virtualization.info

The last time AMD showed a public video of a live migration it turned a number of heads.

The company used an unreleased and unrevealed management console for KVM (which is probably what Red Hat is about to launch), showing how a virtual machine could be live migrated between an AMD Quad-Core Opteron (codename Barcelona) and an Intel Xeon DP Quad Core E5420.

This time AMD is a little more careful, and unsurprisingly decides to change its demo product, switching from Red Hat/KVM to VMware (there’s a lot of partnership to build against Intel here).

The VI 3.5 virtual machine is migrated from a Quad-Core Opteron (65nm) to a Quad-Core Opteron (45nm) to the upcoming Six-Core Opteron codenamed Istanbul (45nm):

This video is now broadcasted on virtualization.tv, the new webTV channel of virtualization.info.

Labels: AMD

In September 2007 AMD was the first on the market to introduce an implementation of the much awaited nested page tables technology that promises unprecedented performance for virtualization platforms: the Rapid Virtualization Indexing (RVI).

As every virtualization professional knows, even the most enhanced CPU extension it’s useless without a virtualization vendor that supports it in its hypervisor.
VMware introduced support for AMD-V RVI almost one year ago with the release of VI 3.5.

Today most hypervisors support it (to see which ones you may want to check the brand new virtualization.info Buyer’s Guide) but we still didn’t have much details about the performance boost that this technology provides in a virtual infrastructure.

This week VMware released a very interesting 9-pages whitepaper answering the question:

…We evaluated RVI performance by comparing it to the performance of our software-only shadow page table technique on an RVI-enabled AMD system. From our studies we conclude that RVI-enabled systems can improve performance compared to using shadow paging for MMU virtualization.

RVI provides performance gains of up to 42% for MMU-intensive benchmarks and up to 500% for MMU-intensive microbenchmarks…

These results come out of an analysis conducted with VMware ESX 3.5 Update 2 and the upcoming AMD Quad-Core Opteron 8384 (codename Shanghai).

While waiting for the Intel version of this technology, dubbed Extended Page Tables (EPT) and expected somewhere between 2009 and 2010, this paper is a great reading.

Labels: AMD, VMware

Just one week ago AMD demonstrated that performing a virtual machine live migration from an Intel CPU to an AMD one is possible.

Today SearchServerVirtualization publishes a very interesting article with VMware and Intel comments on that demonstration:

…"VMware currently provides full support for Enhanced VMotion Compatibility, which allows live migration of enterprise workloads across different processor families within the same CPU vendor. This technology provides customers the flexibility to move these workloads across different processor iterations in a stable and reliable way," said Richard Brunner, the chief platform architect at VMware and a former Intel CPU architect.

But "attempting to make cross-vendor x86 instruction sets and features compatible in a VM for live migration puts this stability at risk, and so we have not pursued it," he added…

Probably VMware is right in being so cautious about supporting cross-CPU live migrations, but the author correctly highlights its special relationship with Intel, that invested $218.5 million in the virtualization vendor (and was even rumored to be in bid to acquire it).
Maybe, without such relationship VMware would take some more risks and support the AMD effort.

While waiting for VMware and Intel to change their mind, savvy and well-respected virtualization professionals don’t mind tricking non-production ESX hosts with some CPU masking stunts (see here, here, here and here).

Labels: AMD, Intel, VMware

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