Old laptops as low cost servers? Absolutely, build a homelab, rent a cheap VPS, set up wireguard and viola - instant data center for tens of dollars per month. It's not production grade but you'll learn a ton.
But colocation?
Strip away the learning component and add an production uptime requirement - why would you even consider using crusty old laptops for this? If you have production grade needs, look to a standard cloud provider or, at the very least, a colo facility where you can put production-grade equipment.
> Your old laptop packs more CPU power, RAM, and storage than their entry-level offerings - and with us, you'll pay just €7/month for professional hosting
This is basically the same price as the cheapest options on Hetzner: https://snipboard.io/C9epWo.jpg. Sure my old laptop does have more RAM and a bigger SSD, but I bet it's also less reliable than Hetzner's servers, and is likely to suddenly die some day. So is the tradeoff really worth it? It's hard for me to believe that this is a genuine improvement for most things. The only definite winning case I can think of is if I have a process I want to run, but I don't care if it just suddenly stops working. But when would that ever be the case? and to save a couple dollars per month?
> I bet it's also less reliable than Hetzner's servers, and is likely to suddenly die some day
I’m a happy Hetzner customer but I have had servers that I rented from them die a couple of times.
I rent physical servers from them that have been previously rented to other customers. At some point hard drives fail.
However, I have solid backup setup in place (ZFS send and recv to other physical hosts in different physical locations) with that in mind, so I haven’t lost data with Hetzner. But if I naively did not have any backup then data would have gotten lost a couple of times.
Of course. Just pointing out that one can’t assume that just because the hardware might be server grade, doesn’t mean that the risk of hardware failure is negligibly low. And that one always needs to have offsite backups.
Not sure how Hetzner works, but do they have IDRAC type access to their servers and/or remote hands available to fix stuff? Guess you'd be on the hook for that sort of thing here, making the Hetzner price more appealing if they do include that kind of functionality.
Its a page hosted on CLoudFlare's "pages.dev" service. Their method of contact is a Google Form which does have an email address on this domain "CoLaptop [dot] com", but that as a web address does not work.
Not sure if this is legit... I could see it working well enough if they require the laptop to support at least say thunderbolt3/usb4 then they can use a single connection interface to a management/dock interface that includes a network connection (1gb/2.5gb)
The trouble is a lot of laptops won't power-on with the screen closed and have heavy sleep/suspend behaviors in general. Not to mention general airflow in whatever shelving system is used with the laptops, assuming 2-4 laptops per shelf, per 1u. Not to mention, one would probably want/need some means of ensuring appropriate driver support, or an appropriate Linux or other setup for said hardware.
While I can see it working, depending on shipping costs can definitely see some problematic bits.
+ The usual limiting factor in data centers is power, so laptops could be more optimized for greater cycle efficiency per power than comparable old servers.
+ Laptops are generally compact and so achieve greater rack densities than individual co-lo servers. I'm thinking about 34 or 51 laptops could be stored in 9 or 10U either 2 or 3 rows deep by 17 wide.
+ Shipping a laptop to a co-lo data center is cheaper than a 1U server.
~ Reusing electronics saves e-waste and reduces unnecessary consumption, either old servers or old laptops.
- Laptops lack ECC RAM.
- Laptops typically don't use nearly as fast CPUs or RAM as contemporaneous servers.
- Laptops are limited in their storage options.
- Laptops lack remote, lights-out management of real servers.
- Repairing old failed laptop components is more difficult than old servers.
~ Old laptops tend not to have usable batteries, so there's unlikely to be much an inherently distributed battery backup capability.
- Old laptop batteries of various origins could be a li-ion NMC fire hazard at scale.
~ Reusing old stuff at any sort of scale would prefer standardization, and it's sometimes difficult to amass many of the same discontinued model.
Conclusion: Do it if it works for you. It's kinda cool.
Eeek, I can't imagine what this is like if it scales. What happens to the fire risk when theres 20,000 laptops with aging batteries all sitting together? I hope they take the batteries out, however many laptops use batteries to smooth out power fluctuations.
Laptops aren't designed to be servers - peg your laptop CPU and GPU at 100% and see how long it lasts, I've done this before and the answer is about "2 months", yep sure, this effort isn't targeting that workload, but how many bad apples does it take to start a fire? In their page they say "kubernetes server - no problem" kubernetes DOES keep the CPUs busy, not pegged, but busy enough so that they wont step down their frequency.
I admire the effort to reuse old tech, but boy oh boy would I not want to be a sysadmin here!
My old Lenovo t420 has been running 24/7 pegged as a multi-camera DVR since 2011, no issues whatsoever. Of course the battery is removed, but I don't see many decent laptops struggling running under load for prolonged periods.
I worked for a place that did something akin to this in the early 2010s. Someone figured out how to add 32-bit company laptops to the virtualization cluster (likely because they were using one as a stand in for a server that at the time would have been in the works but not yet purchased) and so once that work had been incurred they just kept "retiring" unserviceable company laptops to the cluster. Imagine a standard wire metro-rack crammed in a telecom closet beside a normal server rack. Now imagine that metro rack literally full of Toshiba Satellite Pro's from about 2005-9. The cluster hosted virtual machines for testing.
No fires, no hardware problems. No special cooling other than the mini-split that was in the closet to cool the server rack. They just kept trucking. But modern hardware is much more high strung and I don't doubt you'd have weird failures.
Edit: Back then VMs were how things were done and RAM was seemingly always the bottleneck by a mile, so the cluster did add up to a meaningful amount of extra performance compared to not having it.
uh yeah i mean we 'colo' at work because its cheaper than buying a windows server with multiple RDP licenses. We have some legacy stuff that must be run on site.... so we buy $200 laptops and people can remote in for years.
But colocation?
Strip away the learning component and add an production uptime requirement - why would you even consider using crusty old laptops for this? If you have production grade needs, look to a standard cloud provider or, at the very least, a colo facility where you can put production-grade equipment.
> © 2024 CoLaptop. All rights reserved.
Website copyright is out of date by two years... And the website has been online since then. https://crt.sh/?q=colaptop.pages.dev
> Thank you for your interest. Please submit the form below and we'll get back to you within 2 working days.
> - Team @ CoLaptop.com
Also colaptop.com is not even registered anymore. If I had to guess the pages.dev site stayed up but the domain and email are nowhere.
https://lowendbox.com/blog/little-machines-in-big-datacenter...
I am sure that some of them either already colocate Pico ones too, or are willing to do so if asked.
I wonder if Hetzner knows their aim.
> We might modify your laptop to remove or power down the battery, wireless radios, etc. to ensure it can be used safely in the data center.
Yeah, just use the DC's UPS.
This is basically the same price as the cheapest options on Hetzner: https://snipboard.io/C9epWo.jpg. Sure my old laptop does have more RAM and a bigger SSD, but I bet it's also less reliable than Hetzner's servers, and is likely to suddenly die some day. So is the tradeoff really worth it? It's hard for me to believe that this is a genuine improvement for most things. The only definite winning case I can think of is if I have a process I want to run, but I don't care if it just suddenly stops working. But when would that ever be the case? and to save a couple dollars per month?
Edit: Maybe this is what github is doing :P
I’m a happy Hetzner customer but I have had servers that I rented from them die a couple of times.
I rent physical servers from them that have been previously rented to other customers. At some point hard drives fail.
However, I have solid backup setup in place (ZFS send and recv to other physical hosts in different physical locations) with that in mind, so I haven’t lost data with Hetzner. But if I naively did not have any backup then data would have gotten lost a couple of times.
Just monitor them so you can act proactively.
The linked one is VPS, so all trouble fixing is easier.
Its a page hosted on CLoudFlare's "pages.dev" service. Their method of contact is a Google Form which does have an email address on this domain "CoLaptop [dot] com", but that as a web address does not work.
I'm not sure they have their act together.
- Mount something in a rack not firmly attached to brackets or a shelf
- Install anything with a battery larger than you'd find in a RAID card
Not to mention all the other ways this is sub-par in terms of airflow, density, serviceability, out-of-band management, etc.
I get the allure of it, but I wouldn't really want my gear anywhere near a bunch of laptops stuck in a cabinet.
The trouble is a lot of laptops won't power-on with the screen closed and have heavy sleep/suspend behaviors in general. Not to mention general airflow in whatever shelving system is used with the laptops, assuming 2-4 laptops per shelf, per 1u. Not to mention, one would probably want/need some means of ensuring appropriate driver support, or an appropriate Linux or other setup for said hardware.
While I can see it working, depending on shipping costs can definitely see some problematic bits.
+ The usual limiting factor in data centers is power, so laptops could be more optimized for greater cycle efficiency per power than comparable old servers.
+ Laptops are generally compact and so achieve greater rack densities than individual co-lo servers. I'm thinking about 34 or 51 laptops could be stored in 9 or 10U either 2 or 3 rows deep by 17 wide.
+ Shipping a laptop to a co-lo data center is cheaper than a 1U server.
~ Reusing electronics saves e-waste and reduces unnecessary consumption, either old servers or old laptops.
- Laptops lack ECC RAM.
- Laptops typically don't use nearly as fast CPUs or RAM as contemporaneous servers.
- Laptops are limited in their storage options.
- Laptops lack remote, lights-out management of real servers.
- Repairing old failed laptop components is more difficult than old servers.
~ Old laptops tend not to have usable batteries, so there's unlikely to be much an inherently distributed battery backup capability.
- Old laptop batteries of various origins could be a li-ion NMC fire hazard at scale.
~ Reusing old stuff at any sort of scale would prefer standardization, and it's sometimes difficult to amass many of the same discontinued model.
Conclusion: Do it if it works for you. It's kinda cool.
Laptops aren't designed to be servers - peg your laptop CPU and GPU at 100% and see how long it lasts, I've done this before and the answer is about "2 months", yep sure, this effort isn't targeting that workload, but how many bad apples does it take to start a fire? In their page they say "kubernetes server - no problem" kubernetes DOES keep the CPUs busy, not pegged, but busy enough so that they wont step down their frequency.
I admire the effort to reuse old tech, but boy oh boy would I not want to be a sysadmin here!
No fires, no hardware problems. No special cooling other than the mini-split that was in the closet to cool the server rack. They just kept trucking. But modern hardware is much more high strung and I don't doubt you'd have weird failures.
Edit: Back then VMs were how things were done and RAM was seemingly always the bottleneck by a mile, so the cluster did add up to a meaningful amount of extra performance compared to not having it.