Cameyo's new app virtualization platform allows 8-10 users per GPU

With workable GPU-accelerated application virtualization, running racks full of servers with GPUs at low utilization to fulfill Citrix 1:1 requirements can become a thing of the past.

Understanding virtualization impact on compute, storage, and networking Match Group's Henry Lin describes the different VMware products that the company is experimenting with across its stack.

Virtual desktops are often not exciting, due to their common limitations. Typically, these are thought of as relatively slow platforms on which to run such thrilling software as Microsoft Office, or QuickBooks, or some accursed application written in Visual Basic 6, which for inexplicable reasons, was never modernized yet is just important enough to your business that it remains in service. Certainly, GPU-accelerated applications such as AutoCAD, Adobe Premiere are possible on Citrix, though the pricing is likely steep enough at scale so as to prompt IT decision makers to abandon the prospect of virtualization entirely.

Virtual Desktop firm Cameyo is attempting to counteract that with the announcement of GPU acceleration for their core product. The difference, Cameyo co-founder and CEO Andrew Miller tells TechRepublic, is that GPU to user allocation need not be 1:1, introducing significant cost savings in the process. "We're thinking around eight to 10 users per instance for GPU based apps, depending on the application," Miller said.

SEE: Server virtualization: Best (and worst) practices (free PDF) (TechRepublic)

"There's two ways to do it. There's utilizing our fully hosted service where we have elasticity built in. Once the system detects that the server is using 50 to 60% of CPU, GPU, or RAM, it will automatically start spinning up new servers in the background. There's no manual interaction that needs to be done," Miller said. The second way is to deploy Cameyo's virtual application product as a self-hosted (on-premises or cloud-powered) solution. 

Cameyo's pricing structure is $25 per user, per month, with GPU capabilities adding an extra $25-30 per month, though this is based on actual GPU usage time. "Rarely, in the testing that we've done, have we seen anybody using GPU 24/7. We don't want to charge based on a 24/7 usage of the GPU cause as you can imagine that is incredibly expensive," Miller said. "Compared to Citrix, we typically see about a 10X cost savings when we start adding GPU."

Because of Citrix's origins as a virtual desktop platform, their offerings are built more around virtualizing desktops than individual applications, which makes multi-tenancy a challenge—as a result, Citrix deployments typically have a large footprint in order to accommodate large numbers of users. In an example migration, Miller notes that a customer with a 160-server Citrix deployment "is able to get down to anywhere from 10 to 12 servers to run the environment."

Roughly 65% of Cameyo's customers are self-hosted, according to Cameyo CMO Robb Henshaw, who notes that this is common among larger organizations and enterprises, while education customers and SMBs favor the fully-hosted solution, which Cameyo operates on top of Google Cloud Platform. 

Cameyo's GPU-accelerated offering is available now, initially on Windows, with support on Chromebooks planned by the end of 2019. Other device support is "to be announced in early 2020." 

For more, check out "Intel's 'Barlow Pass' Optane DC persistent memory comes to workstations" and "NXP i.MX 8M-powered MaaXBoard offers long-lived support, RPi HAT compatibility" on TechRepublic.

Also see

istock-872006502.jpg

Image: Getty Images/iStockphoto