Published by Robert Pogson February 20th, 2010
in Uncategorized.
I read a decent article about adoption of thin clients in education and in the middle of it all, I read, “At best, the upfront capital cost [of thin clients] is 5 percent cheaper, and at worst, it’s a wash,” Sloan says.” Of course this follows a report that a school spent $15000 for a thin clients solution equivalent to $25000 worth of thick clients. I guess analysts have to make a living, but you would think they would reflect reality. The only way thin clients are going to break even with thick clients in capital costs is if you use that other OS and HP clients and servers. If you use $100 thin clients and good used or purpose-built equipment, you are laughing. At least Slone recognizes that thin pays in the long run through lower maintenance.
Look at some software costs:
|
TOS |
GNU/Linux |
| Server Licence |
$1000 |
$0 |
| Client Licence |
$40 |
$0 |
A server for AD/file/print with 2 gB RAM can handle 20 users with GNU/Linux, so the “extra” cost of using GNU/Linux terminal services is -$1800 . Seems like a good deal to me. Sizing the server reasonably scales out a long way. I budget about $25 per user on the server as I save more than $100 on the client hardware because of smaller case, CPU, memory, power-supply and no drives. I can run gigabit/s on CAT-5 if needed so the cost of network upgrading is minimal on any system wired in the last ten years. Take that, Sloan.
The analysts can say what they want. We can figure it out.
“Worldwide thin-client sales grew from 2.9 million in 2008 to 3.4 million in 2009, a 17 percent increase.”
Source: IDC
Although only 1% of PCs shipped are thin clients, a lot of thick clients are being re-purposed and new thin clients last a long time because of no moving parts. 17% growth in a down economy must mean something, eh?
- Robert Pogson
Published by Robert Pogson February 20th, 2010
in Uncategorized.
Over the years I have found the acceptance of GNU/Linux thin clients good because of the increased access (more seats) and increased performance (responsiveness of servers v thick clients). Still, there are many who have not seen this and question the acceptance of thin clients for education or business. I found an article reporting on the results of three tests of acceptance of thin clients in three different scenarios in an academic environment. The third trial, which I consider definitive, inserted thin clients in amongst PCs and provided identical logins and desktops from M$’s terminal services. Thus, the users were blind to the use of a thin client. The machines looked like PCs for the most part and booted PXE etc. and used RDP. The result was 92% acceptance and only the performance with USB was noticeably slower with the thin clients. USB2.0 v 100 megabits/s may, indeed, be noticeable but there are many environments where USB is not an issue at all. USB 2 & 3 are faster than even gigabit/s networking. This would mostly be a consideration for large documents and files rather than smaller, more commonly encountered documents.
Compare that acceptance with the increased performance from a well-endowed GNU/Linux terminal server and there are many good reasons to use thin clients in education. The study noted that users preferred the boot-up of the thin clients because it was faster. Other advantages possible are the ability to leave a session and come back to it later even from another station. This is wonderful for students who have to move around on schedule and may find themselves closer to a different machine when next they are free to use a computer.
For the most part, I have replaced old thick PCs with new servers and thin clients. There is no clinging to the old ways from that perspective. It is just unreasonable to assume any non-profit organization has the ability to replace old PCs with the state-of-the-art new PC periodically to stay up to speed while they can upgrade a few servers for much lower cost. My cost of server per user is about $25 these days, not the $100-$500 cost of some PCs. For that I get the advantage of huge RAID, RAM, multiple cores and gigabit/s networking. I will give up sluggish USB to get those more frequently needed resources. If there are some users for whom faster USB is important they can use thick clients. It should be a minority in most schools.
- Robert Pogson
Published by Robert Pogson February 20th, 2010
in Linux in Education and technology.
I spent some time studying the information about virtualized desktop PCs on NEC’s site. They describe the configurations and advantages and disadvantages very clearly. It is worth visiting if
you are curious about how they do it.
I was puzzled by several things:
- NECs graphics suggests the “initial cost” of a virtual PC setup is less than the cost of doing the same thing with thin clients and that the regular business PC had a lower initial cost than even a thin client
- NECs graphics suggests the power consumption of a virtual PC setup is less than the power consumption using thin clients
I am assuming that ordinary thin clients connected to a terminal server are the basis of this comparison. As the server for the virtual PCs has to have more resources per PC than a terminal server, I do not understand how this can be. Each virtual PC has RAM for the guest OS in a virtual machine which amounts to 256MB or more these days. That memory multiplied by hundreds or thousands of units amounts to something. Similarly, one needs more servers to run the same number of seats this way with a given amount of RAM so I cannot see how the power consumption can be less unless that depends on maintaining a high load factor dynamically. Being able to slide the session to another server “live” is an advantage and could cause power savings.
As an example, suppose we have a server with 64 gB RAM and we allocate 1 gB per virtual PC. We can then run 64 users and their virtual PCs on the one server. As a terminal server, we could allocate 1gB to the OS and use 128 MB per user for their data and put 504 users on one server for the same power consumption and initial server cost. That’s how it might work with a ‘NIX OS with shared memory. If one does not have the ability to share resources like libraries and executables, they are stuck. That is what they are doing, using XP Pro. We might not stick 504 users on that server, but we certainly could run far more than 64 average point-click-gawk users on it.
It could be the licensing costs for that other OS defeat the advantages of thin clients for that other OS. A hefty licence for the server, a CAL for each seat and a licence for the OS on the terminal means you are paying three times for the same thing.
Here is a prime example of the burden M$ puts on IT. If you use M$’s software, you get no benefit from the new technology. (Thin client is not new but 64gB on a server is something affordable these days.) I see this all the time in schools. Add up all the RAM on the thick clients and you can run far more users than if it were on a GNU/Linux terminal server instead. The last place I worked had 1 gB per XP machine. That is 24 gB in the lab. I could easily have run the lab on 2gB on the server. RAM is not that expensive but every server has its limits and a server with double the slots is much more expensive.
NEC, sadly, tries to discourage potential customers from using GNU/Linux instead of welcoming them to twice the benefit from using GNU/Linux and thin clients instead of that other OS.
Perhaps for some customers this seems like a good deal but while it may reduce some operating costs v thick clients and a file server/authentication server it is not even close to being cost-competitive with a ‘NIX OS on thin clients.
- Robert Pogson
Published by Robert Pogson February 20th, 2010
in Linux in Education and technology.
This fall, HP announced a “new” product to deliver 10-seats-per-PC thin client systems for classrooms. They boast a price of $331 per seat (monitor not included, of course).
Three years ago, I built a thin client system for a school:
- 3 terminal servers – 4gB 4SATA 60 gB drives – $1200 each = $3600
- 1 file/web/authentication server 2gB 4 SATA 500 gB drives $1500
- 96 thin clients $134 each = $12864
- 13 multiseat X PCs for six clients $400 = $5200
- 153 keyboards and mice $10 (HP!) =$1530
- 153 LCD monitors $140 = $21420
Total cost = $46114 /$301.39 per seat
I dare say the performance was and still is better than you will get from HP and we do not need to fight malware either. We used four SATA drives in RAID 1 so four files could be read simultaneously when busy.
So, HP, get off the Wintel treadmill so you can give better value to your customers. With Moore’s Law, the system I built would be about $250 per seat these days. We would save a lot on servers and a little on thin clients and monitors. We saved a lot on licences not paid to M$, too. We used Ubuntu GNU/Linux. Today, I would use Debian GNU/Linux.
- Robert Pogson
17466
12762
206
3
2
23909 
11862 
11714 
4634 
4261 
1642 
198 
14 
2 
0 
0 
0 
Recent Comments