Jaymes Davis, Senior Director of Innovation and Professional Services Organization, Office of the CTO, Sphere 3D
As the technology futurist Peter Drucker predicted in the early 1950s, knowledge workers have prevailed and their tool of choice for end-user computing is a physical PC running a Windows desktop. We are still working towards ubiquitous access for all our productivity needs, the nirvana of end-user computing. To that end, we needed apps that run on server OSs delivered via the Internet to supplement our desktop.
For the last 20 years, virtual architects used a data center methodology of centralized compute and centralized management of servers and data to provide these server apps. Shifting to this centralized model in order to increase efficiencies in server virtualization metrics (users to admins being a major one) uses ever-increasing bandwidth to present the illusion of distributed computing to end users. However, these innovations did not liberate the physical desktop which still accounts for a majority of the content created by knowledge workers.
When Virtual Desktop Infrastructure (VDI) came along, it was dubbed a plausible solution to finally moving a broad swath of end user computing to the data center, analogous to the way server based computing moved to the data center due to server virtualization. Certainly innovation has made VDI from a centralized compute and management platform possible, yet the explosion of mobile (which severely increased end users and end user access) and the limits imposed by a server based technology have made VDI only a marginal success. The need for knowledge workers to get access to information in order to stay competitive has relegated VDI to a small sliver of the desktop landscape. As we evolve further towards the Cloud, this leaves us with a partial solution and technology innovations that further increase our risk profile (more complexity, more cost, more things to manage).
I propose a new architecture, Distributed Desktop HyperConvergence. It is a convergence of form and function for end user computing, placing compute in a shape that matches the user base by size, by workload and by location.
Understanding the Desktop as an End User Device
How long a desktop will hold value to an enterprise in this age of mobile and app explosion is a fair question. Certainly tablets have become a secondary or even tertiary device, while the smartphone has become a first-rate citizen. Yet the PC form factor still has enduring popularity in the enterprise. Google states 77% of hits on their sites are still from PC browsers, and 80% of the consumer sites and services are still based on PC access. Mobile OSs have similar functionality to desktops but lack major versatility such as screen size, which is one of the major factors in end- user productivity.
Looked at from an OS perspective, the life of most OSs in the enterprise are 3-5 years, with interoperability and coexistence lasting for 10 years. Today the OS is itself an app that runs apps that may only run on that OS. For example, Windows XP is still running in a percentage of corporations, particularly in the SMB and in enterprises that are regulatory-compliance heavy.
What about the app-centric revolution where PCs will become extinct and the Cloud native app takes over? Many developers are working towards this goal, but have run into roadblocks. Certain apps take a lot of time and money to rewrite. Certain apps have too few users or the retraining cost of the user base make rewriting prohibitive. Ambitious CTOs have tried the “corporate ASP model” and found barrier after barrier, slowing down the SaaS path. Finally, current desktop apps are built to take advantage of new physical PCs and the apps can be customized by each knowledge worker to improve productivity. Can SaaS really provide that level of customization?
Finally, be careful in what you wish. What happens if IT brings all my desktops back to the data center? Consider, there are more virtual desktops than virtual servers. If you backup 1000 desktops with 100 GB of Direct Access Storage Devise (DASD) each, and contemplate the needed compute and networking, then all of a sudden your data center will become a desktop data center with the functions supported being 70% virtual desktop and 30% virtual server. Can server administrators do a good job managing virtual desktops going forward?
Converged, HyperConvergence and VDI
Server virtualization hyperconvergence is the natural consequence not just of the technologies but the roles and responsibilities necessary for high user to admin density in the data center. Most hyperconvergence platforms are in the evolutionary process, 1.0 moving to 2.0. This transition from converged to hyperconverged platforms builds on “best practices in a box” and solves major Input/ Output Operations Per Second (IOPS) issues presented by the consolidation of physical servers into virtual machines. Hyperconvergence worked great for server virtualization, consolidating physical servers, increasing performance and reducing administration overhead.
Again, we apply this concept of server virtualization hyperconvergence to virtual desktops hoping for the same success. We fall short because desktop workloads are different from server workloads. We found that the total cost of ownership only became feasible for hundreds of users per site rather than tens of users, limiting the potential market for VDI. Administration tasks for desktop workloads are different than for server workloads, both from a desktop availability and a return time objective. How is moving to hyperconverged going to reduce the complexity of the current VDI platforms? Server virtualization has mature migration tools that do not translate well to virtual desktops. Server virtualization has defined characteristics of compute, storage, network, but doesn’t have to deal with graphics or varying user density due to variable workloads of productivity apps.
Are we destined to jam the square peg in the round hole or is it time to re-imagine desktop virtualization vis à vis server virtualization? Not at all. Look for our next blog on Distributed Desktop HyperConvergence which is coming soon.