Distributed processing for solving problems, not profit

Words
871
Reading
4 min
5h
Listen to this post
~6 min

I have been using computers for many years and the processing power available to regular folk has gone up by several orders of magnitude. It seems we can find various ways to use that. The latest is for 'AI'. Vast data centres are being built, but you can also run it on your own PC. That is viable with modern GPUs that were originally intended for gaming, but they essentially contain hundred or thousands of small processors that can be harnessed for other tasks. Another task they can do is mining cryptocurrency and many people will have done that. There is also dedicated hardware that can do it even more efficiently.

Long before the current AI and crypto were around people looked at ways to harness the power of the millions of computers that were online, but mostly sitting idle. People could run an application that would take packets of data and process them when not being used. An early one was SETI@home that took radio telescope data from the Search For Extraterrestrial Intelligence project to look for signals that might be the sign of someone out there. This could be run as a screensaver and had the attraction of some pretty graphics.

SEIT@home screenshot from Wikipedia

I think this may have been the first I ran. Another attraction was to see how far up the worldwide rankings you could get and people would set up lots of PCs to do this. It looks like SETI@home ran from 1999 until about 2020, but I had stopped long before that.

Another I did was about proteome folding looking for ways to cure disease.

The one I ran most was more of a geek thing. distributed.net started in 1997 to solve the RSA encryption challenges. There were some prizes for these, but your chances of winning a share were tiny as thousands of people took part. Some of those had many computers and the software was available just about anything. You could even run it on some gaming consoles.

The same software could also run the Optimum Golomb Ruler (OGR) search that has applications in mathematics and some technologies. That part ended a few years ago.

I ran these projects on my own computers and also on some at work. I had some unofficial permission from some IT guys to do that and eventually had about twenty people taking part there. This was all on non-GPU machines with single core processors. Processing a block of the current RC5-72 challenge might take hours. I had not thought about it for years, but thought I would try it on my current PC that has a RTX 5060 NVIDIA GPU. That gets through a block in under 22 seconds!

Block processing

I ran it for a while yesterday and you can see my statistics. This PC has already run more blocks that my old account did over several years. It looks like hundreds of people are still participating, but the project has only processed 17% of the key-space in nearly 24 years. It may be that it is accelerating due to hardware updates, but is likely to take a few years yet. I estimate that it is currently processing about four times faster now than it has managed on average over those 24 years so it could take less than that if they can scale up the processing power and people are willing to keep doing it. As it is about finding a given key it could actually finish at any time.

Warm GPU

I am just wondering if I want to be running this all the time. It makes my GPU run fairly hot and I notice more fan noise. I can see from the data on my solar inverter that it may use 200-300W extra. The extra electricity cost is not such a big deal given that I am mostly running on solar power. I tend to export any excess and so this would reduce what I could earn from that. The difference is a lot less than what it would cost if I was just using peak time power from the grid.

As I said, this is a geeky thing to use your PC for, but I find it interesting. It is really just proving that encryption is really hard to crack via brute force methods. The actual encryption people should be using should have much more complex keys. Some encryption may be vulnerable to quantum computing eventually and AI is being used to find weaknesses, but we have to hope that things like Hive will not be cracked any time soon. Crooks have bypassed this by threatening people holding lots of crypto with actual violence.

Just this week we heard that some web security certificates were signed with ridiculously short keys. These were just a few digits long when they should be hundreds. One key was literally the number 4! That is just sloppy operating.

Have you participated in projects like these? You probably have what would have been considered a super-computer years ago, so what do you use all that power for now?

The man behind: tenkminnows@tenkminnows: Helping good Hivers level up

proofofbrian@proofofbrian: A bot that checks for tag typos

#BritList: A monthly list of Hivers in the UK

Distributed processing for solving problems, not profit | Ecency