Saturday 25 April 2015

If VDI is the answer for education, maybe it's the wrong question.

The idea behind Virtual Desktop Infrastructure (VDI)  is quite compelling.  Instead of running the Windows operating system on a local machine you run it as a remote instance in a server room and the client connects using a network protocol such as RDP or PCoIP from a 'thin client' which can be supplied at a lower cost than a fully featured PC.

On the face of it this is quite an attractive proposition for education because it addresses some of the core problems that plague ICT teams in larger schools.

Windows Management and Application deployment: All the desktops can be deployed from single 'golden' image and applications layered into the image would be immediately available to the users providing a unified and consistent experience for every user.

Desktop Refresh Cycle.  Since the local clients are no longer running the operating system these can be provided as 'thin clients',  allowing ICT suites to be replaced at a fraction of the cost of fitting out the same area with Windows PC's and giving low end devices to be a new lease of life.

Cross Platform capabilities and Remote Access. VDI offers the possibility of running a Windows desktop across a range of devices including tablets and low specification netbooks allowing the desktop to be accessed anywhere - ideal for the new bold BYOD strategy.

So there you have it  - Microsoft Windows finally brought under control as a unified desktop delivered across a range of cheap end user devices to any location, using a solution that has a proven track record in business.

What could go wrong? As it turns out, quite a lot.



The reason why VDI projects often fail in schools can form a pretty long list but here are the main culprits.

Schools normally underestimate the processing requirements for the backend server farm.
Those desktops in the ICT suite maybe old but if you total up all the processing and memory in those dusty grey boxes it still adds up to quite a bit. This has to be matched in the backend server farm because, contrary to the common perception VDI does not give you something for nothing. If the backend is not specified correctly you end up with something that's slower than the original system not faster.  The hardware and licencing for a server farm to support VDI is an expensive investment. Fixing a VDI solution that has been under specified is doubly expensive.

Schools use a lot of multimedia.
VDI is brilliant at a number of things but extremely bad at a number of others and one of those multimedia. There's a simple reason for this. All the desktops are running on a server - a piece of hardware that has been finely tuned over time to provide a fast response for most things but not console graphics. As the students fire up a browser and drop onto YouTube every single one of those desktop images is sharing the same graphic card . Ever since VDI was introduced numerous attempts have be made to work round this limitation (offloading to the client, protocol improvements, specialized graphics hardware) each one adding cost and complexity and reducing the levels of compatibility for the client.  Could there be a more inefficient way of delivering  a simple browser session - I doubt it.

The cheap thin clients were never cheap or thin.
To make VDI work 'thin clients' have ended up being quite sophisticated machines in their own right with prices that are close to low end PC's.  As new form-factors emerge that can provide a fully featured Windows 8 desktop for $170 it's clear that the savings were never really there, especially when the management licence was also taken into account.

VDI is inefficient.
Its important that the VDI server farm is available at all times. Unlike a PC, when a VDI system is down a thin client is just a desktop paperweight. This means building resiliency into the system and there's no way of doing this without providing spare capacity that you end up paying for for - but not using.

Schools use a lot of peripherals.
Over the years VDI has become better at supporting a wider range of peripherals but it's never going to rise to the challenge that education throws at it.  As a result VDI rarely replaces every device in school and generally struggles in  areas such as media, music and technology. Unlike the business world VDI in education will always be a specialized use-case, normally the covering the ICT and LRC suites. By the time you have excluded all the areas where VDI doesn't apply is it really worth it?

VDI is really just a 'band-aid for Windows apps.
Lets be honest the reason why VDI is being introduced in your school is for one of two reasons.
  • A last desperate attempt to move Windows apps into the age of mobile computing.
  • Trying to extend the life of an aging Windows based ICT suite.
The first one doesn't work. Have you every tried using a five year old Windows application on a tablet through VDI? The user experience is horrendous. Wouldn't it just be easier to find find some modern SaaS services to met the requirement.

I'm not sure that any cost analysis would show that adopting VDI long term would be cheaper than adopting SaaS (or a tablet app based solution) and re-equiping the ICT suite with devices like Chromebases.  The cost of the VDI will go a long way towards providing multiple class-sets of modern mobile devices running SaaS applications.

VDI is not a 'hobby' skill.
Managing VDI is not rocket science but it's still a specialist skill that's very unforgiving for somebody learning on the job. Messing up an image for one PC can be overlooked but hosing the gold image for your VDI farm is far less amusing to the teaching team. How many schools can justify the investing in the training required to manage VDI effectively and what happens when that skill walks out the door to a much better paid job in business?

Schools use a lot of software.
In business the average desktop might support a dozen titles on top of Windows Office and most of these are well understood when it comes to integrating into VDI.  In contrast schools use hundreds of software titles, many of them completely unsuited to VDI. After the investment in training the biggest hidden cost for VDI is the time spent packaging software.


The strange thing is that local processing was never a bad idea in the first place.

The best experience of a windows package is gained through a local installation running on a suitably configured machine placed in the hands of the end user.

A class group running local applications will be resilient, scalable, adaptable and remain responsive under load. After all that's how an iPad works and they've been fairly successful in schools.

If you still need to keep those PC's running, SaaS is the solution - not VDI.

VDI doesn't solve an application problem, it solves a management problem but in doing so only creates a raft other issues. So maybe its time to look at the question again and see it there is a better way of doing it, perhaps not using on-premise servers at all.


Related Post January 2017.
Could Google Cloud Platform deliver desktops for Schools.

Tuesday 7 April 2015

SaaS Cost Shock for Schools.


Here's a shock announcement for all schools considering replacing on-premise servers with Software as a Service (SaaS).

A subscription based SaaS service can always be proven to be more expensive than the traditional on-site solution it replaces.

It's not really a shock, more like simple math. If you sign up for a service that costs $1 a year for an infinite amount of time it's going to cost you an infinite amount of money.  OK - that's a trick but you get the idea.

The point is that evaluating the financial implications of adopting a SaaS service requires a different perspective to purchasing an on-site server to run a VLE.

The most obvious point is that, with SaaS the school only pays for the service and nothing else.

With the on-premise approach the consumer must first invest in hardware to host the service along with a covering warranty, a service level agreement (SLA) for the software and an ill defined, fuzzy grey area around backup, configuration, support  and quality of service.

With SaaS the investment covers the service itself, a proposition that's far easier to understand and plan for.

When you subscribe to a SaaS service a service level agreement is implicit in the offering, the two cannot be separated. As an example Google Apps for Education is backed up with 24X7 phone support, 99.9% uptime and ISO27001 certification.  Microsoft Office365 has something similar. For most paid subscription services a failure to meet the SLA may result in an automatic service credit or an extended licence period.

When an on-premise solution is proposed how often are the additional costs factored in? These include utility bills for power (heating up the server) and air conditioning (cooling down the server that you've just paid to heat up).

There are also charges around providing security, backup as well as as the inevitable upgrade cycle.

These are the obvious culprits but the hidden costs are even more insidious. These include software upgrades and the investment in local expertise to manage the whole complex environment.

How many school (and small businesses) have been thrown into a spin when Dev or Sue decides to leave, taking their knowledge of the IT system with them?  Maybe you can fall back on the documentation but pressures on time and the constant upgrade cycles mean that documentation is rarely maintained or kept up to date.

Is it important that the blue box in the corner has a flashing red light, who knows?

A SaaS solution is simpler due to the fact that the customer only has to manage the service and not the supporting platform. It has no flashing red lights. Well it does, but you don't see them and more importantly its not your problem. When upgrades are scheduled they are handled automatically which means the upgrade cycle for SaaS is significantly faster than on-premise software.

Given any requirement with a fixed budget, a SaaS solution will be cheaper in the short term because so many of the services are free and initial barrier to entry is so low. In the long term SaaS needs to be viewed as part of revenue (like water and electricity) which can be turned on and off, not a capital investment project.

At time when schools looking to build a sustainable, long term IT strategy surely answer can’t be to commission more on-premise servers.

That shouldn't be a shock.