Back to Mainframes… well, sort of

This particular topic has been on my mind for some time now. If we looked at the evolution of business computing, it looks something like:

’60s & ’70s – Users accessing applications that run on Mainframes through dumb terminals. No need for the management of end-user computing.

’80s & early ’90s – Personal computing with client-server applications where the presentation & business logic layers manifested as applications on user’s PCs. Data resided on servers. Starting to need to manage end-user computing environments: ensure compatibility, patching, etc.

late ’90s & early 2000s – n-tier application architecture, all computing logic remains on the server, with limited presentation logic sent to the browser. We got better at standardizing end-user computing environment through packaged OS installs & reduced application compatibility requirements, allowing for more personalization.

mid-late 2000s – proliferation of virtual desktops, where controlled environments are centrally managed and delivered through mechanisms such as Citrix or Teminal Services. Though the individual user sessions share common computing resources from the OS layer down.

today – centrally delivered personalized virtual desktops, with true separation / individuality at the OS layer, with dedicated computing resources that are allocated dynamically. The laptop I lug around is nothing but a glorified dumb terminal. Okay, maybe we’re not there yet, but VDI & related technologies are providing us with the tools that make that a reality.

As virtual desktop delivery become a common practice, what products and services can we expect to see in the space? Here are some of my thoughts:

- 1-lb laptops that does nothing acts as clients to the virtual desktops

- rent-a-desktop for businesses that have seasonal spikes (ie. tax prepare shops during tax seasons, schools no longer subjected to out-of-date PCs, etc)

Of course, as the distance of the desktop & applications shrink… does that mean the applications themselves go back to being fat clients again? WAN optimization will no longer need to be concerned about application-specific protocols. Instead we’ll simply need to worry about a more efficient way to delivery screen scrapes?