My conclusions on VDI and other things

I would like to pass onto you my experiences of running VDI in school over the past two and a half years. I do have to say this has not been a pleasant journey and I do have to question my own motivation for promoting the solution in the first place. In retrospect I believe I was motivated more by a desire to try out the new technology rather than what is/was in the best interests of the school. This soiree into VDI has come at a price to the school both financially and in terms of attitudes to new technology.

The solution chosen was VMware View on 100 Wyse thin clients the details of which I have posted in a previous thread (http://www.edugeek.net/forums/thin-c...structure.html). After initial excitement and subsequent disappointments I have finally concluded that, certainly in our environment, VDI is not worth pursuing. This decision is based not just on cost but more importantly on the usability of the solution. Obvious questions spring to mind; was the network fast enough? Was the back end sufficiently powerful? Were the thin clients powerful enough? Was the software setup correct? Were effective trials done prior to implementation? Did you consider other VDI solutions (bear in mind this was the end of 2007)? The answer to all of these I believe is yes. I am happy for anyone to challenge me on this in the interests of moving forwards but numerous periods of discontinuity experienced by staff and students have taken their toll and the decision to move away from this solution, already taken.

If you were to consider implementing a similar solution the following is a breakdown of the approximate costs:-

Clients

Wyse V10L - £280
19" monitor - £85
Keyboard & Mouse - £15

Supporting back end software

VMWare view Premier 100 pack + 3 years SNS - £11,416 + £6,095
Microsoft VDA licensing (needed to run MS OS on a thin client with non-MS OS on it ie Thin OS) - £0.58 per month per client. Based on 100 clients for 3 years (this is a non-perpetual license) this gives - £2088

Back end hardware.

Let’s say 2 ESX servers and a SAN storage solution - 2 x £7,000 + £8,000

Total cost - £79,599. Giving an annual cost of £265.33 per unit

To run for a further 3 years, assuming none of the physical kit breaks down and we do not extend the warranties or replace anything, this would cost an extra £8,183. Annual cost per unit of £146. If we were to factor in backend replacement and warranty extension then this figure would be higher probably in the region of £185 pa per unit.

To give some perspective to this, 5 years ago I purchased 150 Dell GX520s with 2Gb of memory and a 180 Gb HD at a cost of £450 per unit (monitor, keyboard and mouse included). These units are still in circulation. Annual cost per unit - £90.

Are there advantages in running VDI? Yes certainly they address issues of energy efficiency, reduced carbon footprint and ease of management but the cost in terms of usability and fitness for purpose for us was too high. This is not to say we are abandoning virtualisation, quite the contrary, we have made huge strides in reducing our carbon footprint over the past 4 years by virtualising almost our entire server side environment. This has been incredibly successful and I would recommend that you do this as soon as is practicable as it makes you more agile and frees you from the constraints of the hardware you sit on.

So what are we doing with the thin clients and the back end hardware? Well we have refactored two of the ESXs as HyperV servers (those who know me realise just how big an issue that one was!) in a cluster with shared storage and are running virtualised Windows 2008 R2 terminal services on them. They seem to work OK on the Wyse V10Ls no issues with printing, or USB (no need to pay for TCX licensing). Yes there are limitations with some software not running on them but then I'd rather have that than have machines that no-one wants to use. Why have we done this? Basically to make the best of a bad decision and oh yes, cost. The rest of the kit is being used to host just 30 VDI machines that staff use to remote into school. This is the only part of the solution that was successful. If you want to try out VDI then I suggest starting with this (the starter pack + 20) and Ericom are good external access solutions. In the end the decision to use VDI is yours but please be warned you will have to live with this decision for a long time.

What are the lessons I/we can learn from this:-

1. Never let just one person make a decision on the choice of an IT solution. They will always be influenced by their own needs and aspirations. A group decision is always much more objective. Ask people in other schools who have the solution in place, go and have a look. I have personally visited over 50 schools in and around my region, some good some bad, but all with something to offer.

2. Be very wary of being at the forefront of new technology it can reap great benefits but it can also set you back a long way (you only have to look at Sun's attempt at entering the managed service market in education to see that one). You have to ask the question "Will implementing this technology really be of benefit to the school in the long term?"

3. I was wrong. This was a hard one for me to admit but it has to be said. However all was not lost. We have learnt a lot about the technology (and ourselves) and more by luck than judgement we have subsequently made good use of the equipment purchased in the pursuit of technological advantage.

On a completely different tack, if I can indulge with the distillation of my knowledge of IT gained over the past 25 years. Some of these statements may seem brief, obvious, crass and bold but believe me behind them are many years of sweat and tears. Take them or leave them as you wish.

• Buy Dell Optiplex small footprint PCs (GX520, 745, 755, 380 etc) they last 5 years, are quick, robust, parts are cheap (in years 4 and 5)

• Reduce your laptop count, they are a liability.

• Don’t give teaching staff laptops to take home. They don’t use them and it takes money away from devices that can be used by the students. (Yes, I was a teacher)

• Buy the Casio AJ-A130 projectors, no filter, long bulb life.

• Check the alignment of your Smartboards. If they are out and cannot be realigned effectively, get Stelgis to check them even if they are out of warranty. Apply pressure and they will replace them.

• Virtualise your servers with VMWare. Not cheap but definitely the best.

• Use Veeam Backup never goes wrong does what it says on the tin.

• Buy big brand name servers and storage. We use IBM for ESXs and SAN. Switched on 4 and a half years ago, never gone wrong. Best decision we ever made (unlike VDI)

• Use APC 3000 UPS best compromise of cost and effectiveness

• I usually suggest buying perpetual licenses and hate Microsoft but the new Campus agreement due in March may change all of that. Give it a look.

• Never dismiss anything until you’ve tried it.

• Never spend more than £1,000 on a piece of software unless it will be used by a lot of staff. Make sure your IT user group is consulted first.

• Spending more than £500 on helpdesk software is a waste.

• Buy HP managed switches eg 28/2900, 4200, 5400, 3500 series not cheap but do have a lifetime warranty.

• From above Procurve manager plus switch monitoring/management software, excellent.
• Get you fibre checked for problems once every 5 years.

• Upgrade Office and Windows once every 7 years. Unless it’s Vista like in which case wait another couple of years.

• Give System Centre Configuration Manager a try. Makes OS and software deployment so much easier. Yes it’s a bitch to setup, but definitely worth it.

• VMWare ThinApp (Application virtualisation). OK but far too expensive to maintain (high SnS costs). After 2 and a half years of using it, I would not recommend it.

• Stick with technologies you know or have seen working well. Remember ask the kids for their opinion if you are visiting. You’ll always get an honest answer.

• Spend money on solutions that add value to your implementation. As an example the solutions offered by Richard Willis (SalamanderSoft) help integrate the information in MIS into the IT solution. They really come into their own when you have your own AD, Exchange and Sharepoint. Excellent suite of products.

• Use Smoothwall – great product and don’t listen to the rubbish about it not working in a virtualised environment, it does.

• Don’t let staff take sensitive documents off site. Get them to use remote technologies (VDI, Ericom).

• Don’t spend a fortune on a VLE implementation; it’s a waste of money. The issues of VLE use are not necessarily technological but are to do with the mind set of staff. Money is best spent on training.

• Bit of a contentious one but, solutions bigger than a school don’t work. I have seen so many large solutions fall into oblivion over the years, why? Well the only conclusion I can draw is a lack of ownership and support. At least at a school level the ownership and support are there and there are some people who give a damn.

• Always treat schools that blow their own IT trumpet with suspicion. After visiting many of these, I have always been disappointed.

• Stay in close contact with schools in your area. This has two benefits:- first you know where you are in the scheme of things and secondly you can share ideas, costs, problems etc.

• Purchasing huge amounts of kit to reduce balances causes problems 4 to 5 years down the line. This really needs to be hammered home to SLT.

• Related to the above, computers don’t last forever.

• When trying new technology/software do all the testing and then leave/forget it for at least a month. This will get the initial excitement out of the way and you may be able to make a more objective decision. Also incude several members of staff in the testing. If no-one has talked to you about it during that month then it's not worth continuing. We have been doing this for the past 2 years and it has cut down the amount we spend on departmental software by 40%

• Finally one for SLT. If you pay peanuts you get a monkey.

I hope this will be of help to you and that you can avoid the pitfalls highlighted in this rambling.

Dave O

Dave,
Thanks! Lots of good advice there, stuff to make you think before jumping in. And what I know about I agree with. Just a little clarification please: "Purchasing huge amounts of kit to reduce balances..." - do you mean costs?
Cheers,
Ian

thanks for sharing. I know you've given a breakdown of the cost based on 100 units eg. x100 view licenses, but -and sorry, i haven't done the maths- does that mean you rolled out 100 VDI clients or was it a whole school move to VDI ?

That really would have been a substantial leap. Even 100 clients is quite a lot to have started out with, i guess you did discover in the end that you had more success with a specific usage scenario [the 30 VDI clients for teacher remote access]. I guess maybe it's sometimes difficult to justify a significant spend on server/SAN equipment if it isn't immediately servicing 100's of clients, but it's a real shame that the concept of starting small and experimental and having the headroom to scale up more often than not justifies the initial outlay. Or if you found it not ideal to scale to more than 30 VDI no reason why the rest of those backend resources should remain idle. That's the power of server/SAN virtualization, it's not about showing how bleeding edge one can be with presenting desktops. :)

i think what a lot of people who've gone down the VDI route may have discovered is that a lot of what they were trying to do can be accomplished with more traditional, and mature terminal services solutions. Plus the newer complimentary products like software streaming. That's not to say a move toward server virtualization and *some* thin client device rollouts is futile. From a centralization of resources and energy efficiency perspective there are clear benefits. But whole school rollouts and a complete dependency on server based computing is another thing entirely.

Ian

No. What I am eluding to here is something which many schools face or have done in the past. For example you get close to the end of the financial year. Budget projections have been poor (not pointing fingers here but LAs, SLT and finance really need to get their act together). You are left with a balance which for all sorts of reasons cannot be carried forwards eg its represents more than 5% of the budget and must be spent before 31st March. Who do you call?? IT because we all know who can get rid of huge amounts of cash quickly! So unless you've got your long term plan in place, panic buying ensues and you end up with kit you don't really want or need at this time and a headache in a few years when it comes time to replace. You then have to deal with comments like "We gave you £400,000 four years ago what more do you want?". Someone who can actually manage a budget is the simple answer.

Dave O

Thanks, Dave. Yes, been there, too - in fact nowadays we plan for that scenario and have a few sensible 'wants' in our back pocket just in case!
Cheers,
Ian

Yes it is based on a roll out of 100 clients. We have approximately 800 PCs and laptops and this therefore represents only 12-13% of our estate. God forbid we do all of them!

At the time 100 did seem a reasonable amount to test on and if you read my previous post you will see that we did not spend anything like the figures it now costs but I did not want people to think it is in any way costs effective (for 100 clients) certainly over less than 8 years and how many IT solutions do you know that last 8 years or do you want to last 8 years. The extra SAN and ESX/HyperVs were purchased in the salad days of Harnessing Technology and as they are housed in a different location to the server side give us some extra resilience for DR. So not a total disaster. If the solution had worked then yes we would have scaled out and probably reduced TCO. We have not abandoned thin clients totally we will still be using them for terminal services but eventual replacement will be with a PC unless some killer technology comes along in the next 3 years. To be honest the reduced carbon footprint is a fallacy as the back end pumps out enough energy to make up for the savings at the client end.

I totally agree with your final paragraph and I hope that my post will at least alert people to these issues.

Dave O

Fantastic post, Dave ... frank and honest, and I wouldn't mind a chat with you sometime to cover a few other things that get left unsaid about attitudes to technology with staff, SLT, students.

Hi Dave,

Fantastic Post There, I have sent you a PM as it would be nice to chat with you further about the above. just to try and understand where the problems were, and what kind of image you had deployed to the clients etc. I just want to try and get a view for the situation you were in. (for further reference more than anything).

I am not saying that you are wrong in what you have said, you have seen it first hand yourself. I have had a FULL ICT IMAGE Including Adobe Fireworks etc. running on VDI without any issues at all. You may know it already but the end expirience generally lyes in the way the Client Template has been created and configured.

James.

Quote:

---

Originally Posted by Dave_O

Ian

No. What I am eluding to here is something which many schools face or have done in the past. For example you get close to the end of the financial year. Budget projections have been poor (not pointing fingers here but LAs, SLT and finance really need to get their act together). You are left with a balance which for all sorts of reasons cannot be carried forwards eg its represents more than 5% of the budget and must be spent before 31st March. Who do you call?? IT because we all know who can get rid of huge amounts of cash quickly! So unless you've got your long term plan in place, panic buying ensues and you end up with kit you don't really want or need at this time and a headache in a few years when it comes time to replace. You then have to deal with comments like "We gave you £400,000 four years ago what more do you want?". Someone who can actually manage a budget is the simple answer.

Dave O

---

i've been in that position, and while it can seem like all your christmas's have come at once, it really does stress and panic you. Particularly if you start having to condense your long-term development aims into a few days or weeks and rush getting pricing and availability.

often development plans are developed with a view to what may be available down the road, but what isn't necessarily available now. And there's nothing like buying part of a solution knowing to complete it properly will require funds that won't be available 12-18-24 months hence. That's if what you've bought hasn't already been superseded by then. oh joy.

VDI strikes me as exactly one of those projects where you have your eye on it as a medium term project with a view to completing once the products were more mature, but you know you'd end up biting the bullet of a v1.0 solution to get money spent by a deadline. Not saying it's ever a panic buy, but that it's worth a punt on if you haven't got the time to be that methodical and can afford to go a bit bleeding edge.

And well, sales bods have been calling with the usual patter about how they've rolled it out here and there.... so why not. Even though you know it just sounds like way too much faff and in truth compromises, to get what it boils down to... is a desktop on a screen. Or the soliutions as were, in 2007. Things have probably moved on a bit since then.

Quote:

---

Originally Posted by EduTech

Hi Dave,

Fantastic Post There, I have sent you a PM as it would be nice to chat with you further about the above. just to try and understand where the problems were, and what kind of image you had deployed to the clients etc. I just want to try and get a view for the situation you were in. (for further reference more than anything).

I am not saying that you are wrong in what you have said, you have seen it first hand yourself. I have had a FULL ICT IMAGE Including Adobe Fireworks etc. running on VDI without any issues at all. You may know it already but the end expirience generally lyes in the way the Client Template has been created and configured.

James.

---

maybe you could share information on what setup worked for you ?

James

I am quite happy to discuss this or any other issue with you. I love the "I am not saying that you are wrong...". Come on of course you are and that's what makes these threads so great. OK I'll bite... What technology are you using? What spec is the LUN your master/gold image is on? We can certainly go through all of this but to save time. I have over the past 2 years followed every best practice published by VMWare, changed hardware (tried stuff that you only really seen on the starship enterprise), used an image with only XP SP3, office 2003 , Smartboard notebook, used different types of broker changed the configuration of the LUNs used SAN accelerators. I finally ran out of ideas. On the plus side what I don't know about VMWare View is nobody's business. So if you can show me an implementation of an equivalent size to the one I have, running on a normal school day that does not have a Cray as the back end serviced by solid state drives and yes I tried them, then I'm prepared to eat my words.

Dave O

:) Ok Maybe I am finding a little hard to believe,

I'm going to struggle to remember all the information as it was at my previous place of work, first of all we were not using VMWare Solution we were using SUN/Oracle Hardware and Software (1 SAN + 3 Servers (2 for VDI and 1 for SGD) and the environment was to provide only 40 Desktops to Sun Rays.

I find it hard to believe that the image would not cope with XPSP3/Office2003 at the moment even though i am not there i know it is running XPSP3/Office2010/Adobe Software + Others. and the client machines were only assigned 1024MB RAM Each but alot of tweaks were done on the image itself i.e. Theme Enhancements / Disable Indexing / Paging File / System Restore Disabled etc.

Andy/Kim @ CutterProject will be able to go into more detail as to how it was configured as they would have it logged somewhere, + Andy did the initial configuration on the Solaris Boxes and Storage.

James.

Couple of things I missed off the original post (senility setting in)

I know some people are interested in this type of informaion so, based on 1650 staff and students, 800 PCs (life expectancy - 5 years), laptops (4 years), 70 printers (4 years), 90 IWBs (God knows), projectors (4 years), VMWare SnS (far too expensive), other maintenance stuff (very small amount as we have perpetual licenses) toner costs (reduced from £26,000 pa to £6,000 pa by implementing quotas. The funny thing here is when people ask for more we just give them so there's no real limit!) and LA broadband costs (there's another story altogether) we have spent on average £140,000 pa over the last 3 years. We have been able to smooth this out over the past 5 years due to refusing to panic spend (see original post) at the financial year end. We are projecting expenditure of £150,000 pa over the next 3 years just to maintain the status quo. Hope this helps give some perspective on where we are and what we have got. I do have to say that after years of me bleating on about the need to continually reinvest, the message is getting through, shame its come at a time when there's no money!

Dave O

James

So no real life functioning solution in a school context.

Dave O

Quote:

---

Originally Posted by Dave_O

James

So no real life functioning solution in a school context.

Dave O

---

Yes, The above is in a Secondary School and is still there at this moment in time being used by students and staff daily. (SGD is being used as External Access for Teachers)

James.