Thin client cluster servers

Since we first heard about the tuxlab project (first by the Shuttleworth foundation and now Inkululeko technologies) we have been keen supporters of the model in the CoE school's project, now named e-Yethu. For two reasons, firstly open source software in schools, which is (we feel) the way things should go. That said we never stop schools from using Microsoft products, but then the software is their own problem to sort out, as we obviously can't provide it. The other is the thin client lab model, which is one less costly as the only real expense is the central server and also less maintenance as the only machine to maintain is the server. Maintenance is a massive issue, especially when you start taking on more schools. Right now we have 7 schools who look to us for support and then just me providing that support. And when that is not your only full time work things can get a bit hectic time wise.  Anyway, so if a school wants our help specifically we try to encourage as much as possible for a thin client model lab running edubuntu.

As I said, in this model the server is the only really expensive item, but sometimes that expense is even too much. Schools can't source the money and sometimes we even struggle to source the money. Desktop computers however pose less difficulty sourcing, they are cheaper so easier for us to buy and we are to get more donations from other parties of desktops. We have also found, rather obviously, that the single thin client server poses a risk in terms of a single point of failure - so it might be more scalable to maintain as the number of schools increase but when it fails its more chatastrophic.  For these two reasons we thought wouldn't it be awesome if we could cluster the LTSP server over a number of machines (namely desktops that we tend to get given on a regular basis).

There are four cluster options I would like to implement: LTSP DNS round robin cluster; Ubuntu's new cluster server implementation; DNS round robin cluster; and single system image cluster. Options 1 and 3 seem to be the same thing, but are not. In 1, the portion that is clustered is just the LTSP service (so there is still a single point of failure with dhcp and nbd services - edubuntu implementation - from one of the cluster servers, which also acts as a primary file server). In option 3, however, I want to try and decouple the dhcp process a little more from the nbd services so that regardless of the dhcp server another server can provide the tftp services for the client that is pxe booting, meaning that any of the cluster servers could crash and the others would continue working as if there were no change.

I hope to have a number of entries here detailing how each option was eventually implemented, deployed and tested.


Display comments as (Linear | Threaded)

    No comments

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
Standard emoticons like :-) and ;-) are converted to images.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.