A Case Study in Virtual Workstations for CAD28 Nov, 2018 By: Alex Herrera
Herrera on Hardware: The experience of architecture and engineering firm Mead & Hunt presents a compelling proof point for a new CAD computing paradigm.
While mitigating version control and performance penalties may have been the driving motivation, Mead & Hunt’s decision to use cloud storage and sourcing of project data also delivered on another critical IT goal: hardening security of that data. Not only is version control less vulnerable, but so is data security and resilience. Sensitive corporate IP — or even more concerning, a partner’s IP — never leaves the cloud, so files can’t be snooped via a coffee shop Wi-Fi connection, read off a stolen laptop, or copied onto a flash drive for transport.
Yes, once you give up control of your data to the cloud, that data is only as secure as the cloud provider makes it. But for several reasons, a provider’s security is in all likelihood far more secure than the average “on-prem” datacenter (one located on the premises of the company that owns the data). First off, consider the onus on the provider to keep customers’ data secure. Do you think AWS doesn’t have security as priority one, knowing what even one high-profile data breach could do to their business and bottom line? Second, with the luxury of having built out their infrastructures relatively recently, providers such as AWS were able to create from scratch with security in mind, and without many of the holes that supporting legacy hardware and software can often open up.
The Second Step: Trading Physical for Virtual
So that critical first step to the cloud — centralizing data that all users access remotely — paid off for Mead & Hunt with valuable improvements in productivity, robustness, and security. But taking that first step also begged a possible and equally compelling next step: centralizing the workstations in the same place. After all, if all the data is already up in the cloud, why do the processing all the way down on the client? Might as well move the workstations to the data and leverage another host of advantages for users, administrators, and management alike.
Seeing that opportunity, Knauf’s team embarked on a transition, trading in tried-and-true deskside and mobile workstations in favor of virtual workstations in the cloud. Mead & Hunt turned to Workspot, a provider delivering turnkey virtual workstations built on a foundation of Microsoft Azure GPU-enabled cloud-based machine instances. (It’s worth noting that Workspot alone provides the virtual workstation solution to the customer, with Microsoft Azure as its back-end cloud provider. So there’s no paying Azure; instead, all is covered through Workspot.) With Workspot, Knauf’s team can provision any number of virtual workstations for Mead & Hunt staff, no matter where they’re located.
Workspot’s virtual workstations are built on the Azure cloud. Image courtesy of Microsoft Azure.
Mead & Hunt personnel now have several configurations of virtual workstations to choose from, each of which can be provisioned dynamically, daily or even hourly (rather than once every product lifetime, as is the case with physical client workstations). Consider the specs of these Azure N-series machine instances, configured specifically for graphics-intensive workstation applications. Today, they range from the NV6 instance, with six CPU cores, 56 GB of memory, a dedicated NVIDIA Tesla M60 GPU, and 340 GB of storage (temporary, initialized every time the virtual machine is spun up), up to the NV24 instance, with 24 cores, 224 GB of memory, four dedicated M60s, and 1.4 TB of storage. Compare those specs to a typical entry-class workstation of four cores, 16 GB, and a GPU perhaps a half to a quarter as powerful as an M60.
Mead & Hunt personnel use HP laptops as clients to access Workspot/Azure virtual workstations, but they theoretically could have chosen almost any device. Because all the processing is done on the cloud-hosted workstation, the end-user devices could just as easily be desktop PCs, Macs, workstations, thin clients — even tablets or phones. After all, that’s one of the major appeals — or at the very least, an appreciated bonus — of a virtual solution: simple, agnostic client devices.
An Onus on Network Performance
Outsourcing the workstations and the storage to the cloud is a significant portion of the solution, but the network connections users will rely on to access them are also essential. In the case of Mead & Hunt, recall that the storage is on AWS and the virtual machines are on Azure. Fortunately, while a cloud provider would most certainly prefer to source all your cloud IT business, each needs to be as agnostic as possible, precisely to support approaches like Mead & Hunt’s. And with high-bandwidth, low-latency network options available to bridge infrastructures, Mead & Hunt had no problem with Azure virtual machines operating on AWS data; Knauf cited delays between the two of around 5 ms.
Getting from Azure to Mead & Hunt client devices scattered around the country is the second potential bottleneck to address. After all, if bandwidth is too low or latency (the round-trip time between clicking the mouse and seeing the visual result of that click updated on-screen) is too high, then all the value of virtual workstations is nullified. Poor image quality and/or awkward delays crush productivity. Consider how annoying and ineffective a high-latency video conference call becomes when participants talk over each other — and that’s only a short part of your day, compared to the bulk spent at your workstation.