I’ve been in new new role with Xtravirt now so seven weeks or so now, and it’s interesting at this point to take a quick checkpoint and look at what I’ve observed already. I’ve been in the End User Computing game for some 14 of the 16 years I’ve been in the IT industry, in fact, it wasn’t even called EUC back then. I think the euphemism was something like “bloody end users!”, but of course we live in enlightened times these days, and we have to give them a more businesslike moniker.
The point is that EUC now is primarily a virtualised environment. Since I started my new role, I’ve been exposed to XenDesktop, XenApp, Login VSI and a whole raft of other tools. The interesting thing about the “new” EUC space is that it forces you to turn traditional desktop approaches on their heads. For example, Windows is written based around the fact that it usually has a local disk all to itself, that it can thrash to it’s heart’s content, without having to worry about running into other resources trying to land grab from it. In a virtualised EUC environment, this is no longer true. There can be several dozen virtual machines booting at once, logon storms, anti-virus scans and a whole batch of other processes going on simultaneously that are competing for the same disk resource.
Additionally, in the days of server consolidation (the phase 1 of mainstream virtualisation, if you will), capacity planning tools such as VMware Capacity Planner or PlateSpin Recon would be set up to capture performance metrics from physical servers, to essentially baseline what CPU, disk and memory resource was being used, so that the virtual equivalent could be appropriately sized for performance. In the EUC space, this is no longer sufficient. As well as capturing the previous metrics, we also need to look at additional detail based around the end user experience. If logon to a server console is slow, generally no-0ne but the admin would notice, and as frustrating as it might be, it’s generally tolerated and goes unreported. In the EUC world, when several dozen users logon at the same time and experience is degraded, IT will get to know about it pretty quickly.
As such, the likes of Login VSI help to determine the performance of the EUC experience using real world examples such as Outlook, Flash and manipulation of large spreadsheets. Traditional capacity planning tools are very useful for obtaining basic figures on specifications, but lack the insight to analyse application performance and the impact on a virtual desktop environment.
Away from such matters, it’s also interesting to look at applications. As I remarked at a BrainShare event presentation several years ago (before iPads and VDI in 2007), the apps drive the platform, not the other way around. Generally, users don’t care if it’s Mac, Linux, Windows, iPad, Android or Etch a Sketch, as long as they can get access to their line of business applications in a usable manner. The underlying layer of the OS generally just becomes another commodity item. I didn’t think I was being particularly visionary back then, just a pragmatic view based on the way I approached things as an end user.
Whilst enterprise applications such as Microsoft Office come with tools for the virtual environment, many core business applications are written in house and are proprietary to the business. As such, they tend not to have enterprise deployment tools, have extensive user communities or knowledge bases, and are written on the “good enough” principle. Again these apps are written with the assumption that the endpoint is a largely static thing, that the hostname doesn’t change and that it never moves around the network or across continents. In the virtual EUC space, this is no longer true and we must now be creative into fooling the app into thinking it’s still living in the traditional desktop environment.
It’s been seven weeks of change, steep learning curves and a change of thinking, but I’m enjoying every minute and it’s certainly the challenge I was hoping for.