Come along with me on a small thought experiment:
Imagine for a minute, you’ve been given a task by the United Nations Commission on Computers & Stuff.
- Your job is to optimize all of the un-optimized digital images in the world (to take less storage space).
- Your mandate dictates that you can only do lossless optimization. For context: there are tools that reorganize the contents of image files such that they take take up less storage space, but no data or image quality is lost. This is in contrast to compressing an image more, where you discard data and lose image quality.
- In my thought experiment, you somehow have the funding, computing power, and the access to all computer systems in the world.
My question is this: Is it better (from a climate standpoint) to:
- Spend the energy once to save all of the storage space in the longer term, or;
- Don’t spend the energy to optimize, and keep spending the energy required for the un-optimized storage?
Obviously the real answer is It Depends on factors like: how much optimization is possible on most images, how long (and on what type of storage) are most images stored. Let’s not ruin a fun (?) thought experiment with nuance, though.
What say you? Does it cost more to squish a big image, or keep it around un-squished?
Since decompression requires time/energy too, assign a last-accessed date to each image, and periodically evaluate and squash infrequently accessed ones; this will result in a majority of the images being squashed, saving a lot of space, but the most frequently requested ones will stay uncompressed until usage drops, reducing energy use from not having to decompress on each request.
Good point Shawn. Compression algorithms definitely varying in computational demands on both the encoding and decoding side.
But what’s your nuance-free gut-instinct from-the-hip answer?
> nuance-free gut-instinct from-the-hip answer
Do nothing.
Yes, how often each image is accessed is important, not just because of the energy needed to store the image, but also where and how the image might be accessed and how quickly it must be delivered. A digital photo frame uses less energy to display an image from its own storage than a web server responding to an urgent request from the Mars colony.
A god (or demon) has given you a single button to press to optimize all images every (and incur the energy cost) – you have to decide to press it or not!
I do not press the button, because burning energy harms people now, whereas humanity may rise to mitigate the harm in the future. Elizabeth Anscombe is my guide.
At long enough time frames, don’t you need to minimize the lifetime costs of storage media? Excess — let’s call ’em “hard drives” — are wastes that need to be replicated eventually (more efficiently than we can do today, likely, but manufacture is surely always going to be very costly relative to computation).
Unless there is reason to believe the cost of the compression calculations per unit will decrease, in which case there’s some future crossover between electricity bills and the cost of building “RAW Storage Facility 865,542,278.”