Ten reasons to think twice about virtual desktops

The business case for deploying virtual desktops was comprehensively ripped apart by analysts at the Gartner Symposium last week.

In one of the most in-depth presentations at the event, Gartner analyst Mark Margevicius cited a wealth of research and case studies to reveal the technical and business headaches involved with migrating to the latest flavour in thin client computing.

Gartner’s most optimistic estimate was that 10-12 percent of enterprise PC users would be on virtual desktops by 2015. And these, Margevicius said, would primarily be for call centre staff, sales staff, remote app developers and teleworkers.

“The economics simply don’t make sense for most enterprise cases,” he concluded. “The traditional PC will remain the primary tool on the enterprise desktop.”

iTnews has summarised Margevicius’ concerns below.

10. Server headaches

The worst assumption a desktop IT manager can make, Margevicius said, is to assume that he or she will get the same performance out of industry standard x86 servers used for applications such as web servers as it would from a dedicated fat-client computer on every desktop.

“Traditional servers are not designed for high density workloads,” he said. “Web applications are very different to how a Windows desktop behaves.”

Margevicius said IT shops first need to work out whether their server infrastructure can support a virtual desktop deployment “based on number of users per core" before making a business case.

He said most deployments can handle no more than between 7 and 9 users per core – which stacks up to a big investment in servers.

Further, users would still require between 1GB and 2GB per user of memory to match PC performance.

On most servers in data centres today, “you are looking at maxing the server out of memory,” he said.

“The limitations of today's servers make [virtual desktops] cost prohibitive,” he said. “Memory is a big deal. Processors are a big deal.”

Margevicius acknowledged that vendors such as Cisco and HP are releasing blade server configurations designed with those precise requirements in mind, so that organisations can “scale out” their infrastructure.

“There is a significant benefit in unified architectures, predominantly because of memory utilisation,” he said. “But just having the server technology isn't enough.”

9. Storage headaches

Administrators would also be wise to check if their storage kit can handle a move to virtual desktops.

Margevicius said administrators are naturally unsure of how much storage to allocate to a virtual desktop deployment.

“Storage is the biggest wildcard,” he said. “If you are going to mimic the environment of a physical PC that comes with 250GB of storage, are you really going to allocate 250GB of Tier 1 or Tier 2 storage to every virtual desktop?”

Most customers opt for 20-30GB of storage per desktop (if it’s a persistent image), he said.

Storage performance is another concern.

“The real bottleneck is the IOPS [input/output operations per second] going across the disk,” Margevicius said.

“On desktop workloads, typically the read/write ratio is biased more towards writes [compared to servers]. On most server workloads, the write is about 15 percent and read 85 percent."

The Windows operating system, he noted, is a “very chatty OS”.

“When you consolidate Windows workloads, too many users will absolutely affect the performance,” he said.

Margevicius recommended that IT managers investigate the use of de-duplication to address storage volume limitations and advanced caching techniques to address read/write IOPS limitations.

8. Network headaches

Margevicius said that less often, virtual desktop deployments also require investments in new networking kit.

He said that all the major flavours of virtual desktop technology have very efficient protocols for communicating across the network (Microsoft’s RDP, Citrix’ ICA, VMware’s PCoIP etc). Some of these can run typical sessions of only 65-85kbps, owing to thin client predecessors designed to link branch offices over 56kbps connections.

“Most network administrators love [thin client], predominantly because bandwidth is limited and therefore more predictable,” Margevicius said.

Nonetheless, Gartner has been asked to advise several deployments in which end users took these efficiencies for granted.

“Latency is a big problem for hosted desktops,” he said. “The question is how long it takes for characters a user have typed to hit servers. If the latency gets in excess of 150 milliseconds, users start to complain.”

Read on for seven more reasons to think twice about virtual desktops...

7. Application performance

These limitations in server, storage and networking assume, of course, that users still want access to the richness of the applications they have grown used to on the desktop.

Understandably, most users revolt if "upgrading" to an inferior user experience.

"Applications with multimedia, for graphics or 3D modelling, still work best on a dedicated machine," Margevicius said.

While improvements are being made to improve the multimedia experience on the virtual desktop, it will be another 12-18 months before these issues are ironed out, he said.

For now, such limitations restrict the utility of virtual desktops to call centres and kiosks, he said, rather than traditional office workers.

 6. Licensing headaches

Margevicius said that too many CIOs assume that the software licenses that exist today on a physical PC are translatable to a virtual PC instance.

"The reality is that is not the case," he said. "Microsoft, for example, has very specific requirements around Windows licensing - a license is physically tied to a given device and is not transferrable."

Some customers assume they are covered under the 'Software Assurance' maintenance they pay Microsoft, which is around US$40-US$60 per device per year.

"But in fact, to deploy virtual desktops on your thin client or iPad, you require yet another license which is not covered by Software Assurance," he said. "It's called VDA or virtual desktop access and it's a brand new license this year. It will cost you US$100 per device per year."

Margevicius said that Microsoft is using the hype around virtualisation to shift to an annuity business model.

"Microsoft is very sneakily changing the pricing model for Windows," he said. "On the traditional desktop you purchased a perpetual license and Microsoft could never charge you another dime.

"But by doing virtual machine licenses, Microsoft has very sneakily changed its model to an annuity scheme. Like Software Assurance, VDA is a non-perpetual license - but by the way, the license is still tied to the device.

"They get you," he told Symposium attendees. "People ask me, when is Microsoft going to enter the virtual desktop market? I answer that they are here. They are already charging you for doing nothing."

5. What about offline access?

While concerns are often raised as to how much time is wasted on the internet, the reality for many information workers today is that a loss of internet connectivity has a serious impact on productivity.

But in the traditional fat client world, there are still tasks users are able to perform when offline - replying to stored email messages (queued for re-connection), for example, or working with office productivity tools -- assuming the company hasn't shifted these to the cloud.

So beyond concerns users might have about application performance or indeed having some level of control over their desktop environment, the lack of offline access to files and applications can be a barrier to user adoption.

"The statistics show that 40 percent of users are buying a PC notebook" for desktop use, Margevicius said.

"There is obviously a requirement to be functional when you are off-site or off-line for whatever reason. And we are still 18-24 months away before an offline version of a virtual desktop is viable."

4. Internal politics

There are other cultural barriers to deploying virtual desktops.

The PC fleet has historically been the domain of a desktop or end-user computing group within large organisations. Some of these IT staff will be reticent to hand off desktop control to a data centre team.

And more importantly - even if the desktop team was willing - the assumption is that the data centre team wants the headache of having to cope with managing end user computing needs.

"Remember, as you design your hosted desktops, keep in mind that scale matters," Margevicius said. "It's not just about adding a server - it's about adding thousands of users to manage - and more storage, space, power and cooling needs to be allocated within the data centre design."

He warned that a large organisation's PC group is "not people who have experience with servers" and may need to be led in some respects by the advice of the data centre team.

"It's much bigger than a PC implementation and touches many places," he said. "When you create virtual desktops, it is an infrastructure play that crosses all disciplines."

Read on for the top three reasons to think twice about virtual desktops...

3. Not necessarily easier to manage

Margevicius said it would be incorrect to assume that virtualising the desktop makes it easier to manage.

There are two types of virtual desktops today.

The first is the pooled environment - in which regardless of which virtual machine any user logs in to, they see a fresh instance with the same look, feel and applications as their peers. This version of the virtual desktop is, almost without question, easier to manage.

"It is naturally very scalable and the least costly to implement. But is far more static than many users actually like," Margevicius said.

The second is the persistent environment, in which each individual worker has the ability to maintain a persistent layout from day to day.

"When I log in from wherever I am, persistence is maintained to me - it is a replication of the physical environment into the virtual," he said. "Users are likely to prefer this model, but this is far more expensive and complex to manage."

Gartner anticipates a future virtual desktop that can be managed like a pooled environment, with "the technology in place that can maintain persistence," he said.

"That means de-coupling the things that I do as a user for customisation and storing it independently of the image," he said. "It is not here yet, but most definitely on its way within a few years."

In the meantime, Margevicius said that the roll-out of virtual desktops doesn't magically "make PC management issues go away.

"You are just virtualising the mess," he said. "It moves - but it doesn't go away."

He expects another iteration of improvement in management tools needs to evolve before he could confidently say virtual desktops are easier to manage.

2. Benefits are hard to quantify

There is, according to Gartner, a good possibility that upgrading to virtual desktops can save an organisation money over time.

But the problem most CIOs face is having to justify the upfront capital costs of the new infrastructure required.

"The Total Cost of Ownership of your desktop environment will drop," he said. "But most cost savings occur in hidden and indirect ways.

"If a user reboots less frequently, if the number of helpdesk calls goes down and there is less time spent on calls, these are real benefits. But are they quantifiable for the organisation?"

There is only one way to make such a business case stack up, he said.

"It is a very good idea to understand your existing support costs," he said. "The most important argument as to why virtual desktops should be deployed is to argue from an operational point of view.

"You need to look at that army doing break-fix and distribution and be able to say - I can reduce headcount for 40 percent to justify this investment." 

1. The economics don't stack up

If there are no headcount reductions to be made, the business case is going to struggle to stand up, he said.

"The costs that matter most today are capital costs," he said.

"When you do comparison of PC pricing to the real investment in server, storage, network, licensing, management tools and cabling required for virtual desktops, the cost of the virtual desktop is 1.4 to 1.7 times more than the traditional PC," he said.

The cost of virtual desktops could drop to 1.5 over the year or so," he said, "but it will never get as cheap as a PC does.

"That is too much for many organisations to stomach and we agree."

What do you think? Does the business case for virtual desktops stack up?