Intel Developer Forum 2009 is going on this week at the Moscone Center in San Francisco. This isn’t an event that I’d ordinarily travel to, but since it’s only a block from my office, I figured I’d drop in on the few sessions about client virtualization.
This afternoon I attended a session called "Client Virtualization and Intel vPro Technology." Overall I was pretty disappointed with the session. The presenters broke it into two parts which was tough because the whole session was only 50 minutes to begin with.
The first part was about desktop virtualization in general and focused on what Intel calls the "Dynamic Virtual Client," which is basically their word for selecting the proper form of desktop virtualization at any given time. (VDI, client-based VMs, OS streaming, app streaming, etc.) While I of course agree with the concept, it's interesting that Intel's not actually doing anything to help the system flow between these different models--they're just kind of saying that all these different models are possible.
The second half of the presentation was the part I thought I was waiting for, where they'd talk about how all the various types of Intel VT technology would be applied on the desktop. (I think most of us are familiar with what VT does in our servers, so I was excited to learn about the desktop-specific applications and advances. Unfortunately this part of the presentation was fairly generic. We got what appeared to be the standard VT slide deck that wasn't really desktop-specific at all.
Sure, there were some mentions of desktops, but only in the context of how VT can improve VMs in general. For example, I know VT-d can allow guest VMs to directly map I/O devices and that can increase performance, etc. But when it comes to desktops, I want to know how that specifically applies to things like USB. And what about GPU or graphics virtualization, something that's key for desktop virtualization but has so far not been addressed (by Intel VT in general or in the session)?
I was also curious to know a bit about Intel's desktop virtualization announcements with Citrix, VMware, and Symantec. What exactly does that mean from Intel's perspective?
I actually learned more about Intel vPro and how it relates to desktop virtualization from my short conversation with VMware's Robert Baesman at VMworld a few weeks ago. When talking about VMware CVP (VMware's client hypervisor), Robert mentioned that vPro technologies like TxT would allow the system to authenticate that the hypervisor code hadn't been tampered with to ensure a secure boot.
The one interesting tidbit I picked up from today's session was that Intel's Extended Page Tables (EPTs) will be coming to the desktop and mobile processors in 2010. EPTs basically create a system-wide map between the addresses a VM thinks it's using for its page tables and where the hypervisor is storing them. This then allows the VM to directly access and create its page tables, a process that currently requires an expensive "VM exit" for the host to act as the middleman.
Regardless of the quality of this session, it's clear that Intel is putting bit of virtualization technology into its desktop and mobile hypervisors that it can. I feel as strong as ever that in a few years, every device will run a hypervisor and the hard core management and security capabilities will happen below the VMs.
So I guess that's about it for my thoughts on Intel and the IDF. What do you think about Intel's desktop virtualization direction? Are they moving down the right path? What would you do differently? And what about AMD?