Thin client performance in BSF schools

[donaskmi]

I work in a school that uses thin clients for word processing and the like but in the ICT labs we have fat clients as thin is too thin when running Flash, Dreamweaver and Fireworks with Netops in the background.

 

My question is this; How do BSF schools get around the problem of graphics and memory intensive software running on thin client only architecture?

 

 

I am an ICT teacher but years ago I once was a network manager so don't kick me out the forum yet. My school has applied for BSF funding and if given the go ahead we will have managed thin client architecture. Will I still be able to teach with all these programs running?

[CyberNerd]

We run flash, fireworks and dreamweaver without problems on thin client.

 

We are non-BSF and run graphical/memory intensive applications on XP/OSX

[rasssp]

From what I know (Bradford BSF Phase 2) graphic intense applications will be on 'fat' clients. As SUN refer to them.

 

We have a small test group of thin clients similar to how the BSF schools in Bradford are set up.

 

We have Office 2007 and Adobe Master suite on them. That's all they can handle really.

[Butuz]

There are two main things that dictate what kind of software you can run on your thin clients.

 

1) The spec of the actual thin client itself. Less powerful entry level clients have serious trouble animating even simple flash games. Here, it doesn't matter what kind of servers or network you have - any animation will be poor. A much more powerful thin client will run animation/video considerably better - in some cases very acceptably.

 

2) Number of thin clients per server. If you try and run 40 copies of Photoshop on one average spec server you are going to run into serious problems. So put simply you need to find the correct balance of clients per server for whatever application you want to run. If that means running photoshop on thin clients forces you to only have 10 clients per server, then so be it.

 

Thin client devices and servers and software are getting better and more powerful all the time. The old addage that you can only run Office and some Internet on thin clients is now utter nonsense.

 

Butuz

[rasssp]

The guys on here from cutter supply our test kit.

 

The Bradford BSF partner is SUN

 

Thin clients are SUN Ray 270's.

 

The addage may be out of date, but from testing it still seems true.

[broc]

It doesn't matter how powerful the thin client workstation is if the data you need is not local. You are then at the mercy of your network & server infrastructure. It's all down to bandwidth.

 

Compare the speed of access to data stored on a local SATA drive vs a local server running RAID shared by a number of users, then look at using a remote server accessed across shared broadband with that server being shared by multiple users. Make the client wireless and it gets worse again.....

 

In my opinion, there are some end-user applications (video editing comes to mind) which will never be comfortable run from thin clients.

[Butuz]

It doesn't matter how powerful the thin client workstation is if the data you need is not local. You are then at the mercy of your network & server infrastructure. It's all down to bandwidth.

 

As long as the link between your Terminal Servers and your File servers is nice and fast then the network and server inftrastucture speed shouldnt be a problem. Infact it should be less of a problem on thin client than it would be on fat client as you won't have fat clients all over the school dragging massive amounts of video data round, and also your terminal server should have considerably faster disk access than a workstation, particularly for reading data.

 

, then look at using a remote server accessed across shared broadband with that server being shared by multiple users. Make the client wireless and it gets worse again.....

 

All true. Anyone expecting to be able to run any kind of whole school network (thin or fat) through LEA/RBC provided 100/10Mbit broadband connections is going to be in for serious dissapointment. Frankly in my oppinion wireless is not worth using either. I don't have 1 WAP here.

 

In my opinion, there are some end-user applications (video editing comes to mind) which will never be comfortable run from thin clients.

 

True.

 

Butuz

[torledo]

There are two main things that dictate what kind of software you can run on your thin clients.

 

1) The spec of the actual thin client itself. Less powerful entry level clients have serious trouble animating even simple flash games. Here, it doesn't matter what kind of servers or network you have - any animation will be poor. A much more powerful thin client will run animation/video considerably better - in some cases very acceptably.

 

2) Number of thin clients per server. If you try and run 40 copies of Photoshop on one average spec server you are going to run into serious problems. So put simply you need to find the correct balance of clients per server for whatever application you want to run. If that means running photoshop on thin clients forces you to only have 10 clients per server, then so be it.

 

Thin client devices and servers and software are getting better and more powerful all the time. The old addage that you can only run Office and some Internet on thin clients is now utter nonsense.

 

Butuz

 

 

Sorry for the long post chaps, but i've been thinking a lot about this recently and i want to get most of my thoughts down rather than start a new thread.

 

@butuz - Would those thin clients that handle media better be using some form of multimedia redirection or even offline use on the client to process the applications locally ?

 

My understanding is that for 80% of workloads a thin client or VDI solution is sufficient......to an extent you are the mercy of the network infrastructure, but then i guess that's why you'd have some applications installed locally on the TC, such as Ie/Firefox, JRE, Windows media player etc so that you can atleast get basic web browing and a browing experience should the TS farm have issues. That would be the fallback scenario. But then your thin clients are looking a little fatter, and they still need to be maintained and patched.

 

As for the photoshop example, is it really acceptable to have to make the investment in more servers, or have to provide dedicated resources to SBC for intensive applications just to accomodate a modest number of users who would still not get as good an experience as if it were running locally.....the result of such a decision would mean reducing the ability to virtualize those servers to get the maximum investment - you'd have to hand all resouces over to serving photoshop apps. It's a balancing act, it's why SBC depends on the environment where it's going in....it's not the be all.

 

VDI seems to be the halfway house between true virtualized clients and traditinal SBC like terminal services. Not all applications can run in a TS environment, and the possiblity of streaming an entire, customizable desktop to a user seems to be what's encouraging it's use. But there still seem to be a lot of compromises, how fat does the thin client need to be just so that you can run 80-90% of workloads ? going back to the multimedia aspect, i notice one company provide a software solution where the local media apps handling more intensive mm on the thin client can be integrated into the virtual desktop view to make the interaction seemless to the user. It disturbs the user experience to have a virtual desktop for most tasks but then to have to switch to local apps for other tasks.

 

I personally believe that embedded virtualization on the client is the end game for true flexibility and ease of mgmt....that in combination with an OS/App streaming solution will offer a true virtual desktop experience.

 

citrix and intel have launched an initiative to do just that using the starting point of an embedded hypervisor in conjunction with intel vpro and citrix desktop provisioning solutions.

 

Citrix, Intel Collaborate On Desktop Virtualization - Hardware - IT Channel News by CRN and VARBusiness

 

and becuase there is no exclusivity in a solution [i.e intel could work with other players to leverage a differnet hypervisor and vpro] such as this i expect it to catch on in a big way with other players.

 

For the time being i question whether you can justify the infrastructure investment requirments for a complete SBC or VDI solution when there are so many issues to deal with. It's very nature indicates a substantial initial cost outlay followed by a lack of flexibility to a degree. Sure, you've got power savings but at what cost ? End user experience ?

Edited January 23, 2009 by torledo

[donaskmi]

Thanks for your very prompt and in depth replies - It seems to be dependent on the very many ways of implementing the architecture that will make or break the running of mem intensive software.

 

We try and lead our ICT teaching here on the latest technologies but from what you guys are saying it looks as though its down to how Sun / Bradford partnership implement the architecture. At the moment it looks like we'll be fully BSF'd in three years. It would be interesting to know how present BSF schools get on running CS3 suite etc on the platforms that Sun have implemented? - Have SUN installed "fat"er clients for ICT labs and how well do they run? Does the school have to specify its ICT needs in the process or do I assume that SUN / Bradford have it all to hand? Does anyone know if they are running thin clients in the academies?

 

Thanks for those that have commented already.

[broc]

Thanks for your very prompt and in depth replies - It seems to be dependent on the very many ways of implementing the architecture that will make or break the running of mem intensive software.

 

Don't get confused between memory intensive & data intensive; depending upon architecture, there's nothing to stop you loading up a thin client with fast memory & lots of it especially if you have a 64bit O/S; memory intensive should not be a problem for a thin client. The crunch is where you have lots of data outboard of the client, where you may be bandwidth constrained reading/writing the data.

[johnwinkley]

My first post on edugeeks, having lurked for a while, firstly just wanted to say thanks for the excellent information that the community provides. More than once over the last six months I've found answers on here that I was struggling to find elsewhere.

 

I have been tracking the discussion about thin client for a while now - I'm involved with a couple of schools in Bradford who are likely to be moving to the BSF solution shortly, which as others have said is a largely thin client (mix of wired desktop and wireless laptop solutions).

 

I have a couple of questions:

 

1) Graphically intensive rather than "processor intensive" applications are increasingly important to general ICT use in schools, I think. While a small but growing proportion of use is in power applications like photoshop, it's the day to day youtube/newsvideo sites that sees broad curriculum use. My limited technical understanding and research suggests that there are technology improvements coming which reduce the streamed bandwidth from servers to thin clients for this type of content, and use graphics processors in the thin clients to render the video. Anyone know of a good site with a backgrounder in thin client where I could do some reading up?

 

2) User acceptance testing of thin client seems to be a grey area. I've struggled to identify any tools specifically designed to measure performance on thin client networks/clients, or indeed to find any specifications for what reasonable user performance requirements might be given that the nature of network traffic on thin client is different to "traditional" networks. Does anyone have any experience in this?

 

Thanks, John

[wesleyw]

I would have thought that ICA protocol using 20kbps should be able to run 500 users on a 10mbit line concurrently so 5000 with a 100mbit connection. Now the problem will be whether or not the 10/100mbit line to the school is contended or not? If not then it will be a serious problem. But there is always BT Net.

 

 

Wes

[cjohnsonuk]

My thoughts on this:

 

1) As we don't know what hardware we will be using in the next 5 years we should really try to separate application delivery from desktop hardware so that we have maximum flexibility. With clients for iphone, blackberry and all major OS's solutions like Citrix will allow more flexible teaching by allowing a wider choice of devices to be used.

 

2) With many authorities now having extensive interschool networks (IIRC most of our schools are now linked with 100Mbps lines) the ability to centralise services and benefit from economies of scale should allow for the option of schools grouping together to purchase and deploy/deliver software across locations. The more forward thinking authorities may already be thinking of extending IT services such as account/directory management, data storage, email and applicaiton delivery to schools not included in the BSF program such as Primary schools. Many Primary schools don't have a full time technician and moving the aspects of software delivery requiring technical support offsite to a data centre where they can be properly managed seems a great benefit to all with everyone benefiting.

 

3) With anytime anwhere learning being such a big focus and MLE/VLE being the now "not so new" buzzword we're providing all the data and instructions for students wherever they are but there seem to be very few solutions that also provide the tools to carry out the work. Arguably people like Google are doing more with Apps to enable anywhere learning than the traditional content based VLE service providers. Being able to deliver applications into student homes (and management apps into teachers homes) does two things: 1) it provides access to a standard and familiar set of software tools for all students regardless of their ability to afford to purchase it (and so also reduces the temptation to pirate it) and 2) it keeps the data centrally so that is properly backed up and there is no more excuse of "the dog ate it" or of teachers leaving their laptops in the back of taxis. OK so not everyone has internet access yet but it brings great benefits to those that do.

 

There is still a place for fat clients for specific applications such as video editing. For such data intensive applications there probably needs to be specific data storage provision (cache?) as well, local to the school if not local to the suite. However to minimise the administrative burden of managing the fat PCs the "run of the mill" applications can still be delivered to these desktops using thin client technology once again providing schools with the greatest flexibility in their choice of hardware and helping standardise the environment so that students can focus on learning curriculum not how to use IT on this particular computer. So Macs runnining iMovie and MS Office 2007 (as a remote app) as (opposed to OSX MS Office 2008) would not be an issue.

 

Anytime anywhere learning: tick

Addressing the digital divide: tick

DPA issues addressed by proper data control : tick

reliable, resiliant, flexible : ticks

 

Now all I have to do is persuade everyone else that I'm right

[cjohnsonuk]

Anyone know of a good site with a backgrounder in thin client where I could do some reading up?

 

Many thin client now pull the uncompressed media to a local cache and decpompress locally and reintegrate it back into the desktop image. This way there is no more lag than if a fat client PC had downloaded and displayed the image. Citrix are doing this now, NEC were showing off something like this (thin client with full screen video) a couple of years ago. Thin clients are getting fat enough to be able to have their own codecs for media files and decoders for flash etc.

 

Regarding performance I'm, always amazed at just how quick MS word launches on our thin clients which are 4 years old and running off one server. That's 80% of your IT users happy 80% of the time.

Edited June 11, 2009 by cjohnsonuk
forgot to add this bit

[cjohnsonuk]

Don't get confused between memory intensive & data intensive; depending upon architecture, there's nothing to stop you loading up a thin client with fast memory & lots of it especially if you have a 64bit O/S; memory intensive should not be a problem for a thin client. The crunch is where you have lots of data outboard of the client, where you may be bandwidth constrained reading/writing the data.

 

Forgive me if I've missed your point here but the network between the data store and the thin client server is going to be much easier to beef up than the network between the data store and every desktop seat. Only the screen updates go to the client, the data only goes as far as the thin client server so for data intensive applications the large quantities of data never get dragged all over the network slowing everyone else down, only between the two servers, something that can be addressed within the confines of the server room and which would not affect anyone else's network performance.

 

Or did I miss something

[Butuz]

Or did I miss something

 

Nope!

 

Butuz

[webman]

This is all well and good; but I can see two pretty big problems, and that lies with the WAN link between your thin client terminals and your data centre. When it fails or performs sub-optimally, the school loses everything - curriculum, admin, communications - the lot. Even worse - if the data centre itself fails (or its section of the WAN), then schools in the whole county lose everything.

 

Another question to raise over centralising this sort of thing - who owns the data? The school? The LEA? The managed service?

 

It also makes for a pretty large attack vector. It only takes one vulnerability to potentially leak a whole lot of data to the outside world, or let a whole lot of viruses/trojans affect the whole network, again - potential to take down the whole LEA education infrastructure.

[PiqueABoo]

Reality check: "Primary schools record an average bandwidth of 3.2Mbps, which is expected to increase to 4.1Mbps by April 2009. For secondary schools the levels are expected to increase from an average of 15.7Mbps to 21.1Mbps by April 2009."

[SYNACK]

I would have thought that ICA protocol using 20kbps

 

Only if you disable sound, high resolution, high colour depth, printer and local device fowarding etc. This figure is unrealistic and as others have said it add a huge single point of failure. Add to that that most school connections are rubbish and highly contended anyway and you are asking for trouble.

[Butuz]

The thing is - if your centralising the infrastructure across the authority, the cost savings in initial purchase cost, replacement lifecycle, electricity usage, support costs, by replacing thousands of 100watt fat clients with 6watt thin clients will give the LEA an enormous pot of money with which to spend some of it on the WAN infrastructure. I would expect dual redundant 100mbit lines at least.

 

Large multinational companies do this kind of centralised consolidation every day and have been for the last 10 years. It works, and with recent advancements in server hardware, virtualisation, and improvements in desktop and app delivery, it will continue to work even better.

 

Butuz

[webman]

Reality check: "Primary schools record an average bandwidth of 3.2Mbps, which is expected to increase to 4.1Mbps by April 2009. For secondary schools the levels are expected to increase from an average of 15.7Mbps to 21.1Mbps by April 2009."

 

You didn't specify a source or a context for that quote, so this might help others. C3 Education/BESA report, page 4.

 

That is at the moment for using the internet to browse websites; not for running the whole computer network continually throughout the day over thin client protocols.

 

Each school might have the necessary bandwidth; but will the WAN and the server infrastrucutre at the other end be able to cope with hundreds of schools? Are the RBC and managed service reliable enough to sustain the required throughput and performance?

[torledo]

The thing is - if your centralising the infrastructure across the authority, the cost savings in initial purchase cost, replacement lifecycle, electricity usage, support costs, by replacing thousands of 100watt fat clients with 6watt thin clients will give the LEA an enormous pot of money with which to spend some of it on the WAN infrastructure. I would expect dual redundant 100mbit lines at least.

 

Large multinational companies do this kind of centralised consolidation every day and have been for the last 10 years. It works, and with recent advancements in server hardware, virtualisation, and improvements in desktop and app delivery, it will continue to work even better.

 

Butuz

 

redundant WAN links won't make a difference in the event of some sort of datacentre failure. Even for all the talk of DR sites in central datacentres, the reality it isn't always a straightforward failover back up and running in no time......and it can take quite a lot of time to get all services back up and running....a service provideer might have a recovery time objective of a couple of hours but how often has the entire DR process been tested since it was implemented for example.

 

Throwing millions at a problem should in theory allow you to be better prepared, but centralized infrastructures tend to be pretty complex beasts, added to over time to increase complexity, and all sorts of change procedures and policies in place. And $h1t happens after all.

[Butuz]

Good points.

 

I would say that if you are providing centralised ICT services for a whole LEA, i.e all primary, secondary, special, librarys etc you would definatly want at least two redundant datacentres. Modern SANs and Vitualisation make redundant datacentres far cheaper and easier to set up than they once were.

 

All eggs in one datacentre is never a good idea! Even I am trying to get a second server room set up in my school which can act as a redundant backup incase the main server room should ever somehow be destroyed.

 

Butuz

[bossman]

@cjohnsonuk:

 

HDX MediaStream (SpeedScreen Multimedia Acceleration) is what I think you are talking about.

Can be read about here:XENAPP SNEAK PEEK - HDX MediaStream for Flash Tech Preview - Juan Rivera - Citrix Community

 

enjoy.

[bossman]

@Butuz:

 

Well let's take it a step further and save even more money, lets outsource to another multinational in a different country and let them run our educational IT systems at 1 tenth of the cost of running them in this country.

 

We might as well cut out the middle men (LAs) and save heaps in making all the IT staff in education redundant just think of all the money we can save and the misery we can inflict on people all in one go. :)