Recently I had a conversation with ClearCube CEO Randy Printz. ClearCube is a desktop virtualization company that sells turn-key solutions based on hardware zero clients connecting to a combination of VDI desktops and physical blade workstations. But in today’s world of virtual desktops, is there still a need for blade-based solutions, or is this a relic of the past?
ClearCube has been making blade workstations and clients with chip-based PCoIP for quite a while now, since back before Teradici partnered with VMware and there was no software-based PCoIP. These days, much of ClearCube’s business is with high-security industries like finance, government, and defense—areas where the physical separation of the back-end blades makes a lot of sense (and is often required by law). They have specialized products for these industries, such as blades with physical jumpers that will lock out USB drives on the client end and a multi-network client device with a built-in KVM.
Looking forward, ClearCube is anticipating increased sales of their zero clients thanks to prices that should be coming down soon. But when thinking about blade-based desktops versus VM-based desktops, I wonder if the blades really have a chance in the long term? Virtualization is always getting better and more cost efficient, and in fact two more historic “blade only” use cases have been knocked out of the picture by recent announcements.
The first use case is environments requiring high-end graphics. Doing this with a remote desktop used to require a blade workstation with a physical GPU, although that’s starting to change now. At Citrix Synergy last May, Citrix showed a preview of HDX 3D Pro that allows a physical GPU to be passed through the hypervisor and exposed natively to a virtual machine. The relationship of GPUs to machines is still 1-to-1, but now at least they can be virtual machines. More recently VMware announced that they will support NVIDIA’s Virtual Graphics Platform, which will also result in GPU to VM pass-through.
The other recent change in the “blade versus VM” game is that using PCoIP hardware encoding for the remote desktop used to be limited to blade environments with a physical PCoIP add-in card. But at VMworld last year, Teradici announced the Apex 2800 PCoIP server offload card. This single card can offload up to 300 million pixels per second of PCoIP encoding, typically enough for 50 or 60 users to share the same card. Better yet, the offloading is seamless and dynamic—there’s no need to log off of a VDI session in order to log on to a blade.
Of course, somebody will have to put all this together, and in the meantime maybe a blade is still the best way to do it. And we still need blades for situations where physical separation is mandated for security reasons. But now that we have GPU pass-through and hardware PCoIP offload capabilities, are blades becoming an anachronism, or am I being too quick to dismiss something that’s still useful?