A Different Take On PC Manus' 'Recycling' Schemes
Timex writes "Linux Devices is reporting that the COO of Linux startup Symbio Technologies, Roger Del Russo, claims that the recycling programs of Dell and HP are nothing more than a means of getting older systems out of use, to be replaced by newer systems. Mr Del Russo then makes the case for putting the older systems to good use as thin clients, using the Linux Terminal Server Project."
Recently, I set up a small Debian-based computer lab using LTSP at the school I attend. Currently, it only has the server and 5 thin clients, but it is astounding what can be done with this hardware:
Server: 800 MHz Pentium III, 1 GB SDRAM, dual 20 GB hard drives.
Thin Clients (all diskless and netbooting thanks to LTSP): 233 MHz Cyrix III with 64 MB RAM, 233 MHz Pentium II with 48 MB RAM, 333 MHz Celeron with 32 MB RAM, 300 MHz Pentium II with 64 MB RAM, 150 MHz Pentium MMX with 32 MB RAM (IIRC).
This setup can have 6 people (one on the server plus the thin clients) running Gaim + Mozilla + OpenOffice with very little slowdown. Rarely is it noticeable. And it is very easy to manage; since there is only one computer with Debian and all the data and whatnot on it, only one box has to be updated, had new users added, etc.
Total cost so far: $0, it is all just hardware out of my friend's basements and mine. The fact that no money has to be spent to get something like this to work (and work well) really shows that a new, shiny 2.4 GHz Pentium 4 is not needed just to check email and browse the web. The science department at this same school actually just recieved 24 new 3.2 GHz Dells, and the most that is ever done on them are Excel spreadsheets!
Imagine the money that can be saved by using old but completely useful computers instead of upgrading every couple of years when it is unneeded... (and spending more money on licenses for new versions of Office and Windows and etc.)
- The client broadcasts for IP and other info using PXE or Etherboot.
- The client downloads the net bootable Linux kernel.
- The kernel boots and mounts a read-only root FS over NFS.
- The client self-configures and starts XFree86 v. 4.3
- XFree86 connects to a second computer and the entire session is exported to the thin client.
There is no apparent difference to the user. They don't even understand that they are working on the server at this point. With a Gig LAN card on the server and 100Mb on the clients, the session lag is unnoticable.BTW, this is a true thin client, unlike Winterms or the Linux equivalent. Exploit in the kernel? Just drop a new one on the server and update the name in DHCP (or make a symlink), and, presto, all the hundreds of clients have been upgraded with no downtime the next time the are cycled. There is no need to burn flash on hundreds of clients. Any application holes can be filled in the time it takes to patch the server, with no reboots required anywhere, and no tech to be onsite at all.
In short, though you seem to think so, nothing in the user session is running locally and the user is not limited by any old hardware. The clients can be used for another fifteen years with no penalty is they don't fail. Upgrade the software, upgrade the server hardware (or cluster), and you've got faster, better user sessions for free.
Check out for more info.
Put identity in the browser.