Carolina Cloud – One third the cost of AWS for data science workloads
carolinacloud.ioWe're Carolina Cloud - managed data science infrastructure at ~1/3 the cost of AWS.
I left my job earlier this year after watching companies get crushed by cloud bills for workloads that didn't need hyperscaler complexity. Some examples from my previous life: - $1k/month for a basic 16 vCPU VM - $50k/month for a high-RAM instance - Over $1k/month for notebook platform start-stop execution
We built Carolina Cloud for data scientists and small teams who need serious compute without the sticker shock. Our sweet spot: if you're running VMs, notebooks, or RStudio and not deeply tied to AWS/Azure/GCP service ecosystems, we can save you a lot of money.
What we offer: - Standard Ubuntu VMs - One-click Marimo notebooks - One-click RStudio Server and Shiny hosting - S3-compatible object storage (launching soon) - Prepay discounts for commitments as short as 2 weeks - SOC2-certified, HIPAA-compliant datacenter in Charlotte, NC
Simple pricing: $0.005/vCPU/hr, $0.005/GiB RAM/hr, and $0.0001/GiB of hot storage/hr on AMD EPYC Turin processors. A 32 vCPU, 128GB RAM instance runs ~$240/month vs $800+ on AWS.
We're not trying to replicate every AWS service - if you need Lambda + Secrets Manager + S3 with pre-signed URLs, stick with AWS. But if you're a hedge fund running backtests, a biotech team analyzing genomics data, or a researcher who just needs a beefy VM without surprise egress fees, we're 1/3 the price.
Check us out at console.carolinacloud.io - happy to answer questions about our infrastructure, pricing, or why we think there's room for regional clouds built on owned hardware.
I'm very interested in your product. However, I'd like to report a strange phenomenon: whenever I open your website's homepage, although it doesn't seem to use a lot of memory, my Chrome becomes extremely laggy, and afterwards my entire Mac OS becomes very slow. The first time I encountered this, I couldn't determine the cause. I closed all applications and rebooted my computer. After working perfectly for a while, I reopened your website, and my Chrome and OS became almost unresponsive. After rebooting my computer again, without opening any other applications, just Chrome and your website, and it immediately became nearly unresponsive again. Therefore, while I'm not entirely sure, it seems highly related to your website.
We fixed this, please try again, also if anyone knows how to run a Three.js shader without exploding people's laptops I am very open to suggestions.
curiously other pages with three.js demos at the landjng page like svelteflow.dev/ doesn't have this issue
i think the trick is to not attempt to render at each frame (that is, don't try to provide something similar to a game engine). just render it based on user input, like the page I linked
It could be the three.js shader on the page that they use which renders in the background. I think you may be able to disable shaders to test but not sure. Been awhile.
Want to point out, it does run pretty quick on my iPhone surprisingly (and faster than my other dev laptop :-P). Almost forget how much power phones have...
Yeah, I had the same experience. It made the page basically unusable for me.
Oh wow, same thing on my oldish Macbook Air from 2018, using Firefox. Completely unusable.
It's good of you not to jump to any conclusions!
Same on my low end android phone
It is also terrible to scroll on iOS with power saver turned on.
Same here.
I’m confused about what the licensing situation is that allows you to provide RStudio Server instances. Does Posit allow third parties to do that?
> Carolina Cloud
> Charlotte, NC
> bojangleslover
Username checks out!
Seriously though, this looks really cool. And I'm always happy to see other NC folks representing on HN.
I could see using your service for some stuff, so who knows, we may be sending some business your way. It wouldn't be much (for now), but hey...
If you need more trial credits than our pitiful $1 email me at derek@carolinacloud.io. We were inundated with crypto miners from Russia so we had to ban all .ru domains from signing up + shrink the free trial.
People on HN are obviously legit users so anyone reading this please email me at derek at carolinacloud.io and we will give you $250 free credits at least or maybe more if your workload requires it!
What type of workflows do you support? My startup is running a lot of Whisper on user-provided audio clips. Is that something we could run on your servers? If so, I'll email you for trial credits.
Something is really really off with the website I would suggest a remake of the front-end as unfortunately it seems to miss the mark. The product seems genuinely valuable so don't get lost in that crazy background thing - focus on what information should be delivered - outline the important stuff - rn it broke my browser and it feels like you don't have the right focus. I closed it and just don't want to be bothered engaging more with it otherwise than writing this stinky commentary unrelated to what you actually deliver. Keep up the great work it seems you have done the most !
Thanks for the honest feedback. We just fixed this, I was hoping people wouldn't stick around on the landing page for too long since our console lets you do almost everything without needing an account. That assumption was obviously misguided!
What service are you using to run your cloud on top of? Do you have your own hardware (which I feel is unlikely) or are you using any other cloud provider to build on top of
I was interested in building my own cloud (which had a specific niche) and the biggest issues which happened for me personally were that most companies had the issues that if you wanted to create your own cloud on top of, you were responsible for the abuse from your clients and they can shutdown your account if threats persisted/It wasn't really clear for many companies so I was always somewhat worried about it, so I am curious as to how you are dealing with it or what your thoughts are on this matter.
We own all of our own hardware and we rent space in a colocation center in Charlotte, NC. We used to run out of my house since I had 8Gbps unmetered from Google Fiber but ultimately went to colocation since it was relatively cheap and has the backup power/network infrastructure.
Ooh very interesting! Thank you for your response but I have some questions
My family (sort of) has some relations with internet providers in the area so if I actually do follow this, its not something that I am exactly fearing about.
Now regarding the fact that you ran it out of your house, I suppose you had a completely different firewall/fibre to prevent your devices to get any sort of (malware?) since the servers would be running untrusted code
I am more so interested in these two aspects:
Were you worried about the fact that your public ip was available or anything similar? Like I had this fear that it was equal to effectively ddosing myself even though my ip of the current router only gets geolocated to a place 150 kms away from me, I (still?) have that fear
I am not thaat familiar with networking but did you have your own ip ranges or did you provide ipv4 ranges to your customers?
Also how were you managing the fact that maybe your house could've been swatted (I am not sure about the security of north carolina) but I have this deep (irrational?) fear that anyone might do nefarious things since effectively I think that we are only as strong as our weakest link and I am curious if something like this happened with you or did you have any such thoughts?
login.carolinacloud.io resolves to 216.227.218.162 which is owned by https://www.tier.net/ which has a datacenter in Charlotte.
even if the datacenter provider owns the IP ranges and thus gets all the abuse mail, they are kind of incentivized to keep you as a customer? since you are paying them
also, if you get your own IP ranges then the provider won't care much about e.g. your clients scanning the internet / running malware
Another topic is DDoS to your clients which can exhaust the uplink bandwidth... Don't know much of it
We've yet to have any DDoS problems. We don't even use cloudflare proxy which we feel keeps our API and frontend super fast and responsive.
yes, pretty unlikely to have them, as long as all your customers are actually ML people
> they can shutdown your account if threats persisted
That problem can NEVER be avoided at any level unless you run absolutely everything (which is almost impossible).
What everyone does is have a system to quickly pass on and also shutdown who's 1 layer down. You receive a report and deal with the client.
> I was interested in building my own cloud
At the end of the day the problem has nothing to do with clouds. It happens everywhere e.g. if you rented out a house and someone did something illegal with it... how do you avoid it? All the same.
We actually ran out of my house in our very early stages at which case we owned everything. We had Google Fiber 8gbps for $150/mo. The only issue was the Google Fiber terms of service — ie they could shut us down at any time. This was one reason we went to colocation.
> at which case we owned everything. We had Google Fiber
So you didn’t own everything. Google owned the IPs and network.
It’s the same colocating in that your network providers can be shut down.
I guess you're right. Is it even possible to truly own everything yourself? Even if you open your own data center, at some point you have to agree to peer with big fiber providers like Segra, Lumen, AWS Edge Network, right?
> Is it even possible to truly own everything yourself?
Even if you do there's still more to it e.g. blacklists of different kinds e.g. known proxies, spam listings, etc that would still mean you need to kick out your client.
Then there's the bad clients you need to deal with e.g. if they send DoS from your network. Even if you don't get a report it will eat up your limited resources and you may be subject to revenges and bring down your infrastructure.
Conclusion: there's lots to it but so does every business. That's just life.
Yes I agree but there was always this fear I had/still have that if abuse gets too much, some providers might be unwilling to give proper support
This is why honestly I had been talking to more indie cloud providers and realized the whole space of low-endtalk and I think that a good compromise to the whole situation is as if I can rent servers but they can realize that at the end of the day, I am an indie solution myself and this is something which in my opinion is only possible with either large but specific cloud providers (fwiw upcloud although they blocked my account because I hadnt added card details or similar supports it and mostly small cloud providers support it)
If you’re looking for a cheap but reliable alternative, Hetzner offers i7 64GB Ram servers for as low as 37.45€/month.
Of course you’re not in AWS, forget about all the managed services, but we’re talking about 95%~98% cheaper egress costs, with 20TB included in most machines.
We have no egress fees but you're right about Hetzner. They're cheap. We aim to offer more value-added software to make up for it. Also that machine is in Europe, Americans don't like the latency (at least I don't), they have been known to shut people down (within reason).
Also I'm seeing that the most they can go RAM-wise for a dedicated US location is 192G. We go to 512 and will soon go beyond. I'm sure they'll get there soon but that's a consideration as well.
Please go an checkout the whole offer of hetzner is already everything you need.
Hetzner doesn’t offer their cheap dedicated servers in the US, only their cloud offering which is similarly priced to the OP’s example ($223).
What about ingress costs?
Approximately nobody charges for basic incoming traffic to a server. If it's not mentioned it's free.
AWS will charge you if you cross zones.
free
This is a huge market and there are lots of competitors out there. Just FYI that 1/3 of the AWS price isn't really considered cheap when you look at the budget market. Some of the cheap European providers have US presence now as well. That said, the pricing isn't too bad and the lineup looks good. Also, the website also gives me issues on my Android phone.
> Some of the cheap European providers have US presence now as well
Not that offer s3-compatible object storage (which I guess is on the roadmap) and Turin VPS. These prices are legit and I'm not pulling the trigger yet but I'm definitely interested.
Yes we're going to offer this very soon, we have a bunch of SSDs but we're still deliberating on what to use now that MinIO is moving toward closed source.
Landing page or console? We didn't really build our console for mobile though we probably should.
Love that it is in NC! Though, I'm confused who this is for. Is it for genomics data scientists mainly or is it general compute (I would also compare with hetzner and the other clouds if you are going the general compute method). If genomics or data scientists is the main audience, is this going to grow to be more collab notebook plug and play? Very cool and wish you the best!
Really it's for both. One of our cofounders came from a genomics background and so that's why we have the Marimo notebooks + R Studio Server + genomics container. But we've also had a lot of requests for general VMs so we made some of those as well.
Not sure what's going on with the landing page but it's CPU / GPU heavy.
three.js shader. Surprisingly runs super well on my iPhone. ¯\_(ツ)_/¯
Turn on power saver and try again :-/
I looked around for reasonable datacenter colo in the triad and triangle and didn't get anything promising. Where are you physically hosting your infra?
Curious if you are having to buy bandwidth as well. Some of the Midwest data centers include over 30TB of bandwidth in the rack rentals.
And if you are willing to go into the details curious how you are handling bare metal provisioning. MaaS or home grown tooling? Or are you just installing proxmox by hand?
We don't have on-demand API-based bare metal provisioning right now. Sorry if that is misleading on the website. Our bare metal is OTC right now (over the counter).
For the rest of our provisioning (VM and container) I wrote the software myself. It's based on a Django app called the "master" that hosts the console and keeps track of who has rented what etc + a bunch of "host" nodes that listen for instructions from the "master". Pure python, the only thing that's in Go is the CLI.
I looked into Proxmox but ultimately decided I wanted full control. ZFS storage from Proxmox is something I do sometimes wish we had — going to offer s3-compatible storage very soon but I know Proxmox does ZFS out of the box really well.
How do you sanitize hardware between workloads?
> left my job earlier this year
They are probably reselling from hetzner/vultr
We own all our own hardware but we rent from a colocation provider in Charlotte. They provide the backed-up power and network + the IP block, we do literally everything else. We used to be in my house and I was considering installing my own generator (actually not that expensive) but we got a great deal on colocation, shout out Ivan Teoh at Tier.net!
That's a surprising amount of investment, well done and godspeed.
The Hetzner US server on the east coast is in Virginia not North Carolina.
Which actually isn't that far. But I don't think they could fake it if they say the servers are in NC.
Why would they need to buy bandwidth? Many places would offer you a gigabit pipe directly.
Sure, some places charge to upgrade. Some places are metered.
100Mbps is 33TB / mo. 1Gbps is 330 TB / mo. I'm curious what they have and how they prevent saturation from a client or they just pass on networks managed by the DC.
> Many places would offer you a gigabit pipe directly.
i.e. you don't have redundancy? A gigabit is nothing these days sadly. You can get a surge in traffic (DDoS or not) and be stuck quite quickly. You always need extra capacity to deal with issues unforeseen and that adds to the cost.
Co-Founder here! Happy to see this has blown up. If anyone wants free trial creits beyond our crypto miner-defense $1 credit please email me at alice@carolinacloud.io or message me on Linkedin at https://www.linkedin.com/in/alice-pearson-1a0339183/
The console page is showing third party login failure on my phone after I tried to switch to desktop mode because I couldn't see the create instance form. Also the wrbgl stuff on the home page crashes my Android browser after a few minutes.
Did you try to log in with Github or gmail?
I'll bite. How much capacity do you have or some examples of the capacity you're managing?
We have about 1000 cores right now.
We're really excited for the AMD EPYC Venice — 256 physical cores each -> 512 vCPUs -> 1024 vCPUS on a single board with dual-socket. It will probably be about $40k per machine with these RAM prices but we're definitely going to buy a few. A full data center on a single motherboard!
So we're limited on capacity since we own all our own hardware. Please do not use us for auto-scaling just yet. Our software would have no issue with us linking up other cloud machines such as AWS EC2 to our fleet and offering it there, which could help with auto-scaling, but we would not make any money on that and it would be a lot of engineering effort for us right now.
Love seeing NC on HN. Go Heels!
we're 1/3 the price
How will you provide high quality service and reliability while competing on price with the scale and financial might of AWS?
Because in B2B, those things tend to have a higher value than initial cost. Or to put it another way, your customers will be making long term investments by choosing you.
Successfully competing on price in a commodity market requires cheaper access to resources and because price is the easiest way to segment a market, low prices attract price sensitive customers...they are the least desirable customers. Good luck.
You are right for probably 60% of customers. For someone spending $3k/mo on cloud on a company doing $500k/yr of revenue then going down to $1k/mo on cloud for a less tried and true product (us) is likely a bad idea.
Similarly for a mom-and-pop bakery (contrived example) hosting a website for $60/mo, going down to $20/mo (just to keep the 1/3 ratio) also is probably not worth it.
But some of our customers are not like that. For example a hedge fund we have been working with needs 512G RAM and 256 vCPUs for a mortgage model. The data size is not too big and once they get their results they rip it back to on-prem. The complexity is low, ie they just ssh in and do their models. Often they let them run over the weekend.
And these guys are very price-sensitive. In their industry, saving money means more carry for partners and bigger bonuses for quants. These guys are counting nickels.
So I think you're totally right for the large part of the market that we're not really for, and we're not really competing for those types of customers. But we're not really providing much managed service, we're providing a commodity that, assuming you don't need the high-complexity ecosystem surrounding it, can be very nice for customers who are price-sensitive.
Have you found any customers who are too price sensitive for you? Presumably at some point it is cheaper to go and rent bare metal.
Don't put a particle sim on your landing page
Can I attach multiple GPUs to a container?
"one third the cost of AWS" is just AWS with savings plans enabled
> if you're [...] a researcher who just needs a beefy VM without surprise egress fees, we're 1/3 the price
The AWS egress fee is $0.08/GB, whereas Hetzner has $0.00/GB. So, why pay $0.0225/GB?
These companies are the worst kind of scam. If you could really provide a product on par with the big boys but somehow lower than commodity prices, you'd corner the entire hosting/cloud market. But you can't, because they already made things hyper-efficient.
It's like trying to sell someone a $5 hamburger by advertising that some other restaurant sells a $15 hamburger. It turns out that other restaurant also sells a $5 hamburger, it's just not at the top of the menu, because cheap isn't always a sales leader.
The argument I'm hearing you make is that Hetzner needs to license a white-labelled version to distributors. If the servers are really commoditized then why aren't the data scientists in this "AWS is too expensive for data science" market going to Hetzner?
Nice! I'm definitely going to try you for our testing infrastructure.
We’ve removed the Show HN designation for now because it’s not clear that this is something people can play with without paying anything.
If it is possible/easy, please indicate in a reply to me, and we can update the title and post.
Of course it’s still fine to be on the front page of HN if the community finds it interesting. But Show HN is meant to be about showcasing an interesting project you’ve built (with a focus on technology discovery) rather than announcing a product people can buy.
Thanks for the explanation. As a long time user I feel embarrassed to have not known this but I will definitely keep this in mind going forward.