following items Red Hat can't:

• 7500+ professionals providing 24x7, global support in over 145 countries
• Lifetime support policy (7+ years of general product support with the ability to extend to unlimited number of years)
• Premier backporting (Request backport of specific features eliminating pressure to upgrade with every update release)

…Due to dissatisfaction with Red Hat's quality of support as well as a desire to get more value, many users have switched from Red Hat Support to Oracle Unbreakable Linux Support…

The message is specifically directed to Red Hat because Red Hat is the company that promoted Xen for years and then decided a complete U turn by replacing the open source hypervisor with KVM.

Red Hat will (re)start competing with the other virtualization players in September when its new offering will become finally available.
And before any customer even think about jumping on the KVM bandwagon, Oracle wants to make sure that everybody knows how much better they are at support.

Labels: KVM, Oracle, Red Hat, Xen

We still don’t know anything about the new Red Hat virtualization portfolio based on KVM and Qumranet VDI technologies, despite a (claimed) oversubscribed beta program that nobody was able to access.

A key piece of this offering of course is Red Hat Enterprise Linux which was expected to drop Xen in favor of KVM as the default virtualization engine.
The release notes of the just announced beta of RHEL 5.4 confirm this.

The KVM version included in RHEL 5.4 will support RHEL 3.x, 4.x and 5.x guest OSes along with Windows XP, Server 2003 and Server 2008.
All of them are supported in 32 and 64bits, and each OS will be able to run without installing any para-virtualized (PV) driver, despite these components will be available as part of the distribution. 
No mention of Windows Server 2008 R2 and Windows 7 which will hit the RTM next week.

The Red Hat customers using Xen are not entirely left in the cold but their life will be much harder now:

Xen based virtualization is fully supported. However, Xen-based virtualization requires a different version of the kernel to function. The KVM hypervisor can only be used with the regular (non-Xen) kernel.

While Xen and KVM may be installed on the same system, the default networking configuration for these are different. Users are strongly recommended to use one hypervisor at a time.

This beta shouldn’t run more than a couple of months as Red Hat plans to release the new products on September 1, according to LeMagIT.

Labels: KVM, Red Hat

Ten days ago Red Hat announced that its new, much awaited, virtualization offering based on KVM was in beta and that the beta program was oversubscribed.

The reality is that, as far as we know, Red Hat never announced the beta program or the details of its implementation of the Qumranet technology (acquired in September 2008), and never gave the opportunity to sign for it to the general public.
Still today there not a single bit of information about what Red Hat did in one year and a half after dropping Xen in favor of KVM.

Red Hat will take another two months to finally tell the world as LeMagIT revealed earlier today: the general availability of the new virtualization platform is in fact planned for September 1, 2009, which means during the VMware VMworld 2009.
Too bad that this year VMware is not particularly happy to have competitors showing their solutions on the exhibit floor.

Labels: KVM, Red Hat

One of the oldest (and hottest) topic in the history of modern virtualization is if the Xen open source hypervisor can be integrated into the Linux kernel or not.

XenSource tried to achieve the goal for years (while VMware did its best to avoid it), but in December 2006 Linus Torvalds announced the decision to include another virtualization platform in place of Xen: KVM.

KVM was developed and maintained by the startup Qumranet, acquired by Red Hat in September 2008, and at that time was just 6 months old, much less mature than Xen.
Despite that and because of its architecture (at least this is the official reason), KVM has been included in the kernel since version 2.6.20 and Xen is not.

After this and after the acquisition of XenSource by Citrix, the idea of Xen inside Linux seemed definitively archived. But the community is still debating about the topic.

Torvalds’ comment on the idea is lapidary:

…If Xen was a single driver thing, we wouldn't have this discussion. But as is, Xen craps all over OTHER PEOPLES CODE. When those people then aren't interested in Xen, why is anybody surprised that people aren't excited?

Thanks to c0t0d0s0 for the news.

Labels: KVM, Xen

In the past months virtualization.info highlighted several times how Cisco is silently using KVM as an alternative virtualization platform to VMware.
We always wondered why, considering the investment that Cisco made on VMware.

Now, finally we have an answer to give: Cisco invested in Qumranet too.

Qumranet is the startup that developed and maintained KVM up to the moment it was acquired by Red Hat.
And that’s why Red Hat had a minor but very relevant position during the launch of the Cisco Unified Computing System (UCS) despite its virtualization offering is pretty weak now.

The fact that Cisco invested in Qumranet is not widely known and we had to admit that even virtualization.info overlooked this key information so far.
How the investment links Cisco to Red Hat is not clear but it’s easy to guess that the upcoming Red Hat new virtualization portfolio based on KVM will have an early chance to be bundled with UCS.

Now VirtualLogix, the mobile virtualization startup where Cisco invested along with Intel, is the next most interesting company to watch.

Labels: Cisco, KVM, Red Hat

The article appeared on WorksWithU may be the first documented case of a company deciding to adopt KVM instead of VMware or Citrix hypervisors.

To be fair the scenario described by the author is at the lowest limit of the SMB range: a mere two virtual machines for a single virtualization host.
Yet, it’s interesting to read the reasons behind the choice to go for KVM on Ubuntu rather than VMware ESX or Citrix XenServer:

1. Cost.  Although our virtualization requirements are minimal–we need to run only two guest servers on a single host machine–VMware would have cost an astonishing amount of money.  With features like VMware motion factored in, we were looking at a huge hit to the budget–and it didn’t help that VMware charges per CPU, not machine, regardless of whether all CPUs will actually be dedicated to virtualization.  KVM is totally free, in both senses of the word, and offers functionality equivalent to VMware motion.
2. Ease of deployment.  Installing KVM on Ubuntu 8.04 is as simple as an apt-get.  ESX server is also easy enough to install, but having to deal with licensing adds another layer of complication that we’d prefer not to face.  KVM, of course, requires no license.
3. Speed.  Although I don’t have hard numbers, KVM-based virtual machines definitely ‘felt’ more responsive than those running on VMware.  Our experience seemed to confirm Red Hat’s claim last fall that KVM can support five VMs for every three running on VMware on the same piece of hardware.  It was also troubling that ESX server wasted upwards of 500 megabytes of memory–without any VMs running–on system overhead, while an Ubuntu server is a considerably more efficient host.
4. Management.  Apparently it never occured to VMware that systems administrators might be running Linux on their workstations.  As a result, Windows is the only platform on which VMware’s graphical management infrastructure is supported.  A Linux CLI client is available, but I’d like more options than that.  KVM, in contrast, can be managed via the command line, via graphical interfaces (running either on the local machine or forwarded over ssh to a remote workstation) or through the Enomaly web interface.

Be sure to read the whole article to fully understand the process that led to this choice.

Labels: KVM

Earlier this week SDTimes published a brief coverage of the Linux Foundation’s Collaboration Summit, which was held in San Francisco last week.

A very brief note in the article highlights a remarkable information:

For the virtualization crowd, Zemlin [Jim Zemlin, Executive Director at the Linux Foundation] said that, moving forward, the Linux Foundation is encouraging vendors and developers to standardize on KVM, not Xen.

If true this may be the confirmation that the Citrix acquisition of XenServer has compromised the relation with the open source community, despite Citrix is giving back.

It’s interesting to note that the Red Hat acquisition of Qumranet, which developed and maintains KVM, didn’t have the same impact.

Labels: KVM, Xen

In 2006 the ConVirt team started an ambitious project: develop an open source, multi-host management console for Xen.
Initially called XenMan, the tool was then renamed ConVirt and its roadmap was enriched with several highly desirable features.

Three years later the ConVirt team morphs in a company called Convirture, and ConVirt, still an open source product, finally reached version 1.0 with a notable number of features:

• Support for Xen and KVM
• Support for multi-host virtual infrastructures
• Support for virtual machines snapshot, live migration, backup and decommission
• Support for VMs templates and virtual appliances
• Support for storage usage

The product is available free of charge here.

Labels: Convirture, KVM, Xen

With the acquisition of Qumranet in September 2008 Red Hat raised a lot of interest.
The customers that trusted the company when it was promoting its Xen implementation all over the place want to know what will happen to them.
The potential customers that are interested in an open source hypervisor but  dislike the idea of Citrix indirectly controlling how Xen, want to know how serious Red Hat is about KVM.

Last week, finally, the company announced its commitment:

• Next versions of Enterprise Linux (RHEL) will feature KVM.
  The exiting versions featuring Xen will be supported for the full lifetime of RHEL 5.
• Red Hat will release a brand new Enterprise Virtualization Hypervisor (a minimal version of RHEL only supporting KVM and a selected number of drivers).
• Red Hat will release a brand new Enterprise Virtualization Manager for Servers featuring Live Migration, High Availability, System Scheduler, Power Manager, Image manager, Snapshots, thin provisioning, monitoring and reporting.
  This product will be able to manage both RHEL and RHEVH.
• Red hat will rebrand the Qumranet connection broker and management console SolidICE as Enterprise Virtualization Manager for Desktops.

All the products above will be delivered sequentially over the next three to 18 months, with delivery dates beginning in the middle of 2009.

Of course the biggest concern about KVM is that ISVs are not supporting it. And because something like 95% of the existing virtual machines run Windows guest operating systems, a partnership with Microsoft is fundamental.

This is why just two weeks ago Red Hat joined the Microsoft Server Virtualization Validation Program (SVVP).
Without it, the company could sell its new virtualization platform only to those customers that run Linux guest OSes, and that is a small niche compared to the huge virtualization market.

Yet, much depends on how the new platform will be priced (something that the company didn’t announce yet).
The free XenServer can damage Red hat much more than VMware.

Labels: KVM, Qumranet, Red Hat

One of the biggest challenge when adopting a new virtualization platform is securing the ISVs support.
Without it moving from the market leader to a more innovative or cheaper solution is a risky business.

It’s the case of KVM, the open source virtualization platform that is part of the Linux Kernel since version 2.6.20 and that is attracting a large number of developers (away from Xen, we were told).

KVM may be very cool, and the fact that Red Hat acquired its maintainer, the startup Qumranet, certainly ignites high hopes for the platform.
But the reality is that, at today, KVM is still too young to feature the ISVs support that VMware, Citrix or Microsoft can offer.

Excluding IBM, which just started to its Lotus Notes, Symphony and a bunch of other applications on the Virtual Bridges implementation of KVM, no other major IT vendor is officially endorsing KVM.

As often happens, Microsoft is the key to change this situation: it’s now more than clear that virtualization is being used across the globe to virtualize and consolidate in large majority Windows boxes.
If Microsoft officially supports Windows in a KVM virtual machine then the other ISVs will follow, and the customers can start adopting the solution with confidence.

With much surprise it’s possible that the unlikely event already happened.

As most readers remember Cisco is using a mysterious virtualization platform inside its Wide Area Application Services (WAAS) 4.1 appliance. 
Despite the company stays mum about the VMM used inside WAAS, in the past months virtualization.info received a remarkable number of confirmations from different sources that the appliance is almost certainly powered by KVM.

Now Cisco is selling WAAS 4.1 and its new virtualization capabilities certifying its use as a platform where the core Microsoft services (part of Windows) can be consolidated.
To do so Cisco joined the Microsoft Server Virtualization Validation Program (SVVP) in August 2008. But the SVVP program doesn’t include Windows support.

This means that, through the SVVP program, Microsoft is supporting its core services inside the Cisco WAAS 4.1 virtual machines (which are, we are almost sure, KVM virtual machines), but not the operating system itself.
So, who is supporting Windows exactly? 

The only two possibilities are that Cisco is in charge for the OS support, and it seems unlikely, or it’s Microsoft that is making an exception and is supporting its operating system inside KVM, at least the Cisco implementation of it.

The reason why all these details are unclear, and Cisco customers should investigate before buying WAAS 4.1, is the business relationship of the two software giants: Cisco is definitively in bed with VMware and it’s preparing to make a major announcement, while Microsoft is totally in love with Citrix, which probably isn’t too happy to know that KVM is being supported so quickly.

If it will emerge that Microsoft officially supports Windows on KVM, this may further boost the Red Hat chances to attract customers with its upcoming new virtualization platform.

Labels: Alliances, Cisco, KVM, Microsoft

Yesterday the open source virtualization KVM reached another milestone: the build 83 now includes support for the AMD Input Output Memory Management Unit (IOMMU) technology.

An IOMMU makes I/O virtualization more efficient by allowing VMMs to directly assign real devices to guest operating systems. It's not possible for a VMM to emulate the translation and protection functions of an IOMMU, because the VMM can't get between kernel-mode drivers running on the guest OS and the underlying hardware. So, in the absence of an IOMMU, VMMs instead present an emulated device to the guest OS. The VMM then translates the guest's requests, ultimately, into requests to the real driver running down on the host OS or on the hypervisor.

Less than one month ago the version of KVM that is included in the Linux Kernel 2.6.28 introduced the support for Intel IOMMU called VT-d.

Now both implementations are supported and the AMD patch will soon reach the mainstream kernel (probably 2.6.29).

Labels: KVM, Red Hat

The just released Kernel 2.6.28 includes more than 104 patches for the virtualization engine KVM, included in Linux since 2.6.20.

One of those patches is specially important as it allows the mapping of physical PCI device to a specific virtual machine through the Intel Virtualization for Directed I/O (VT-d) technology.

Intel introduced VT-d in early Q1 2006 but so far only Novell and Oracle supported it in their Xen implementations (as the virtualization.info Buyer’s Guide highlights).

The PCI direct access grants higher performance but lower flexibility in a virtual infrastructure: for instance a VM can only map as many devices as are physically present in the platform.
Nonetheless it’s a critical step to bring high-performance virtualization on consumer equipment (something often called client hypervisor) like laptops.

Red Hat, which is the main contributor of KVM after the acquisition of Qumranet, has all the interest to achieve the goal as the company executives clarified last month.

Meanwhile KVM continues to get new features: in its last build, KVM-82, the platform allows users to nested virtual machines when running on AMD CPUs.

Labels: KVM, Red Hat

Virtual Bridges is a company founded at the end of 2006, that always offered commercial flavors of QEMU for Linux, BSD and Solaris platforms.
After KVM made its appearance Virtual Bridges started to implement support for it on its products for Linux. Where KVM is not available KQEUM is automatically used.

In August it significantly extended its scope by releasing its first VDI connection broker for KVM: Win4VDI.

Compared to other products in this space, Win4VDI doesn’t connect the user to the actual guest OS, but rather to the underlying host. From there the user session is started.
In this way Virtual Bridges can leverage the authentication methods and profiles that the host is using.

The choice has been courageous.
Even if KVM is part of the Linux kernel and even if its maintainer, Qumranet, has been acquired by Red Hat, the spread of a new virtualization platform must surpass a huge obstacle: the ISVs support.
And at this point no ISV formally supported its applications inside KVM virtual machines.

Despite that, Virtual Bridges has been rewarded as IBM just closed an agreement with them to resell a bundle made by:

• Canonical Ubuntu Linux (which is adopting KVM in place of Xen since February)
• Virtual Bridges VERDE (a subset of WIN4VDI that only supports Linux guest OSes)
• IBM Lotus Symphony, Lotus Notes and the other Lotus applications (dubbed Open Collaboration Client Solution)

The whole package is available at $49 per concurrent user.

So the move is remarkable because IBM is the first big player supporting (and actively selling) KVM-based virtual infrastructures. But it’s also remarkable considering how heavily IBM invested in Xen in the past.
After the acquisition of XenSource by Citrix, a number of entities behind the open source hypervisor development were reportedly unhappy and decided to shift to KVM. And this seems the first concrete step that demonstrate how unhappy IBM is about Xen.

True or not, looking at what IBM just did, we can have an idea of what Red Hat could do.
The difference between the two vendors is that Red Hat is in a much better position to sell an out of the box VDI package: it controls the operating system, it (indirectly) controls the virtualization platform, it controls the connection broker, and its role in the industry as OS provider certainly gives much influence on what ISV applications will be supported on top.

Now, considering that, besides Qumranet, Virtual Bridges is currently the only other vendor offering a connection broker for KVM and that its experience has to be somewhat limited, the real question is: why IBM didn’t do this with Red Hat instead of Ubuntu and Virtual Bridges?

Labels: Alliances, IBM, KVM, Virtual Bridges

Two weeks ago the Red Hat CEO hinted at his upcoming virtualization strategy but was careful enough to not say when KVM would be integrated into the company enterprise distribution.

Now CBR reports that Red Hat may be ready by the first half of 2009.

By that time the company will completely replace Xen with KVM in Red Hat Enterprise Linux (RHEL), but will continue to offer support for the former virtualization plaform for another seven years.

The company Vice President for EMEA, Werner Knoblich, insisted that KVM is better than Xen (or VMware ESX) when talking about large-scale deployments (thousands of virtual machines) because the virtualization engine fully leverages the Linux kernel capabilities while the bare-metal hypervisors cannot.

True or not, such comment highlights how Red Hat is looking at KVM for cloud computing much more than for server consolidation.

Labels: KVM, Red Hat

Since months now a serious number of companies and open source contributors is looking at Red Hat to understand its new virtualization strategy.

The company took a major step in June when it thrown out of window years of efforts on Xen to fully replace it with KVM.
Just two months after, Red Hat acquired Qumranet, the company that started KVM, that maintains it, that managed to inject it into the Linux kernel, and that sells a very interesting VDI solution.

What Red Hat wants to do now with KVM and Qumranet (somebody hopes that their highly performing VDI protocol SPICE will be open sourced) is critical to understand what chances has Linux to impose itself as a valuable virtualization platforms against the popular hypervisors ESX (offered by VMware) and Xen (offered by every other vendor except Microsoft).

Some hints of the new strategy surface in a recent interview that InformationWeek arranged with Jim Whitehurst, the Red Hat CEO:

Q: What is Red Hat's strategy with virtualization?

A: …We'll be offering both server and desktop virtualization. The first use-case of server consolidation is just the tip of the iceberg. There's a long-term use for grids of servers running large numbers of desktops. We plan to be the leading virtualization vendor based on the server operating system…

Q: What about cloud computing?

A: Clouds will run Linux.

It’s clear that Red Hat sees in KVM a huge opportunity to differentiate from Citrix, Virtual Iron, Oracle and its worst competitor Novell.
Now the company has to to execute in a much better way that how it did with Xen.

Labels: KVM, Qumranet, Red Hat

At the end of last week AMD announced a breakthrough achievement: migrating a running virtual machine from a virtualization platform to another, each running different CPU brands.

Despite many progresses in this area (AMD-V Extended Migration and Intel Flex Migration), so far the only thing possible was to live migrate VMs between different CPU families of the same vendor.
AMD and Intel never cooperated as much to cross such boundary and in one case we are pretty sure that an Intel executive said that the thing would be unlikely to happen.

Now AMD has found a way to mask the CPU information and operate the migration from an Intel Xeon DP Quad Core E5420 to its own forthcoming 45nm Quad-Core Opteron.
To achieve the goal the company worked together with Red Hat so everybody would expect that the migration happened through Xen hypervisors. It’s not the case.

Red Hat fully embraced KVM as replacement of Xen in June and this is the virtualization platform that was used for the demo:

Labels: AMD, KVM, Red Hat

It may be just a coincidence but the just released Fedora 10 doesn’t include Xen. And this happens just a month after Red Hat unveiled its new virtualization strategy, adopting KVM and acquiring the startup that maintains it: Qumranet.

The reason behind this unexpected drop is explained in the official project newsletter:

No Dom0 Support in Fedora 10

…”There is pretty much zero chance that Fedora 10 will include a Xen Dom0 host. While upstream Xen developers are making good progress on porting Dom0 to paravirt_ops, there is simply too little time for this to be ready for Fedora 10. So if you need to use Fedora 10 as a host, then KVM is your only viable option at this time. If you can wait for Fedora 11 (or use RHEL-5 / CentOS-5) then Xen may be an option for you." …

The distribution lifecycle implies a new major release every six months on average.
This means that Fedora users will have a long time to explore the opportunities that KVM offers. And in the meanwhile Red Hat will be able to further attract them with some new interesting products based on Qumranet technology.

Thanks to Mike DiPetrillo for the news.

Labels: KVM, Qumranet, Red Hat, Xen

Now that Red Hat acquired Qumranet, the maintainer of KVM, the interest around the open source virtualization engine included into the Linux kernel is raising over the top.

Those customers considering the product should be happy to know that the newest version, KVM 75, introduces the memory ballooning.

The ballooning is the most common approach to achieve memory overcommit, a feature available only in VMware ESX at today.

While the real value of ballooning has been questioned several times, it’s really notable to see that the youngest newcomer in the virtualization arena is already stacking up the right features to compete with the most mature products.

Of course KVM is too new and its diffusion too limited to really prove its reliability against the well-known competitors. With or without cutting-edge features, Red Hat will have to work a lot to build confidence in the new platform.

Labels: KVM

Red Hat just announced that it acquired the VDI startup Qumranet for $107M.

The startup left the stealth mode in September 2007 and has just 60 employees, all of them will keep their work as Red Hat employees.

The move is critical and has a major impact on many aspects of the virtualization world.

The Qumranet strategy
Qumranet offers an interesting VDI solution made by a  management console, a connection broker and a new remote desktop protocol.
But more than that, Qumranet maintains KVM, the new virtualization platform that has been implemented in the Linux kernel after just six months of development.

KVM allows any Linux box to become a virtualization platform, and KVM is the only company at the moment able to offer a VDI solution for KVM.
This means that customers looking for cheap, large-scale virtual desktop infrastructures have to buy the Qumranet solution.

The Red Hat strategy
In the last few years, despite a number of announcements, Red Hat didn’t demonstrate a neat strategy about virtualization. But in June the company officially declared its intention to move from Xen to KVM.

Just two months after, a Red Hat executive formally revealed the company interest in VDI. At that point it was easy to forecast a special deal with Qumranet.

(it seems that when virtualization.info presumed the acquisition, the deal was already signed)

Buying Qumranet Red Hat just solved a number of problems:

• obtaining direct control on the development of a virtualization engine (something that the company was never able to do with Xen because of XenSource, and even less after XenSource was acquired by Citrix)
• obtaining a platform which is ready for virtualization ubiquity (KVM is flexible enough to be deployed on servers, desktops, embedded devices and anywhere Linux can fit)
• obtaining a strong position in the growing VDI market
• differentiating its virtualization offering from the competitor Novell (both are currently adopting Xen)
• enforcing its position of open source leader while the competitor Novell is seen as a suspicious Microsoft ally

Both Citrix and Red Hat acquired a virtualization company that develop and control an open source virtual machine monitor (VMM). But there’s a major difference between the two companies.

Citrix never had an involvement in the open source world, and despite the culture introduced by Simon Crosby and his staff (and their tireless efforts), the community has a hard time in recognizing Citrix as a company that can give back. When talking about virtualization Citrix is first and foremost seen as the best Microsoft ally against VMware.
Red Hat instead is a beloved, open source paladin. The company made some mistakes in the past, but its effort in supporting Fedora still makes it a leader in the Linux world.
This difference is now specially important: some entities currently contributing to the Xen development may find much more interesting to work with Red Hat rather than with Citrix (think about IBM).

At the same time some firms heavily relying on Red Hat (think about Oracle), may be in deep trouble now that the company is definitively replacing Xen with KVM.
These entities may need to look at Novell now, or start working on their own implementations.

In any case the acquisition of Qumranet is a major achievement for Red Hat, which has now a unique opportunity to become a real virtualization leader.

Labels: KVM

Over the last few months the speculations around Cisco entering the server virtualization market were supported just by rumors. Below a list of news related to the topic:

• March 2008 - Cisco puts KVM in its IOS (the company denied that this ever happened)
• April 2008 - Cisco or IBM may acquire Citrix
• June 2008 - Cisco doesn’t want to buy VMware
• July 2008 - Cisco may be interested in hypervisors more than what it seems

Rumors or not, today something concrete happened: Cisco signed as member of the Microsoft Server Virtualization Validation Program.

The program simplifies the relationship between Microsoft and the other members so that the software giant can easily support its products on participants’ hypervisors:

The Server Virtualization Validation Program (SVVP) is open to any vendor who delivers a virtualization machine solution that hosts Windows Server 2008, Windows 2000 Server Service Pack 4 and Windows Server 2003 Service Pack 2 and subsequent service packs. The virtualization solution can either be hypervisor-based or a hosted solution.

Cisco doesn’t have any hypervisor so there’s no reason to adhere this program.

At this point it’s hard to believe that the company will stay away of the server virtualization market for much more.

Labels: Cisco, KVM

ComputerWorld just published an interview with the Executive Vice President of Product and Technologies at Red Hat, Paul Cormier, where some interesting details about the company strategy emerge.

First of all Red Hat predicts that within 2 years 90% of the existing servers will run a virtualization platform.

Second, the company sees virtualization as the first step towards a cloud computing world, following VMware which started to spread the idea earlier this year (virtualization.info wrote about this at the beginning of 2006).

Third and most important, Red Hat sees the VDI as the next big thing.

Why this point is specially interesting? Red Hat is only the last of a long series of vendors that demonstrated interest for hosted desktops. But compared to the others the Red Hat vision deserves a little more attention because this is the only company that so far is dropping Xen in favor of KVM.

While open source, KVM is maintained and supported by Qumranet, a US startup (see the virtualization.info coverage here) that offers a commercial VDI solution.

Now let’s consider the equation:

• KVM is now embedded into the Linux kernel
• Red Hat will offer KVM as the virtualization platform of choice
• Red Hat sees VDI as the next big thing
• Qumranet offers the only VDI solution based on KVM
• Qumranet maintains KVM

The question is: how important will become Qumranet for Red Hat in the near future?

Labels: KVM

The open source virtualization platform that made its way into the kernel is now at version 70.

This new build introduces a bunch of bugfixes and performance improvements on virtual network and simmetric multi-processing.

Download it here.

Labels: KVM

No matter what the official version says, the Red Hat virtualization strategy has always been problematic.

It started in December 2004, when the company revealed the plan to adopt the open source hypervisor Xen as virtualization engine for its enterprise operating system Red Hat Enterprise Linux (RHEL) 4.0.
Unfortunately such plan didn't become reality earlier than two years later, when Red Hat finally released RHEL 5.0 with Xen in March 2007.

In mid 2006, being already in late, the company decided to play an interesting game: it declared Xen immature and accused its main competitor, Novell, for being irresponsible in integrating Xen in its enterprise distribution (SUSE Linux).
Probably to support such claims, Red Hat did its best to ship RHEL 5 with a minimal, under any enterprise standard, GUI for Xen: Virtual Machine Manager.

After that, a series of dramatic events happened.
First Microsoft signed a series of alliances with the some of the key Xen contributors: with XenSource, with Novell and with Virtual Iron. Then Citrix acquired XenSource in August 2007.

So, in less than one year Red Hat lost much of its capability to influence the Xen development despite it contributes to the open source project since the early beginning.
It was easy to guess that the company would quickly evaluate an alternative. And the only valuable alternative for Red Hat was KVM, the young virtualization platform that was included in the Linux kernel after just six months of development.

KVM offers a lot of advantages to Red Hat.
For example it ships as an official kernel module, avoiding major investments in maintaining it.
For example Novell is not using it and it's unlikely to do so for a while (it has too much to do with Microsoft and Citrix right now).
For example it brings a new powerful allied, IBM, which clearly has no more interests in heavily investing in Xen.
For example its inclusion in the kernel makes very hard for Microsoft, its allied, or any other uncomfortable player to influence its development.

So yesterday Red Hat did the much expected move to announce the adoption of KVM.

The company will offer a new lightweight distribution which integrates KVM, selling it as the Red Hat virtualization platform (Embedded Linux Hypervisor).
Additionally, the company will offer a new enterprise-wide management solution called oVirt, which is based on the standardized libvirt APIs and is designed to scale up to thousands of virtual machines.

What happens to Xen now? What happens to the Red Hat investment on it?
The official press announcement don't say it explicitly but the choice of words lets clearly understand that Xen is the past and KVM is the future.
It's unknown if Red Hat will continue to support Xen or what will happen to those enterprise customers that adopted RHEL 5 to use the hypervisor (mostly because Xen and KVM virtual machines are incompatible).

In any case it's unlikely that Red Hat will drop Xen tomorrow: the company sits in the Xen advisory board and its support policy implies that any distribution must be supported for seven years.
Red Hat may want to spend this time convincing its customers that KVM is a better virtualization engine: Virtual Machine Manager, for example, has been already re-categorized as a desktop user interface for managing virtual machines.

Both the Embedded Linux Hypervisor and the oVirt management console are already available in beta here.

Red Hat didn't mention when both products will be available as RTM.

Labels: KVM

Fifteen days after the previous release the Qumranet developers finish KVM version 69.
Breaking an impressive series of new architectures supported, this build is only for bug fixing.

Download it here.

Labels: KVM

Proxmox Server Solution GmbH is building an open source management tool for virtualization platforms: Proxmox Virtual Environment.

The product, still in beta (0.9), has some interesting capabilities:

• Bare-metal installation (based on Debian Etch 64bit, which includes the supported virtualization platforms)
• Support for KVM and OpenVZ
• Web-based management console (AJAX based)
• Host-level backup (OpenVZ containers only)
• Support for host-level clustering (Proxmox VE Cluster)

Additionally, the developer is implementing also a live migration capability.

Since both KVM and OpenVZ don't have a GUI it's worth a try even if it's still in early development.

Download it here.

Labels: KVM

Qumranet, the US startup that supports the development of KVM (the hypervisor included in the Linux kernel), announced its first product in September 2007: Solid ICE, an all-in-one VDI solution (the virtualization server based on KVM, the management console, the connection broker and the accelerated remote desktop protocol) for KVM virtual machines.

The company is finally releasing its product which immediately jumps to version 4.1.

This first version has some interesting features to offer, like:

• Support for Windows 2000 and XP virtual machines
• Self-service web portal
• Active Directory authentication support
• Virtual machines pools provisioning
• Fault-tolerant connection broker
• Optimized remote desktop protocol (SPICE) with support for bi-directional audio/video (+30 Fps)

The price per concurrent virtual machine is to $200.

Download a trial here.

The virtualization.info Virtualization Industry Roadmap has been updated accordingly.

Labels: KVM

Just one week after version 67, KVM releases version 68.

The new build introduces support for the upcoming Intel nested pages technology dubbed Extended Page Tables (EPT).
The EPT technology should be available only in 2009/2010 when Intel will introduce a completely new architecture in its upcoming Nehalem CPU.

Despite all virtualization vendors are working to support Intel EPT in their hypervisors, KVM is probably the first platform to offer it today.

Download KVM 68 here.

Labels: KVM

It seems that KVM developers are able to release a new version of their hypervisor every ten days or so, and every time with a new architecture supported.

After the IBM s390 (KVM 65) and the Intel Itanium (KVM 66), it's now the turn of the AMMC ppc 44x embedded processors (used by companies like QLogic, NetClarity, Accusys, etc. for network devices like fibre channel HBAs).

Download it here.

Labels: KVM

In just ten days the KVM developement team moved from version 65, which introduced support for IBM s390, to version 66, which introduces support for the Intel Itanium architecture along with a long list of bug fixes.

Download it here.

Labels: KVM

Today Avi Kivity, the KVM lead developer, published a pretty bold statement on his personal blog: para-virtualization is dead.

While he's saving the I/O para-virtualization drivers (Qumranet, which supports KVM, just released its first para-virtualized network drivers), he also thinks that upcoming nested paging technologies from AMD (the Rapid Virtualization Indexing or RVI) and Intel (the Extended Page Tables or EPT) will provide such performance boost that modifying the guest OS kernel to make it virtualization-aware will become a useless practice.

KVM support for AMD RVI is expected along with Linux kernel 2.6.26, while the support for Intel EPT will not come earlier than 2009-2010, when the CPU vendor will release its next generation processor Nehalem.

The position of Kivity is interesting per se but becomes even more interesting when compared to the recent position of VMware, which is now fully endorsing para-virtualization.

Who knows what Citrix, which acquired the most active para-virtualization promoter XenSource, thinks about this theory.

Labels: KVM

KVM, the The open source bare-metal hypervisor included in the Linux kernel, reaches version 65 landing on the IBM s390 mainframe architecture.

Anthony Liguori, one of its contributors (and once Xen developer), reveals that this first version for s390 already supports 64-way virtual machines.

Other major features are in the hypervisor roadmap.

Some of them will provide KVM memory management capabilities that are famous in VMware ESX: the memory balooning (also available in Citrix XenServer) and the page sharing.
Both should appear in this Q2 2008 (memory ballooning in particular is expected with the upcoming Linux kernel 2.6.25.

Download KVM-65 here if you don't want to wait for the official inclusion in the next kernel build.

Labels: KVM

Almost silently the US startup Qumranet, famous for supporting the KVM development, accomplished a remarkable achievement: releasing para-virtualization network drivers for Windows guest OSes.

The drivers are available for Windows 2000 and XP (signed versions) as well as for Windows 2003 (unsigned version), are for 32bit systems only, and can be used with KVM-61 or later.

Using dedicated para-virtualization drivers, guest OSes can often improve performance as also VMware confirmed.

Haydn Solomon, the Qumranet developer who published the news, provides a step-by-step installation guide just in case it's needed.

Download the drivers here.

Labels: KVM

KVM is the youngest virtualization platform on the scene but the strategic positions it occupied so far beat any competitor.

First, it was included in the Linux kernel after just six months from the launch, and thus today any distribution sporting kernel 2.6.20 or higher can offer it out-of-the-box.

Then it was chosen over Xen for one of the most popular consumer distribution ever: Ubuntu.

And now Cisco includes it in the new IOS-XE, the Linux-powered operating system for its highest-end router: the ASR 1000.

The Aggregation Services Router (ASR) 1000 is the first Cisco router that replaces its traditional in-house developed IOS with Linux and targets service providers and biggest enterprises.
Among the exclusive features offered by this new equipment there is the operating system redundancy, achieved without any hardware module: a first in the networking industry.

How Cisco is able to provide a redundant IOS image? Through KVM virtual machines as Information Week reports.
At this point is still unknown which version of Linux kernel is used for IOS-XE and which version of KVM as well, but the company must be absolutely sure of its reliability to use it in such high-end product.

It's interesting that fact that Cisco chosen KVM for this task, while it's very busy with VMware, investing in VMW, occupying its keynotes, and maybe (it's still an unconfirmed news) releasing a software switch for ESX Server.

Update: This news is bigger than expected, despite few online magazines reported it.

virtualization.info contacted KVM Lead Developer about the story and received a No Comment answer. Other inquiries to Cisco and Qumranet returned no answers at all so far.

Second update: virtualization.info received a second No Comment from Qumranet.

At this point being a Cisco ASR 1000 potential customer or a VMware shareholder we'd like to have some detailed explainations about what Cisco is exactly doing with KVM and why there's an embargo on the story.

Third update: Colin McNamara provides some more details about the the ASR 1000 architecture and the use of KVM inside it.

Labels: Cisco, KVM

It was just a matter of time before one Linux distribution, for consumer or enterprise use, would switch its virtualization engine from Xen to KVM.

The inclusion of KVM in the Linux kernel, along with the controversy influence that Microsoft first and Citrix then appy to the Xen project, led to this.

The suspects that Citrix is not fully commited to the open source project after acquiring XenSouce may be a further factor.

Canonical's decision to drop Xen in favor of KVM may remain an isolated case, but anyway Ubuntu is one of the most popular distro among consumers (DistroWatch reports that it's the most downloaded of 2007).

The influential example may be followed by other distributions like Fedora, which already implements KVM side by side with Xen.
It's worth to remember in fact that Fedora is controlled by Red Hat, which is in hasty competition with Novell. But while Xen is implemented by both companies, only Novell has a tight partnership with Microsoft, which has a tight partnership with Citrix, which is now able to influece the Xen development.

It's easy to imagine how likely is a Red Hat departure from the Xen bandwagon.

Labels: KVM

A new patch on the work could drastically improve performances of Windows guest operating systems for KVM hypervisor, included in Linux kernel since version 2.6.20. This is what Anthony Liguori, IBM software engineer and long time Xen hacker, reports in his personal blog:

...

At this past KVM Forum, Ben Serebin , from AMD, shared an interesting observation. Windows guests only access the TPR with instructions that are at least 5 bytes. The significance of 5 bytes is that that happens to be the size of an absolute call on the x86. This means that you can replace any of the TPR access instructions with an absolute call without the need to do fancy dynamic translation. If you're very clever about hiding routines within the BIOS (it turns out, Windows always has a valid virtual mapping to the BIOS), you can actually rewrite TPR access instruction to instead be calls to functions, that you provide, that access the TPR in a more efficient way.

Avi Kivity posted an implementation of this to KVM recently. The results are quite dramatic. Windows XP installs are at least twice as fast--perhaps even faster. The very latest Intel processors have a hardware feature that ends up with the same result but the nice thing about a purely software approach is that it will work with older processors.

This code hasn't made it's way into a KVM release yet as it needs a bit more testing and clean-up. I suspect we won't see it in a release for a couple more weeks but once it's there, you can reenable ACPI in your Windows guests and enjoy good performance...

Read the whole post at the source.

Labels: KVM

Linus Torvalds just announced availability of new Linux kernel 2.6.23.

It brings several improvements in the hardware virtualization space with the introduction of Xen and lguest support.

Kernel Newbies provides a good summary:

Lguest host support (CONFIG_LGUEST)can be compiled as a module (lg.ko). This is the host support - one you load it, your kernel will be able to run virtualized lguest guests. But kernel guests need to compile lguest guest support in order to be able to run under the lguest host. The configuration variable that enables the guest support is CONFIG_LGUEST_GUEST - but that option will be enabled automatically once you set CONFIG_LGUEST to 'y' or 'm'. This means that a kernel compiled with lguest host support does also get lguest guest support. In other words, you can use the same kernel you use to be a host as guest kernel. In order to load and run new guests, you need a loader userspace program.

Part of Xen has been merged. The support included in 2.6.23 will allow the kernel to boot in a paravirtualized environment under the Xen hypervisor. But support for the hypervisor is not included - this is only guest support, no dom0, no suspend/resume, no ballooning. It's based in the paravirt_ops infrastructure.

Anyway the most important news is support for virtual SMP inside KVM guest operating systems:

• Enable guest smp
• Implement rdmsr and wrmsr. This allows smp Windows to boot
• i386: Allow KVM on i386 nonpae

More improvements are expected also in the OS virtualization space: at LinuxWorld 2007 Andrew Morton said kernel development will focus much on these technologies during the next 2 years.

Labels: KVM, Xen

After almost two years in stealth mode, one of the most interesting virtualization startup at the moment, Qumranet, launches its first product: a VDI solution called Solid ICE.

Solid ICE is made of a connection broker, but also features a server component which adds resources control capabilities to KVM, and a new remote access protocol, called SPICE, which can be optionally used as replacement for Microsoft RDP.

The connection broker has some interesting capabilities in itself, supporting high availability and exposing a web portal for standard PC clients access which is designed to scale up to thousands of virtual machines. Despite that first version will provide basic capabilities to operate the virtual machine, with enhancements to be released over time.

The new protocol adds further value to Qumranet solution, being designed to deliver on thin clients all those multimedia protocols which usually don't perform well into a terminal services session (an approach which competes with NEC VPCC one).

Last but not least Qumranet took care to support several thin clients on the market, developing a dedicated MiniOS (probably a special purpose Linux distribution).

Solid ICE will support Windows 2000 Professioanl, Windows XP and Linux as guest OSes, and it's expected to be available before the end of this year.

Qumranet is interesting for three reasons:

• It's founded by a well-known entrepreneur in virtualization industry (and not only), Moshe Bar, already behind XenSource and InovaWave
• It supports and maintain KVM, the open source hypervisor which has been included in Linux kernel since version 2.6.20 (so every Linux box on the planet will have it).
• It plans to port KVM on other platforms, Windows on top of the list

The decision to leverage growing value of KVM first of all with a VDI solution seems a savvy move. First of all using Linux as hypervisor cuts a relevant part of VDI implementation costs, and with a Windows port in the work, Solid ICE may become a dangerous multi-hypervisor connection broker for competition. Secondarily, a focus on the desktop-side gives Qumranet time to improve most critical KVM capabilities on server-side before launching additional products, as virtualization.info expects.

Despite KVM immaturity, VMware may have problems justifying high entry-cost for its upcoming Virtual Desktop Manager (VDM), which requires the whole VMware Infrastructure to run.

But it's Microsoft that may receive the biggest damage here, losing the opportunity to leverage value of new System Center Virtual Machine Manager 2007 with its own connection broker. It seems now clear that every major virtualization vendor want to have a VDI solution as natural addition of the hypervisor: VMware with VDM, Citrix/XenSource with Desktop Server, Virtual Iron with Provision Networks Virtual Access Suite and now the whole Linux world with Solid ICE.

Both the virtualization.info Virtualization Industry Radar and Virtualization Industry Roadmap have been updated accordingly.

Labels: KVM

After XenSource acquisition by Citrix, another breaking news is going to shake virtualization industry: young virtualization platform KVM, already included in Linux kernel, is being ported on Windows.

At the moment there are no more public details than a mention on official FAQs page:

3.4. Do you have a port of kvm for Windows?

Not in this release.

Company maintaining KVM, Qumranet is still in stealth mode but it's expected to announce its first product before the end of September. During such occasion the virtualization startup may also unveil the upcoming port.

KVM for Windows would have a huge relevance in the virtualization industry, offering an open source alternative to commercial solutions made by VMware, Parallels and Microsoft itself.

Another young company offering an open source virtualization platform is german startup innotek, with its VirtualBox, but at the moment the product is oriented to desktop usage only.

KVM is also being ported on FreeBSD since May 2007.

FreeBSD is currently missing an enterprise virtualization platform and KVM may fill an important empty space.

More news about Windows porting as soon as possible.

Labels: KVM

Quoting from ZDNet:

...

At the San Francisco, Calif. conference this week, Qumranet, a Santa Clara, Calif-based commercial startup that funds KVM, told ZDNet that the company's first product will be unveiled in late September and will ship in the fourth quarter.

It won't be a knock off of rival XenSource's XenEnterprise. Because it leverages the KVM support in the Linux kernel, the new offering can focus on advanced services such as storage virtualization, hinted Qumranet co-founder, president and vice president of R&D, Rami Tamir.

...

Tamir offered up some fightin0 words. "Being latecomer is an advantage," Tamir claimed, then lowered the volume of his voice. "Xen is going away."

...

Following a conference session detailing the merits of Xen and KVM, Qumranet's Tamir acknowledged that several efforts are underway to enable interoperability. He noted that while one IBM programmer is working on code to add para virtualization features to KVM, and thus make it more competitive with Xen, still others at IBM are developing a CIM-based module that will ensure both engines will be supported and managed according to industry accepted management standards.

...

By his estimation, it's inevitable that Xen and KVM are headed for war but he was skeptical about interoperability promises, Rosenblum told ZDNet.

"There will be a battle of open source virtualization engines," he said. "I think one of the issue for both [Xen and KVM] is that current hardware support for virtualization doesn't recurse very well so if you try to run KVM inside Xen virtual machine it doesn't work and I don't think it'll work soon with any kind of acceptable performance. Obviously, on the same box you probably are not deploying both technologies."...

Read the whole article at source.

Labels: KVM

One year after its launch, Qumranet finally arranged first conference about its virtualization platform, KVM, which is now part of Linux kernel since version 2.6.20.

KVM Forum 2007 will take place in Tucson, Arizona, from August 29 to 31, and will put on stage very intersting sessions, presented by speakers from Qumranet, IBM, Intel AMD and others. Among interesting sessions there are:

• KVM Lite, No Hardware Support, Fewer Calories
• VT Roadmap, Hybrid Virtualization, Power Management, Fat vs Thin Hypervisor
• Standards Based Systems Management Solution for KVM
• KVM Live Migration

Look at the whole agenda here. Register for it here.

The virtualization.info Events Calendar has been updated accordingly.

Labels: KVM

Fedora Linux distribution is slowly integrating virtualization as basic OS capability.

In Fedora Core 5 Red Hat initially integrated Xen hypervisor, followed by a basic GUI for it in Fedora Core 6: virt-manager. Now new version 7 includes Linux kernel 2.6.21 and then offers out-of-the-box its second virtualization platform: KVM.

Integration of kernel 2.6.21 also implies Fedora 7 sports paravirt-ops framework and VMware VMI interface. This means new VMware Workstation 6.0 should be able to run it as a para-virtualized guest (with a major performance boost).

Last but not least Fedora 7 updated embedded Xen package to new version 3.1 (formerly 3.0.5).

Read full release notes about virtualization packages here or download the distribution here.

Labels: KVM, Xen

As announced last month, newest Linux kernel version, 2.6.21, includes for the first time an adapted version of VMware Virtual Machine Interface (VMI), able to interoperate with paravirt_ops virtualization interface.

At this point new Linux distribution going to use this kernel version will be able to run paravirtualized in upcoming VMware Workstation 6.0, and at a later time in Server and ESX Server new releases.

But kernel 2.6.21 also introduces remarkable new KVM features:

• Initial paravirtualization support
• Live migration (the guest continues running even while being migrated) support
• Host Suspend/resume support
• CPU hotplug support

Labels: KVM

The most famous liveCD Linux distribution, Knoppix, embraced virtualization completely with its 5.2 version.

The new release in fact sports all open source virtualization packages, including Xen, OpenVZ, QEMU (with KQEMU accelerator), VServer, and UML, up to most recent solutions KVM and Virtual Box.

This almost complete set (VMware Server/Player for Linux are the only missing since they are not redistributable), along with easiness of liveCD approach, makes Knoppix probably the best solution to approach virtualization for newcomers.

Unfortunately Knoppix 5.2 has been presented and distributed at CeBIT 2007 on physical support only. Who didn't attend the event will have to wait April to download the DVD/CD ISO, free of charge as usual.

Take a look at the remarkable features list here.

Labels: KVM

Ulrich Drepper, lead maintainer of Linux C library since 1996 and employed at Red Hat, worte an interesting post about KVM on his personal blog:

With KVM proving more and more that it is viable Xensource and VMWare start sandbagging. They call KVM immature and the wrong approach.

Calling KVM is immature is, well, premature and misleading. Xen has a headstart of several years. KVM is today not supposed to be in the state Xen is. Nevertheless, KVM already has hardware virt support, SMP support, support for 64-bit host and guests (despite what the article says), live migration, and more.

But immature is not the worst complain. Claiming the hypervisor approach is the only viable option is what should get people worked up.

...

These are bogus claims. And you have realize where they come from. VMWare’s ESX is a kernel on itself, one which only few people work on (compared to something like Linux). Device drivers will always be a nightmare unless/until devices get their own PCI devices (once DMA can be virtualized). Nevertheless, ESX is a full OS by itself. Plus, ESX has the service console a Linux OS. The service console of course has to have some control over the hypervisor.

For Xen the situation is similar. Here the hypervisor, after the mistakes of the 1.x series, don’t have device drivers included and use a privileged domain, a complete OS.

This means, both Xen and VMWare do not have less code. I’d say they even have more code that is part of the privileged code base. Certainly a Linux installation hosting KVM domains can be scaled down to only have the kernel, kqemu, and the service console...

Read the whole post at source.

Thanks to Fraser Campbell for the news.

Labels: KVM

Quoting from ZDNet:

Red Hat, the dominant Linux seller, will include KVM (Kernel Virtual Machine) in the next version of its hobbyist Linux version, Fedora, Chief Technology Officer Brian Stevens said Tuesday. "We're packaging it for Fedora 7," Stevens said.

...

However, Stevens said, KVM lags another open-source virtualization technology, Xen, which is the single biggest new feature in the company's upcoming Red Hat Enterprise Linux 5. "There's a year of work, I'd guess, to really make it at parity where Xen is today," Stevens said...

Read the whole article at source.

Labels: KVM

The new Linux kernel 2.6.20 is particularly important for virtualization, officially adopting a virtualization solution, Kernel -based Virtual Machne or KVM (32 and 64bits) as well as support for para-virtualization (only 32bits). (KVM itself is about to support para-virtualization)

Xen project founder and chief architect, Ian Pratt, expressed severe critics about KVM, as reported by DevX News:

Xen is a true hypervisor, whereas KVM is a legacy virtualization solution akin to VMware Workstation, VMserver and Microsoft Virtual PC

...

It lacks the benefits of para-virtualization performance enhancements that have been pioneered by Xen and are now being copied by VMware and Microsoft.

After seeking for long time inclusion in Linux kernel, Xen founder now snub the opportunity:

The Xen hypervisor is an operating system independent hypervisor that supports many different operating systems (Windows, Solaris, Netware, FreeBSD, OpenBSD, NetBSD) of which Linux is just one (though an important one)

...

Putting Xen into Linux doesn't make sense: hypervisors are different beasts from operating systems, so they share little code.

Labels: KVM

In December 2006 the young Kernel Virtualization Module (KVM) has been preferred over VMware and Xen approaches to introduce native server virtualization support in Linux, where will appear since kernel 2.6.20.

Now the project already adds an important feature, support for para-virtualization, thanks to Ingo Molner:

I'm pleased to announce the first release of paravirtualized KVM (Linux under Linux), which includes support for the hardware cr3-cache feature of Intel-VMX CPUs. (which speeds up context switches and TLB flushes)

...

• it provides an ad-hoc paravirtualization hypercall API between a Linux guest and a Linux host. (this will be replaced with a proper hypercall later on)
• using the hypercall API it utilizes the "cr3 target cache" feature in Intel VMX CPUs, and extends KVM to make use of that cache. This feature allows the avoidance of expensive VM exits into hypervisor context. (The guest needs to be 'aware' and the cache has to be shared between the guest and the hypervisor. So fully emulated OSs wont benefit from this feature)
• paravirtualization is triggered via the kvm_paravirt=1 boot option (for now, this too is ad-hoc) - if that is passed then the KVM guest will probe for paravirtualization availability on the hypervisor side - and will use it if found. (If the guest does not find KVM-paravirt support on the hypervisor side then it will continue as a fully emulated guest)

This patch significantly improves KVM performances by 4 times in context-switchting operations, nearing physical machines perfomances.

Labels: KVM

Quoting from Heise Online:

Linus Torvalds has included the virtualization environment KVM (Kernel-based Virtual Machine for Linux) in the tree leading to Linux kernel 2.6.20. In the case of KVM the kernel after loading a special module itself functions as a hypervisor for virtual machines.

...

KVM, which was presented to the public only barely two months ago, thereby easily overtakes other virtualization solutions such as Xen, OpenVZ and Vserver, which are based on other approaches, on the path toward integration into the kernel.

Read the whole article at source.

The original patch announcement discloses some details about current KVM status:

...

SMP hosts and UP guests are supported. At the moment only Intel hardware is supported, but AMD virtualization support is being worked on.

Performance currently is non-stellar due to the naive implementation of the mmu virtualization, which throws away most of the shadow page table entries every context switch. We plan to address this in two ways:

• cache shadow page tables across tlb flushes
• wait until AMD and Intel release processors with nested page tables

Currently a virtual desktop is responsive but consumes a lot of CPU. Under Windows I tried playing pinball and watching a few flash movies; with a recent CPU one can hardly feel the virtualization. Linux/X is slower, probably due to X being in a separate process...

More details about KVM are available at the official site.

Labels: KVM