Amazon’s cloudy desktops creep towards cloudy workstations
Adds cache-like storage currently offered for servers to speed demanding desktop workloads
Amazon Web Services has made an interesting tweak to its “Workspaces” desktop-as-a-service offering: temporary local storage.
The cloud pioneer’s previous Workspaces all had to make do with 100GB for a root volume and the same quantity of storage for user data.
The new Graphics.g4dn and GraphicsPro.g4dn Workspaces add temporary local storage using an AWS offering named “instance store” that the cloud colossus currently recommends as “ideal for temporary storage of information that changes frequently, such as buffers, caches, scratch data, and other temporary content, or for data that is replicated across a fleet of instances, such as a load-balanced pool of web servers.”
The Graphics.g4dn Workspaces bundle includes 100GB of instance store, along with 4vCPUs, 16 GB of RAM, and 16 GB of video memory. AWS suggests that combo can handle mainstream graphics-intensive applications, such as engineering, design, and architectural applications.
The GraphicsPro.g4dn bundle offers 6vCPUs, 64 GB of RAM, and 16 GB of video memory, and is said to be ready for “media production, seismic visualization, GIS data processing, data intelligence, small-scale ML model training, and ML inference.”
Workstation workloads, every last one of them. And not workloads that AWS previously encouraged as ideal for Workspaces.
It helps that both of the new instances run-on second-generation Xeon Scalable CPUs (Cascade Lake) tuned to AWS’ specifications, plus NVIDIA T4 Tensor Core GPUs.
- Alibaba Cloud lets its tiny desktop-as-a-service client leave China
- Azure Virtual Desktop on Azure Stack HCI – for those who want cloudy remote desktops hosted on-prem
- Microsoft's Cloud PCs debut – priced between $20 and $158 a month
The new instances aren’t cheap - $537 and $959 a month respectively – but it’s also possible to pay a monthly reservation fee and hourly rental. Bringing your own Windows license knocks a few dollars off the monthly fees and a few cents off the hourly rate.
AWS last launched GPU-enabled Workspaces in 2018, making these new bundles perhaps a little overdue.
Cloud workstations have been in demand of late because physical workstations don’t thrive in work-from-home environments where they’ll usually go without the high-speed networks required to move data into and out of demanding applications.
Amazon’s cloud networking makes that problem moot and adding the local instance store will mean fewer demands to move data into the new bundles. Users still need clients that run on Windows and MacOS, or streaming tech like PCoIP, leaving plenty of potential for those using consumer-grade broadband connections to experience frustrations. ®