- cross-posted to:
- hackernews@derp.foo
- technology@lemmy.ml
- cross-posted to:
- hackernews@derp.foo
- technology@lemmy.ml
You must log in or register to comment.
What a super weird question. “Cloud computing” is distributed computing. Distributed computing is practically all we have left. Bitcoin/crypto, Kubernetes, Bit Torrent, and endless AWS/Cloud infra patterns. Then we have our happy little Fediverse here.
I feel the author was trying to say “is at home distributed computing dying?” In which case, yes, because Mobile took over and you really can’t do background compute on those. Certainly not like how SETI@Home worked.
Cloud computing is… Not distributed computing.
We’re talking about pushing compute workloads across a distributed set of devices where that workload is linearly scalable by the number of devices involved, compute, storage, failovers…etc scale elegantly. Cloud computing can give you the tools to make such a thing a reality within the scope of the cloud provider, but it most definitely is not distributed computing just by existing.
Also the fediverse is NOT distributed computing either, at least for Lemmy. There is no distributed compute available for Lemmy. You can’t have a few hundred users toss up their own compute to handle loads for an instance. Each instance is limited to a single database, and can have webservers behind a load balancer to spread out the compute. And that’s about the best you’ve got. Not distributed, you can’t just spin up 100 nodes for a Lemmy instance to handle more load and everything “just works”. It’s a very “classic” architecture in a way.
A K8 cluster isn’t distributed computing until you build a distributed application that can elegantly scale with more and more nodes. And is fault tolerant to nodes straight up dying.
Kafka for example, is an actual distributed application. One which you could run on a K8 cluster, it self-manages storage, duplication, load balancing, failovers, rebalancing…etc elegantly as you add more nodes. It doesn’t rely on a central DB, it IS the DB, every node. Lemmy is not.
Just to add, I really hate how mobile operating systems work. They really prevent them from being used more seriously.
I hope “Dex” like abilities become the norm. Clearly, phones can handle most tasks and it could prevent the need of a computer or at the least, be down cycled into a computer for a kid or grandparent.
I have so many old phones laying around that would do fine in some sort of desktop mode for spreadsheets and browsing.
I hate how much Apple has purposefully handicapped the iPad. It has such capable hardware, but the software doesn’t even come close to taking advantage of it.
But God forbid the iPad actually becomes a viable daily use computer (What’s a computer?)
Absolutely! If I could just run MacOS an an iPad when it has a keyboard and mouse attached, I would get an iPad today. There is literally no reason it cannot do this other than they don’t want it to.
I would be happy to make it my sole computing device. I wouldn’t even mind if it switched to iPadOS once undocked. But as it stands I literally cannot do my work on an iPad.
Aside from personal websites and maybe some Lemmy instances, I can’t think of a single application that’s NOT using distributed computing. Hell, Lemmy as a concept is still distributed computing even if individual instances aren’t necessarily.
Lemmy is… Not distributed computing.
If each instance is a separate application than must scale on it’s own, then no distributed computing is occuring.
There is one database, and you can have the instance itself behind a load balancer.
Lemmy is not a distributed program, you can’t scale it linearly by adding more nodes. It’s severely limited by it’s database access patterns, to a single DB, and is not capable of being distributed in it’s current state. You can put more web servers behind a load balancer, but that’s not really “distributed computing” that’s just “distributing a workload”, which has a lot of limitations that defeat it being truly distributed.
Actual distributed applications are incredibly difficult to create at scale, with many faux-distribited applications being made (Lemmy being n-tier im a per instance basis).
Think of Kafka. Kafka is an actual distributed application.
The article seems to use the term distributed computing for volunteer computing. Distributed computing very clearly is neither dying nor fading away.
Volunteer computing is more interesting. There’s a constant trickle of new projects, like distributed sharing of GPU power for running smaller LLM models, random science experiments that have you install an app on a phone to collect data, various Blockchain shenanigans, etc. I’m not sure if they’re getting less coverage than, say, Folding@Home or the Great Internet Mersenne Prime Seach. So perhaps it is fading into the background, but it certainly doesn’t seem to be dying.
Yeah, even comments here haven’t seemed to read the article. As somebody that used to install BOINC on all my machines back in the day, the reason I stopped is that many of the projects I ran (SETI being one) aren’t active any longer. Also, like the article mentioned, I just don’t have a desktop anymore and I am not about to run something like this on a laptop that doesn’t have things like user-serviceable or replaceable parts.
The rise of distributed computing was at a time that CPUs didn’t really throttle down. CPUs in general were just a lot more power hungry. But if you had to leave the computer on for some reason and had spare CPU cycles, it made sense to contribute them to a distributed computing project - the power was being spent anyway and it seemed like a good cause. Today, modern CPU sip power and throttle down and you are actively driving power consumption by taxing the CPU. There is a much less favorable cost/benefit equation today, but in terms of the cost of the power consumed AND the climate cost of the power consumed.
It’s use cases might be more focused, but I’m gonna say no. At least based on the distributed database cluster that I am in the middle of building right now. When folding@home started, the processors in PS3’s were a big deal. But these days a single socket server can have 256 threads and 3TB of RAM in it. Why manage 500 servers when you could manage 4?
I am not 100% certain that this is the correct answer but I think distributed computing still happens but it between containers and VMs instead of physical machines. This is just a guess though and could be wrong.
I mean, every supercomputer is distributed too, the concept is growing if anything. This was specifically about over-the-internet distribution with volunteers.
As soon as someone cracks efficient distributed LLM training (and ideally a distributed, privacy respecting, inference for larger models) that’ll change, and 'OpenAI (such an oxymoronic name!) will openly cry…
At first I thought you were talking about Chorus distributed operating systems, something I used in the 90s, and I thought “yeah, it is dead” :)
But distributed computing still exists in companies, just not used in everyday users like good old seti and folding etc
Gridcoin (which uses BOINC) is an example of a thriving distributed computing project.
There is cool project called sheepit where you can render blender projects, get credits and render your own projects in one fast burst.
If you mean crypto mining malware, probably since crypto is pushing up daisies
