Sunday, January 16 2022

Consumption, Technology, Economics, Sustainability - A Small (therefore beautiful) Study

A recent article about the way Linux gave older computing devices extra life started a train of thought.  The gist of the article was that older laptops and desktops, in particular, ran very well  and securely with the latest Linux distributions as their operating system and applications, when either the operating system on which they originally ran was unsupported due to age or various issues meant that older systems struggled with more up to date software.

I know this to be true. Heavens, the very first Linux server I ran in anger was a re-purposed older server that was long past its original use-by date, but which performed better than the Windows server it replaced, and that was in early 1998. So, the thinking goes, if you have an old system, download the Linux distribution of your choice and extend the life of a perfectly working set of hardware.  One therefore reads or hears of stories of people using this technique especially to help older people who can't see why some item they bought just a few years previously was no longer doing its job, or to supply younger people, who may not be able to afford a new device, which an essential of modern life.

Who can argue with this? It lengthens the life of the one item that does such damage to the world and its people, from the unpleasant and demanding conditions of factories to the need for ever-more raw material with which to build the devices. It's also an act in strong contrast with the economic model of consumption as a driver of "growth", which is taken to be some natural thing, rather than an economic invention.  The most unworldly form of this, frankly, crazy model, can be seen in the comments section of a recent tech press article on messaging systems - see here - in which peer pressure allegedly forces younger people to throw away working devices and buy an expensive replacement (with the associated annual "upgrade" cycle) because their messages have a blue, rather than a green background when their peers receive their messages. Such triteness is, no doubt, part of the problem of such consumption, and the economic model that underpins it.

Coming back to the idea of consumption choices for a longer life, I applied the same general thinking of a long-lived manufactured article when I bought my faithful old Land Rover in 2000. The idea was that such a vehicle would last much longer than a more conventional car. And indeed, this was the case. We had 17 trouble-free years and 170000 trouble-free miles from that Landie, and when we sold it, it still had its original factory-fitted battery. We learnt other things that helped sustainability,. Its first set of tyres lasted 88000 miles. Later, although they lasted half the mileage, we used retreads, which incidentally removed the Landie's tendency to under-steer.  Arguably, we might have bought 3 new vehicles in the time we had that Landie. That's 3 times the manufacturing energy, 3 times the economic hangers-on (marketing, sales etc).  But we averaged 30 miles to the gallon with that Landie.  I consoled myself with the fact that the design of the engine meant its emissions were much lower than other diesels of its era, and indeed, needed no catalytic converters to conform to the standards of the day.  But 17 years later, those standards had changed.  17 years later, the ideas underpinning the original decision no longer held as true. 17 years later, a replacement was probably the better option.

Something similar is true with regard to computer hardware. If you've perused this blog, you'll know we are off-grid, and therefore we are necessarily aware of power consumption of the various items in the household.  An example might be light bulbs.  We started off using incandescent lights, then, as soon as they became easily available, switched to fluorescent lights, and in spite of the cost, or the need to throw out the old incandescents,  we were happy to see our power consumption for lighting drop by about two thirds, a act we used to justify buying more fluorescents for more light, as the power consumption had dropped.  Fast forward a few years and now we have easily available LED lighting. So we threw out the old fluorescent lights although they still functioned as lights, and bought LEDs. What's more, we bought more LED lightbulbs, as they were so power efficient, we could even use them for decorative purposes rather than strictly for their function. And indeed, if we now switched on every single light in the house and the sheds, we'd still use less power than one of the original incandescent bulbs.  Yet we had to buy, consume and discard to get to this point.

Of course, we had fallen into an aspect of the effect described by William Stanley Jevons in 1865, that the more power efficiency improves, paradoxically, the more power is consumed. It's not quite so bad for us, because we use roughly the same amount of power as 14 years ago, but theoretically we should be using far less.

The parallel with computing is also intriguing. We run our own computer services, so we have a server here at home.  14 years ago, the "low power" option was the Intel Atom processor.  This was, of course, low power in more ways than one.  But running Linux on this device meant that it provided services acceptably, and, according to the concept of using Linux to elongate the life of devices, we might have kept using it for many years. But actually, the Atom was by no means a low power processor.  Soon, AMD's offerings were providing much more computing power, while consuming less electrical power.  Our laptops 10-14 years ago were consuming 15-20w of power and battery life was perhaps 2-3 hours.  The laptop on which I am typing this, although an order of magnitude more powerful inn computing capability, uses 4w of power at rest, and battery life is 10-12 hours. We could have kept using the old devices, but the argument of lower electrical power consumption is a strong one.

But in order to taker advantage of these new aspects of technology, you have to participate in the economic model of consumption.

We are now quite close to a point where this may no longer fully be true, but with caveats.  These days, our server at home is a small Raspberry Pi. A tiny device costing less than £50 uses 2-3w of power. Add a disk, and even when it is working hard, it's rare for that power consumption to hit 6w. But even with this device, improvements in manufacturing techniques and other economic factors have meant that the little Raspberry Pi  is many times more powerful, and has 16 times the memory, than its grandparent device, which came to market in 2013.  And instead of that one server, I can now afford a spare, and a spare for the spare, and a development machine... So the device that enables lower consumption options allows me - encourages me - to consume more.

Nevertheless, thinking about providing modern services that we expect using these types of thoughts of balancing consumption versus capability, longevity versus possibility,  to underpin decision points can leave one with better decisions from a consumption perspective than using other, especially, purely technological, decisions. An example of this is the Assynt Digital Archive, with which I am involved.  The Archive began in 2011. We made good technological decisions at that time, for longevity and resilience to data loss. The four workstations and server, along with accoutrements necessary to run the little network, cost about £10000. Considering that the network ran for 8 years meeting its design target and without requiring "upgrades", one could say that the ideal of balancing capability, consumption and economy was met. But last year, the Archive moved from its old location and came under a different managing group.  One aspect of this change was less physical space. The old machines simply would not have fitted into the available space. Anyway, there was something organisationally advantageous to renewing the systems.  This time, we just wanted three workstations. But even with that difference, the total cost came to just £1500. We now base the entire network, server and workstations on the little Raspberry Pi boards. A happy by-product is that the server, which remains on all the time, uses a tenth or less of the power of the old unit.  These systems should last for several years, but even if we have to replace them at 3-4 year intervals, for whatever reason, one could argue that it is a better balance than purchasing much more expensive, market-driven conventional servers and workstations.

The key to all these thoughts is that it is possible to base one's thinking on internal imperatives rather than external market-driven frameworks, and one can achieve a better, though not ideal, balance between technology, consumption, economics and sustainability. But these principles are by no means cast in stone. They need testing, and they need action when they no longer hold true. Otherwise one is mistaking stubbornness for sustainability. But one must then evaluate changes, especially when they are thrust on one from outside, to see whether they meet the standards set.


Monday, November 29 2021

Paradoxes, Power and Performance

Some thoughts on changes in our off-grid supply and demand over the years.

Continue reading

Saturday, October 9 2021

Squeezing a Raspberry Pi service down a slow Internet link.

Our home Nextcloud installation works well, but when away from home, the heavily constrained bandwidth of abysmal rural broadband makes it close to unusable. Fortunately, there are ways of resolving the problem.

Continue reading

Wednesday, May 26 2021

MONIToring network services

Simple small-scale service monitoring - Can Monit out-Nagios Nagios?

Continue reading

Saturday, April 10 2021

Lighttpd and Nextcloud 21+

An irritating but apparently harmless notice when updating to Nextcloud 21 needs a slight change to the web server configuration under Lighttpd

Continue reading

Monday, February 22 2021

OpenSUSE as a Raspberry Pi 4 Server

A work-in-progress series of notes on a migration to OpenSUSE as a server on a Raspberry Pi 4.  These notes are somewhat rambling, as they were written during the development process of the new server, but are left here, warts and all.

Migration project status: The project has been completed successfully. All services are now running under OpenSUSE as of 6th March 2021. I moaned and grumbled my way through this post, adding to it as issues arose or little triumphs were achieved. The reality is that what felt like little hacks, workarounds and customising I had to do, tiny things, really, were probably no more than were helping the stability of the previous system, but time and familiarity had made me forget that's how sysadmins make things work.
This has been an interesting exercise, and I plan to leave these notes here, both for my own reference and possibly as some might be useful to others.

One month later:  I remain pleased with the decision to migrate from Raspberry Pi OS to OpenSUSE for this server.  In fact, I wish I could easily migrate a hosted Raspberry Pi I have.  The most pressing issues I have faced were with regard to various weekly security and performance reports I have written over the years. I am still getting to grips with systemd's timers and logrotate, as some of the reports should run after the logs are rotated, but I am still not clear when these happen. One unexpected bonus is the degree to which btrfs compression works.  At the moment, there is a 25% space saving on the disk.  The next big test will be with the release of OpenSUSE 15.3. Previous experience on x86 suggests this update will be straight forward,  so I am expecting more of the solidity I am currently enjoying.

Continue reading

Omeka under Lighttpd

Documented for those wanting to try this out

Continue reading

Saturday, January 16 2021

Our own xmpp server, with all the trimmings

We can bypass all the data slurping and other issues, by running our own messaging service, including voice and video, and group chats.

Continue reading

- page 1 of 9