What uses the most energy in your life? How can we, as a society, make it easier for you to eliminate or reduce that thing?
Individual and community action are entwined.
@tindall considering that the latency is bad enough that it's basically unplayable just connecting through wifi to a computer wired directly to the router, I have no idea how they can do it
@ben i mean if the mainframe in question is in the basement of your CAT-6 wired apartment building I think it would be fine
@tindall I mean, I'm probably not 50 meters away from my gaming PC right now and the latency is such that I can't do anything that requires hand-eye coordination in a video game
@ben @tindall This is able to be made to work pretty well with the new fangled deep learning stuff. A model can be trained on a game and get really good at guessing. The model is then not anywhere near as big to download or intensive to run as the real game, but is good enough to predict what the next few frames will probably be based on controller input.
@ben @tindall I can highly recommend cloud gaming (yet another form of mainframes/timesharing computers) for various games. Of course you don't get your Counter-Strike Source or alike done perfect there, but I use it to play various strategy and building games. Given that I spend just 150-200€ a year while getting high-end graphics I would consider this a huge win compared to buying a graphics card for 2k€ at this point.
@tindall I spent a lot of time in this space a long time ago, and the thing that I kept coming back to was that the most expensive (and energy inefficient) parts of a cheap computer were I/O, which are the bits you need to make a community mainframe work.
@ajroach42 Fair point.
I/O devices have come down a lot, and networking is fairly ubiquitous now.
The problem I see for mainframe is that what you need is (reasonably) continuous connectivity (batch-process excepted), and that a lot of what's now being transacted on systems involves a lot of data transfer (audio and especially video). For interactive activities (gaming, voice / video comms), broken / slow / high-latency / variable-quality connectivity doesn't cut it.
Your own work with a slow / intermittent network connection is very much a case in point. There's a lot to be said for hub-and-spoke store-and-foreward designs in such a case (including sneakernet / saloon car full of tapes data transfer). Some centralisation, but also reasonably autonomous hubs.
The devices that provided mainframe-class service 3--6 decades ago would barely operate a toaster today (in terms of processing power, not thermal output).
@tindall What specific problem(s) are you trying to solve?
Infotech/comms energy tends to be:
Embedded (energy of manufacture)
Infrastructure (also mostly embedded, though maint & opps as well.)
Actual device use. That's ... often low.
Displays. For energy, you can save much with e-ink, though with numerous compromises.
Datacentres / centralised ops. This is where a lot of energy use occurs, though it is shared across many users.
The monetisation infrastructure. Effectively, ad-tech, which generates bloat all kinds of ways.
I'd probably look at that last first. If transport and land-use are the interrelationship with energy, then I'd say it's the funding mechanism for infotech media that's the big factor affecting energy there.
It's not just direct, but the fact that, say, ongoing web bloat means old devices which function perfectly according to spec but cannot handle present uses need to be replaced. That's both embedded cost and beefier devices than are necessary (though energy efficiency per unit compute power still increases majorly).
But you cannot answer this without addressing what uses / needs you're trying to fulfil as well.
@tindall I've also pointed out that given the location-indifference of Internet technologies, and the loss then of location as a market segemntation device (streetfront, foot traffic, demographics selection), then discrimination based on differential capabilities and recency of end-user kit becomes a way of distinguishing "desirable" and "nondesirable" consumer segments.
It's an awful strategy, but one that I suspect is actively and intentionally pursued.
@tindall The biggest boonies problem right now is bandwidth and comms, which a mainframe-like solution would specifically not address.
What's proved most socially-impactful, based on everything I've seen, somewhat resembles the old photographer's advice: "F5 and be there". That is, pre-set your apature to an intermediate range, and have the damned camera on you when whatever it is that you want to photogaph happens.
I hate to mention the guy by name, but the Zero-to-One notion is spot on: the increment of going from none of something to one, that is, no capability to rudimentary capability, is immense. No other marginal increment matters nearly as much.
(Contradicting this in my head as I write but I'll leave that for later.)
Mobile phones, SMS texts, the ability to get basic information (mostly market and weather) has had an incredible impact on underserviced areas, often lacking in POTS capabilities (much of Asia & Africa). Mainframes would be peachy, but major overkill.
Many typical Western online services are vastly too data-fat to be viable in such locations. Very thin clients / data-feeds are essential. "Recent-time" is far better than "real-time".
@tindall Lucky thing is, a lot of these either exist or have existed.
E.g. for carless, there are well known cities that have a lot of cycling, but there's also islands like the Prinkipios islands of Istanbul where cars aren't allowed. Some of these islands are fairly large. There's no disruption to life, people ride bikes, there's air to breathe and nice sounds to hear. One doesn't need to fear for their lives like a rabbit when walking. Some even have horses. (There were horse carriage taxis too, which were abolished because the riders were abusive towards the horses [which is great because when they werent assholes to animals, they were assholes to pedestrians]).
For mainframe computing, we overdo computing by an order of magnitude to begin with anyways. Besides games, nothing we do needs to be as heavy as it is. More interestingly, heavy stuff like games can indeed be "mainframed". There's this nVIDIA thing where you can subscribe and play the games on their computers, with yours as essentially a thin client. Apart from that, if we used digital documents and media properly, out file sizes and processing requirements could fall by orders of magnitude. Most MS Office documents and most PDFs are superfluous. We could use much smaller programs that are equally usable. And if we rid of ad tech, the web would be orders of magnitude smaller.
As for density of living, in the past, the types of available housing was way more diverse. Big families living in family homes in neighbourhoods. Widely available, cheap single person dwellings (studio apartments, bedsits, microapartments, dorms) with shared spaces (sports, hobbies, recreational, even kitchens and bathrooms sometimes) in cities, which can spread out w/o cars if there's nice transit and cycling. The reason these are not widespread anymore and still receding is housing becoming a financial investment, and cars being very lucrative to car manufacturers, oil companies, et al. If landlords and capitalists were kicked out of the equation, we'd have diverse, green, and accessible transport and housing.
And that's where the problem is. Nice things don't drown capitalist lords in juicy, abundant ROI. Before the pandemic, I lived in a 50m² apartment in Ankara, and it was more than enough. Bus, metro and the occasional cab was good enough for all my travels, 7/24, no exceptions. Then, in bigger cities like e.g. Tokyo or Hong Kong, they have very smart solutions to make the best use of the space they have available https://www.youtube.com/watch?v=TYVJbupG3Xg https://www.youtube.com/watch?v=vMk289_WAUU Elsewhere, all cities have dorms that function similarly. There's no reason municipalities don't operate public dorms for people living alone.
In sum, a lot of worlds problems are solved. The solutions are nice, pleasant, and already out there. Problem is they are not lucrative to the right people.
@cadadr @tindall On the note of computing, we're definitely there with things like LTSP and even raw XDMCP (full X11 sessions over TCP). And, of course, as always, Sun was here first (or at least very early) with its Sun Ray hardware back in the late 90s / early 2000s. These could do more than just X11 (Oracle built Windows support), but at the time X11 was the assumption.
@tindall Energy budgets are pretty well-established. The US DOE produces its sankey diagrams (energy flow charts) through the LLNL, there are similar data for the IEA and elsewhere. Pick up virtually any book on energy or climate (IPCC) and you'll find a breakdown. These aren't mysteries.
Here's the 2020 LLNL / US chart.
Interesting, straight off the top: the usage is once again well below 100 quad (quadrillion BTUs, thanks, standard system, a/k/a 105 billion billion joules, ~30k TWh, 17 billion barrel oil equivalent (bboe), 2.5 billion tonoil equivalent (btoe), 3.6 toncoal equivalent (btce).
If you look through the archive site, you'll find a set of projections from about 1975, projecting use forward to 1980 and 1990, as well as back decadally to 1950. Looking at those, thirty years ago the US should have been using far more energy than it is now, if 1950--1975 trends had continued.
@tindall All of that out of the way ...
The biggest use is ... industry, just edging out transportation (this was a surpriise), though veyr roughly you can divide energy use into thirds: transport, residential & commercial, and industry. There's also electrical generation, but that's simply an energy mode that's then allocated to end use.
There's a lot of focus on the "rejected energy" element, though IMO that's largely misguided. Most of that is based on theoretical and practial efficiencies, e.g., thermal motive energy (including tranport and electrical generation) runs at ~30--40% efficiency simply due to Carnot process. Unless the energy cycle itself is fundamentally changed, that's a given. There are efficiency gains possible, but it's fairly modest.
The major modifiable elements are transport and discretionary energy use. Transport (about 25% of net energy) is about 70% private passenger, 30% bulk cargo. Modes matter.
The biggest determinant of transportation energy use isn't vehicles or modes, it's land use.
If you spread things out, you need to move things around. A lot. Density also enables and disables numerous transportation modes. Most especially, high-capacity transit (heavy rail, high-speed rail, light rail, busses, trollycar / trollybus) all rely on high concentrations of residences, workplaces, commerce, and industry.
@tindall And some modes are notably efficient. Ships and trains, by units of mass*distance/fuel are phenomenally efficient, up to 500 ton-miles/gallon for rail, perhaps double that for marine or canal shipping.
Trucks are much less efficient but far more flexible. And passenger cars are comparatively abysmal.
With bulk cargo, scale again matters. Endpoint distribution is hard (the last-mile problem). Rural delivery is similarly difficult (package-delivery services love the high-margin urban areas, but defer to national postal services for rural delivery, typical of any networked-distribution system: mail, electricity, POTS telephony, cellular telephony, broadband internet, retail chains, ...).
The biggest impact of the automobile has been on land-use patterns.
@tindall Other energy uses also derive from land-use.
With denser urban regions:
Per-unit heating and cooling costs are decreased (shared walls -> reduced heating/cooling loads.
Communal foodservice. Whether informal or commercial foodservice, meal prep and production occurs at scale. (The low-ooccupancy unit, highly-appointed kitchen is a recent novelty. Earlier times had larger multi-generational houses, rooming-homes, public houses (pubs) for food, village ovens, etc.) The space allocated in residences to food storage, prep, cooking, and cleanup is substantial.
Infrastructural efficiencies. #StrongTowns discusses this aspect at length.
@tindall Density isn't everything, and there are many other elements to be considered. But as a single point of intervention, it's probably the least-cost / highest-benefit route.
Vastly more effective than pushing the individual-responsibility canard.
@tindall And one more on land use / density and energy consupmtion. This via Shane Phillips: denser housing has major savings in both transport and non-transport (heating/cooling) energy loads:
@tindall have you read Ecotopia by Ernest Callenbach ? It was written in 1975, but feels ever so actual. It is not without some defaults and weird "a white guy definitely wrote this" moments, but gives a really nice overview of somewhat a more balanced word. I really recommend it.