Monday 14 February 2011

Is it finally the year of the Virtual Desktop?

I still remember it clearly, July 2006 and suddenly from VMware a new acronym was released into the wild 'VDI'. It wasn't that the virtual desktop had not been thought of as yet as there were already a number of options available and the 'offshore developer' and 'test' use cases were already being utilised by some pioneers, it was just that all of a sudden the concept had its own identity. Riding on the wave of success of the new VMware ESX platform (remember GSX Server?) virtualisation was suddenly looking grown-up and the possibilities were causing a notable stir.

At this stage being an Aussie working for an investment bank in London we had already deployed with reasonable success a Microsoft Virtualisation Server (MSVS) based solution to meet the needs of the packaging   authoring and test environments internally (and with 6000+ packages there was a lot of need!). Shifting focus to VMware's ESX 2 for the server development environment seemed like a logical 'grown up' next step to the virtualisation journey. The MSVS environment already enjoyed a limited amount of automation utilising the COM  automation libraries so developing around the VMware SDK was a natural progression. I was lucky enough to work with some pretty cleaver guys (although I would never tell them I thought as much) and at the end between us an automated service was developed that included the provisioning of the ESX Hosts, Windows Server guests and a pretty slick P2V process. All this helped the environment supporting the development environment to grow at a significant pace in various locations around the organisations international sites.

It was around the third quarter of 2006 that an issue with the limited availability of electricity to our London based Head Office was becoming a focus point and ideas were being thrown around on how to resolve it, step forward 'VDI'!  The challenge was raised, could the lessons learnt through virtualising server workloads be applied to good effect to provide a platform to serve desktops and in doing so reduce the power requirements to a standard end users desk?

I wont go on about the pain of the next 6 months but will say that having worked closely with the engineering and development teams of VMwareWyse and Leostream we rolled a desktop replacement solution into our London, New York and Singapore offices that provided a fully distributed, redundant, 'work anywhere, return home' desktop service built on the newly released VI3 platform. There were a lot of lessons learnt on the way in areas such as the guest build (we deployed the virtual desktop solution on the back of the Windows XP migration), developing an effective management system, managing end-user expectations, training support personal and suddenly realising significant advantages available.

How brilliant a virtual desktop solution is for an organisation, it can provide:

  • Workplace flexibility - No longer being constrained to a specific desk, office or even country when accessing your desktop. How good is it to be able to work from home from your own personal machine as if you are in the office.
  •  Business Continuity - Say good-bye to those pools of cold machines sitting at a recovery site just in case a site suffers some kind of significant outage. Think how easy a BCM process can be if you are not tied to a specific location and you can leverage high availability of your data centres. 
  • High Security - No longer having any risk of loss of data if machines go 'walking' from an office. All content is secured within virtual containers in a secure data centre facility. It is by no small coincidence that a major fan of the technology was our internal security team.
  • Improved support model - This is an easy one, you put desktops on the floor you need people to walk the floor to support them, you put your desktops in the data centre and stateless thin clients in the office then you can refocus some of those people on supporting the desktops from a central location reducing the time and effort required. Add to this the ability to fully automate the management and provisioning of the environment and you can realise a pretty slick quality of service.
  • The Green Office - Who would have thought it, a technology that brings significant benefits to the corporate office and to the environment at the same time. When the virtual desktop is paired with a thin client device drawing only 7-14 watts of power an office can easily realise a reduction of 100 watts of power per desktop device. In corporate offices this can result in a significant carbon emission reduction not to mention a reduction in the power bill (have you ever asked how much your company pays for electricity?).  The other benefit is the quality of power provided to the floor, with a desktop there is a risk of corruption and loss of data when a sudden loss of power is experienced, this is not the case when using a stateless thin client - are office wide power supply assurance services such as UPS required to cover everything or even anything. What level of savings can this provide to facility fit-outs?
I will not gloss over the journey to a mature virtual desktop solution, as any early adopter can testify, there have been many challenges on the way. Pain was experienced in many areas including:

  • Bridging the usability gap between a standard desktop and virtual desktop
  • Making the cost model fit 
  • Transforming a support model to fit the service
  • Experiencing pain points at scale
  • Beating the Monday morning 'access storm' blues
  • Managing the capacity in way to ensure the best cost versus performance balance
The advantage any adopter in 2011 has though is that there is now a lot of intellectual collateral out in the wild from real world 'lessons learnt' at enterprise level scale. There is also the benefit of the maturing of the desktop brokers and connection protocols on offer from vendors such as Citrix with XenDesktop and VMware with View. In the final stages of the VDI evolution all the focus had swung towards the protocol and the value added features it provided. This will continue to evolve rapidly as will the continued efficiencies in the back-end infrastructure helping the cost models to make solid economical sense (this is a big topic and I will dedicated an upcoming blog entry to this topic alone) along with the ease of acceptance of the offering with the end-users.

I spent a lot of time since 2006 around industry technology round tables and councils hearing everyone talk about there 'proof of concept' and 'pilot' virtual desktop environments with very little moving beyond. The atmosphere of these changed quickly in 2010 with everyone seeming to ramp up and start to take it seriously. Bring on 2011 and I now spend more time talking to clients about virtual desktops then servers and you definitely do finally get the feeling'2011 could well be 'the year of the virtual desktop'.



+++++++++++++++++++++++++++++++++
As a sub note to this in 2014, there are two things that have changed in the last three years

  1. As it should have been, VDI is no longer approached as the answer but as a part of an organisations overall End User Computing strategy. This is a good thing as the drive for change is from the business perspective and not from the technology which is what VDI is
  2.  Technology to support VDI has advanced hugely which in turn has resulted in dramatic cost reductions while at the same time, improving the overall user experience. This has happened at the transport protocol, server hosting and at the storage layer. I have been lucky enough to spend quite a number of hours with EMCs new all flash array XtremIO within our lab environment and am still blown away by what I have pushed this array to do in performance and storage (thanks to inline dedup) you can get this device to do. Capacity  for 2000 desktops in the space within a rack that normally would be associated with a network switch.... sign me up :)