Had a random thought, at my university there were many computer labs scattered across campus, most computers most of the time were idle and not in use.
Is there a piece of software you use that takes advantage of those idle resources to use as your cluster?
If not, would you like one? I'd love to build an open source project around it, sounds like a super fun project to contribute to research.
Well, there is a possibility I may be getting access to 5+ desktops. I am looking to build a cluster so I can run various programs on it, typically written in Python. These programs typically involve various needs to connect to internet, make API calls, and communicate through IP addresses, if that is important to note. I just want this cluster to execute these programs. So my few questions:
- What Linux distribution is optimal for this case? I was looking into Ubuntu Server, but I wasn’t sure about its speed and reliability.
- How could I construct this cluster? I know there are numerous tutorials, but I wanted to see what experienced individuals thought would be the best.
- How can I setup the host node (on construction completion) to execute these programs? I know this probably has a simple and obvious answer, but I am just trying to cover all bases for this.
Any more advice would be greatly appreciated! Thank you for your time.
Source: https://www.hpcwire.com/off-the-wire/e4-announces-the-risc-v-based-monte-cimone-cluster/
It contains 12 mini-ITX motherboards and each board has a SiFive Freedom U740 SoC. I think this means they simply use SiFive Unmatched for the base system. It is fine, as their purpose is not winning TOP500 or such.
”Monte Cimone is the first RISC-V ISA cluster specifically designed, built, and validated for co-design activities targeted to enable its use in the HPC ecosystem and having an operational environment as the primary target.”
It is exactly what a development board is designed for, isn't it?
Cineca, the largest supercomputing center in Italy, have already ported several HPC applications and libraries for verification, and will continue the effort with Monte Cimone. According to the article, for instance, an InfiniBand software stack is being ported on the system thanks to the boards' PCIe slots.
This cluster will help future RISC-V supercomputers in EU as well as other countries. Bravo!
In minnesota I had lots of hardware connections through various contracts where when they were getting rid of old equipment I could just take a couple for free. Since moving I don't have any hardware connections and would like to build those relationships again. I want to get 10 or so minipcs to run as a cluster for projects. Anyone know where I can get free or cheap pcs?
An idea I've been toying with is setting up a cluster of raspberry pi's for running of my docker containers. Currently I only have a hp microserver running open media vault. I managed to get docker on it and have containers for pihole, bitwarden, nginx and nextcloud. I'd still like to have more containers running but my poor microserver is starting to chug a bit and I've had docker itself outright crash on me.
So I had an idea to offload all those containers onto a cluster of say 4 raspberry pi's and have the microserver just running as a nas with nothing else. The reason I was thinking a cluster might be appropriate is for the failover. If say the pi running my pihole dies the pihole container can migrate to another pi and I can still use it. Does it work like this or do I have a misconception of what a cluster computer is and this doesn't work at all?
Genuine question: Are there actual benefits other than potential cost savings?
Is it a raspberry pi enthusiast preference?
I think I can fit up to five compute units on my main board, which I manage through Kubernetes. I don't believe the ones I'm buying are designed to work directly when plugged in through pcie, and need some type of container. Would this be worthwhile for mining Monero? Would xmrig work with Kubernetes with some tinkering?
I want to set up a small cluster but don't have much room so SFF computers would be ideal. A Pi cluster simply doesn't have enough computer power for my needs. I am hoping to find computers with processors that support DDR4 ( so at least 6th gen Intel or Ryzen), and come with 4 GB of RAM (ideally 8 each). No need for storage as I have plenty of old HDDs.
Price range I am hoping to stay between $120-$175 per computer and buy around 4.
Hello all,
I've been apart of the golem network for a few years now using my old and retired linux pc. I'm thinking of creating a small cluster of simply nuc pc's to create a linux server to mine golem. Is anyone running a system where they receive an individual golem task which is distributed between many computers before being solved and sent back.
I'm also not 100% sure on how to scale. Do i buy a few high powered cpu's (More GHz less cores) or should i buy many low powered cpu boards but i would end up with more cores.
I'm wanting to play with cluster computing. I'm wondering if it's worth using in a dedicated server for a video game.
Space Engineers uses CPU and RAM more heavily than most other games I play, so that seems like a good challenge for a cluster computer. But can anyone tell me: is it feasible?
The game requires minimum 3gHz and 8GB usable resources. Would stacking two 8GB Rpi-4GB theoretically get me there? From my understanding they have a 1.5gHz processor.
How exactly does cluster computing work? Willing to do some reading, just not sure where to find introductory information on the subject!
I was recently thinking that second hand laptops can offer much value if run on Linux. Later on I thought about if it is possible to combine their powers. First I thought about VNC or whatever it is called, that is to share the display. Then I thought that the computational power of the client lappys will remain largely unutilised. Then I thought about sharing a keyboard and mouse across them. Finally I thought if it is possible to share the storage, i.e. the root across them.
Initially I thought about booting all of them with shared root from a NAS or just use one of them as storage. Then use them individually. Like running GIMP on laptop A on a i3 and Inkscape on laptop B on a i5 both having their root partition on laptop C.
I can imagine how complex it can be. The most obvious disadvantage of this is the network latency. The possible race condition that may arise while multiple devices asking for disk read/write to the storage is also a concern. Also are there any microarchitecture level wiring that may prevent a software to run in different architecture ? How the kernel will handle this varied setup ? How much practical is the idea?
Background : It is possible to have many laptop, even desktop handed down because they cannot run propitiatory OSes well.
I'm about to buy an XJ (2014 3.0 V6D Portolio SWB) - used, with around 50k miles. Noticed a crack/dent in one of the rear light clusters, which the sales garage are fobbing me off with excuses about replacing. They mumbled something about Jag needing to key/program the lights and needs to be done by Jag.
So really just looking for some info before I finalise purchase - what are the rear light clusters like to replace in terms of cost etc?
I also noticed that the main display screen in the dash is super super low resolution and looks like it was replaced with a cheap knock-off (non-Jag) with crappy pixelated resolution. It looks worse than my older XF Portfolio display was on the 2012 model. Is the resolution supposed to be crappy or has the previous owner replaced with a crappy alternative?
Cheers in advance!
I have to swap out my instrument cluster in my 01 Hyundai Elantra because I'm not getting illumination by the backlights and a few of the other indicator lights. I troubleshot the thing to hell, so I think I'm just going to swap it out.
Will the new (used) cluster display the correct mileage, or is that stored in the cluster? If it's in the cluster, I suppose that's a problem I need to bring to the dealer?
I would like to point my phone camera at an icon and be told what app or company it belongs to. Because logos change overtime and have varieties this would involve cluster analysis. Labeling clusters in this fashion seems like it would also work for rocks, sticks, et cetera. Could this be how language developed?
Apologies if this post isn't welcome here, I couldn't find any rules.
I'm a college senior in physics in the US with an amateur interest in computer hardware and limited experience with programming via my degree.
I wanted to build a cheap server out of used components to host a dedicated server for myself and my friends to play various games on. When looking for used parts I came across an opportunity to purchase several old Dell Optiplex 760's for $10 each. They have intel core 2 duo 7200 CPU's inside them and (reportedly) are all functional. An individual one of these wouldn't be powerful enough to run a server on, but I was thinking that multiple might. I've worked on my university's supercluster doing simulations, so I got the idea that maybe I could build my own mini-cluster out of these cheap machines. Would, for example, a minecraft server, with it's proprietary software, be able to run in parallel like this? Or steam servers for any number of their games?
Also, the systems are old enough to lack windows 10 support. Assuming I wasn't able to get them running window's 10 reliably, would running these servers on windows XP or Vista cause any issues?
I have the four-week winter break off as of right now so I'm willing to dedicate a fair bit of time to getting this working, as long as it's possible.
Notably, I would also be in this for the experience, as I enjoy messing around with and learning about computer stuff. So it does not need to be the most practical option, it only needs to be viable, which is why I'm wondering: Is this possible/reasonable? Or would I encounter too many issues for this to be worth it?
Edit: Thanks for the responses everyone. It's finals season so I don't have time to reply to everything (I really didn't even have time to make this post but I fell down a 4 am rabbit hole) but I appreciate all the info and will try to reply when I get the chance!
I'm cleaning up a small domain and I've never worked with Failover Clustering before as a junior admin and I'm finding several computer objects identified as Failover Cluster Virtual Network Objects in AD.
My question is - with no documentation available is there any way to find out what computers are tied to these objects in AD? I'm guessing more than one of them is stale but I have no way of checking.
Apologies for my lack of knowledge.
Thanks all.
HI all, hope you're staying safe and well!
I have a launchpad setup, where a launch computer in the form of a Raspberry Pi Zero is what commands the countdown and ignition. At the moment, I have a pack of eight AA batteries which provide more than enough charge for a quick ignition, passing through a small relay which allows the Pi to control the circuit.
I'm now looking at a new rocket design with clustering for the first time - in particular, three Estes E12s. My issue is trying to get a good, clean and reliable ignition. At the moment I understand around 3-5 amp for each ignitor (wired ideally in parallel) would be a good number to aim for. Obviously I'm aiming for the higher end of that, which wouldn't be possible given the relay can only switch up to 10A at a time. This would give 3A max for each ignitor. Therefore, I'm looking for a better way to reliably ignite the rocket!
My absolute backup is a car jump starter (the portable kind with USB ports etc.), which according to the box can give me at least 200A. More than enough, but too much for the relay (which is very important!). My solution at the moment, is to buy two more relays and pass another 8-pack of AAs through each - I know this works individually and wouldn't be too costly. I was wondering if this sounds like the best idea for you guys, or if anyone has some better ideas for triggering multiple ignitions from a computer?
Any help is much appreciated!
Relay link: https://core-electronics.com.au/gravity-digital-relay-module-arduino-and-raspberry-pi-compatible.html
Anyone who's done a lot of work with file systems knows that sometimes you just find weird shit in there, especially on computers that see a lot of use by techie but forgetful types, like me. And I've lent my laptop to my younger brother a few times, who's even techier but not especially forgetful. More irresponsible, as in I found a Tor browser installed on my machine the last time he borrowed it.
Which was exactly why it was the last time he borrowed it. Tor's got its legitimate uses, but mostly it gets used for shady shit, especially when it's a sixteen-year-old we're talking about, like my brother. That was just a couple months ago, so yeah, finding a strange folder in my little impromptu bout of hard drive spring-cleaning wasn't a surprise.
What was a surprise was the sheer size of the thing. Bigger than the entire drive it was on, by several orders of magnitude. Which is of course impossible, or rather is of course a glitch, which made me sigh and think I'm gonna have to just format this bastard, it's something I'm overdue for anyway, I've just been putting it off for all the usual reasons.
But then I started actually poking around in there, and...maybe glitch, maybe no, because everything I clicked on had actual data attached. HUGE quantities of data attached. No images or 3D models or anything, that would have made the size of the thing even more unbelievable, just endless statistics, the kind you'd expect to see in an especially dry astronomy journal, not that I've ever read one. Only way more detailed; I may not be a specialist myself, but I'm morally certain our scientists don't know that much about planets around other stars.
Finding Earth in the directory wasn't a surprise either; of course it was included, why shouldn't it be? All the data was there, only every one of the planets I'd been perusing before had been frozen lifeless rocks or star-blasted hellscapes or just your basic gas giant. Earth's got life. Earth's got civilization, which meant a lot more data. Like a whole separate Wikipedia in there, describing different cultures, tech levels, social norms. Domesticated animals ("associated species") and plenty more. And...
wait
wait
what the fuck
no really what the everloving Hell was this?
Threat level. For a delirious moment I expected to see "Mostly Harmless" there, but no, the rating was
... keep reading on reddit ➡So a botnet is a group of computers that report to a centralized C&C server and can be used for distributed computing... Could a cluster computer be described as the same? What are the practical differences between the two?
Hello guys, I love Kubernetes, fell in love few years ago and never look back. I use it for almost every kind of deployment for my application. I have been thinking if somebody has any experience or knowledge, is it possible to join your local computer as a worker node to cloud Kubernetes cluster( hosted on Linode). I have a decent computer with lot of computing power and RAM and GPU. When my computer is on, I could distribute a work me efficiently and less costly. In theory I could target this node with any kind of work that stays idle in cloud until my local worker node goes online. In that sense, my local node would be bouncing between online and offline status. What do you guys think? Thanks
