Alex Little

Up and running

Yesterday I came back from the Small Is… festival organised by Practical Action and Engineers without Borders (EWB).

I’d gone over partly just out of interest – when I was working at Aptivate in Cambridge I was sharing a house with some of the interns at EWB helping to organise the festival. Partly to see what my brother was up to when he runs the solar power workshops etc that he’s run at many festivals over the past few years. But also to help out Alan from Aptivate in setting up a solar powered thin-client network.

Just before I finished my short placement at Aptivate, Alan and I had spec’d out what we’d need to be able to run a small thin-client lab on solar. Alan had bought the equipment on Friday and so, Saturday, at the festival, was the time to actually try and get it all running. All worked out really well and within a couple of hours we had the system up, running and 3 terminals (including the server) connected to the internet, all running for a single solar panel.

Moving away from the rain

What we’d do differently next time:

• plug the PV panel into the battery before 3pm – so it had some time to get charged up and so demonstrate for longer
• be more consistent with the power supplies – we had 4 different power outputs that we needed to supply (5v, 12v, 16.5v and 19v). Using the same type of terminals would have helped with this – we had 4 different types of terminal.
• add up the current more carefully – we blew several 10amp fuses because we’d put too much load on
• lay our the server/terminals etc onto a board and have everything neatly arranged to screwed down, to make it easier to explain to people what was going on – rather then the spaghetti of wiring we ended up with

Equipment used:

For the PV system:

• Lorentz 95Wp PV panel (model: LA95-12S)
• Rolls 12v deep cycle battery (model: S12-128AGM)
• Morningstar SunSaver-20 charge controller (model: SS-20L-12V)
• + assorted wires, fuses etc

For the network:

• IBM Thinkpad X41 (running Ubuntu 10.04 with LTSP)
• Samsung N netbook fitted with Pixel Qi daylight readable screen (configured to PXE boot)
• Aleutia T1 fanless PC (configured to PXE boot)
• Aleutia PC (I’m not sure of the model – configured to PXE boot)
• 2x Iiyama Prolite E2271 HDS 22″ monitor – one of the only low power monitors we could find that will run from a 12v external power supply
• Netgear DS104 4 port hub
• Vodafone GSM mobile broadband dongle

From the outside

In the past, I’ve worked on quite a number of research projects developing prototype software and web applications, where, due to the nature of the project (pilot/prototype/research) you never quite get the time to invest in making these systems as robust, well structured or documented as you’d like. Yet at the same time, a tension starts when there is a push to get these systems used in the real world, and they become to be treated as production level systems. No, it’s not quite that black and white, research projects need some real world testing to prove their worth and production systems will never be perfect.

With our new thin client labs and OpenSolaris server I feel we have a similar tension. On one side, these labs were a facility for us to try out new ICT policy and infrastructure for the university, for example, disabling flash devices, using university mail accounts, amongst many others. Yet at the same time, our lab is the only functioning lab available in the Engineering College. A review yesterday by the ECBP team, found that there was only one other lab (approx 30 PCs) which was functional and had relatively modern PCs (less than 5 years old) – but this lab isn’t yet networked (it’s in the computer science building – which, almost a year after being occupied by the department, still hasn’t been networked except recently for some of the staff offices). So there are 40 networked terminals available for a student population of over 2000. That’s a ratio of 1 terminal to 50 students in a college teaching engineering, technology and computing.

So we have a question of priorities, do we stick with the research goals and risk the labs not being usable due to the decision we’ve made (but then have something interesting to write up), or do we focus on providing students with a lab they are able to use effortlessly?

For me, it’s an easy decision to make. We must focus on providing a robust and reliable lab (contrary to the other labs) for the students to benefit from and not worry about the fact that they’re sharing passwords, using webmail rather than university mail accounts and the like. This doesn’t contradict some of the bigger research aspects we wanted to look into (e.g. demonstrating we can create a more robust and scalable architecture than the usual PC labs), but does mean that we very quickly needed to make compromises (such as allowing the use of flash drives) we didn’t want to have to make so soon.