New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
[LET's 2024 #1 TOP PROVIDER*] Free VPSes, Shirts, Hoodies, and Firearms! -- Final Yearly VPS Chance!
This discussion has been closed.
Comments
@crunchbits any chance of restock?
They've also been in the business for quite a while offering HPC services. They own 2 of their datacenter locations (I think), own all of their other infrastructure, and have the means to provision rather expensive GPUs like the A6000.
Little bit different from the 17yr old kids renting dedis or selling LXC containers.
WHERE IS HIS 4TH FINGER ON THE LEFT HAND???
AI ate it
It's a fair concern. That whole most recent drama is just... mind-boggling. Especially the provider's response. If things came to that point where we simply couldn't make it happen, I would know before it was too late and give some time to pull data or at least pull disks and rotate them up for people to grab data. While we're not a 1-man shop, a lot still rides on my shoulders specifically so if something were to happen to me I can't say it's entirely impossible for the doors to close. I can't speak to rest of the teams desire/ability to continue operations. Either way, it wouldn't just be rug-pulled.
Since I doubt you're the only one with this concern, I'll explain a bit more behind the scenes (with emphasis on yearlies):
@Crab is 100% right. Have backups! You can lose data even without a catastrophic event like a deadpool. It could be as simple as filling up your VM too full and then being unable to boot/recover.
These yearlies are profitable (though not enough to be our sole business) simply because it's letting me leverage 2 things we already have:
Being on LET and seeing this hardware that we had 100's of in storage since ~2018/2019 allowed me to bring already sunk-costed gear online, recoup some of those costs, and fill in where we might have ~1kW left in a rack which is too small for most GPU chassis (safely). The only trick here has been getting rates knocked down a bit on software that we use to run everything and IP space cost (if it's on a leased range instead of owned). If you were to do Yearlies at this price continually, you'd have to oversell them a lot further to make it worthwhile per node and/or cut storage sizes down (just in case any start-up wants a blueprint). Problem is the return on those chassis just isn't enough, and the entry cost into the hardware is still too high IMO for its age. It's still a perfectly viable platform--all of our internal stuff runs Xeon E5 v4 era--but not quite as sexy as Scalable/EPYC.
However, the cost of storage is just creeping too far up. Where I used to get large batches of good quality enterprise high-endurance SSDs 1.92/2TB ~$30-50/ea, the cost is now $100-$150 each and quantities are much much lower. That's the main reason for having to stop the product. The alternative is to cut storage in half and run 960/1TB units or double-up on tenants per node but I'd rather not--anyone who got in has it locked and that'll be that.
We're about to send out notifications for PA moves. Once that happens and we've given enough time to get through existing customers whatever stock is opened up/left over we'll throw up public but I assume it will be gone quick.
We used to own our own locations, but as of the end of 2023 the last one finally closed out/completed sale/title done. It's kind of a weird situation, back when the focus was purely "how much power can I get, and how cheap is it?" it made more sense. Parts of the PNW were under 2c/kWh for up to ~5MW. I think even to this day above that, it's still the 4c/kWh range which is amazing for the kind of reliability and robust infrastructure delivered. That being said, all the additional costs that go into it (generators, landscaping, building repairs, cooling systems, security, general maintenance and testing, etc) need to be factored in and we were just kind of in that weird middle-ground where I could skip all that and run a solid-enough tier 1 facility but it was going to be hard to grow beyond that. We still had to deal with all the extras of running your own site that squeezed an already near-understaffed company. It didn't help that Microsoft was our neighbor to the East, Dell bordered us on the North, and Microsoft was buying everything up to the West so we were being squeezed a bit: couldn't find electricians without bringing our own in from 2.5+ hours away, nobody willing to give access to fiber (outside of County/PUD network which maxed at ~1G) to some small shit like us because the big daddy was paying to run brand new dedicated ~800 mile runs across 3 states. Our use case outgrew the facility, and the amount of capex(+timeline) required to maybe compete would have just been too much.
Also will start sending out messages this evening to prize winners. I needed some time to recoup myself, and check the plan with @skorupion about how best to log everything.
You continue to amaze me @crunchbits , your Top Host tag more than deserved.
I sincerely wish you all the best in your business endeavours!
Thank you for the very detailed response. It makes me and I'm sure many others feel a lot better. I wish you continued success, you definitely deserve it.
@FAT32 WE LOVE YOU!
I'm more interested in OKC location. Not going to happen soon?
looks like the 4.5gb islands now have PA as an option and have gone from zero to -1 availability
Website is down ? 402 payment required?
nope
Must be me.
Must be. Everything working on my end too.
PA is loading......
Waiting for @crunchbits and their #DOMEBIGPAPI promo code (and bigwapi ofc)
Be the first to know:
https://crunchbits.monster/
It felt like you were talking about a few hours, now it's been several days, can you tell us when the move will be possible, please?
Is your VPS down or something? It's not as if you have to wait for PA to start using the VPS.
Well, moving to a different datacenter means new IPs most of the time.
This is why I would rather wait to transfer my data from my old provider too. I have an invoice due to be paid in a few weeks so I would rather hold off and transfer everything once PA is ready. If it isn't ready within a few weeks then I will transfer everything to WA in the meantime but I would rather only have to move everything over once (and not risk breaking something by transferring everything to another data centre).
I hate bugging people about stuff so I am just waiting to see what happens but it would be ideal if PA is ready soon so that I can move everything straight over without complications.
All in crunchy
Unsure if they’ve mentioned the process, but I had assumed that they’ll move your VPS over themselves and inform you of the new IP address (before or after), and that you won’t have to manually move the data if you opt into cutting over to PA. Shut it down from WA, taken over and bring it back at PA. That way it’s quick and painless for the users (other than fixing DNS) and they get the old server resources back without needing to support some handover period.
They commented this won’t scale for larger storage of >1TB, but it’s how I expected the 1.5/4.5 plans from this topic to move.
I may be wrong but.
I'm also just waiting, because they mentioned having 200-400 of these up in PA by Feb 1st. I figured why do the work twice when I can just wait a few days and do it once.
all quiet in the western front.
Ah man, regretting not picking up 2 of these now
Took longer than expected, I should know better by now
. Early testing and two live WA units caught a nasty firmware error on these chassis (slightly updated platform from normal yearly units) that only became visible under moderate-high load which put a halt on everything until we were confident it could be repaired. Unfortunately, not something readily and easily available online so we had to work through it internally and make a permanent patch in-house.
That being said, there was no guarantee on timeline for PA. It's best effort, gear is racked and ready. The only thing I hate more than paying for rack space, transit, and power that is unused is putting customers on it when it's not ready and dealing with a barrage of tickets/support requests and a damaged reputation where nobody is happy. Trust me: I want you all on it as soon as possible so we can re-monetize the leftover space/cleared space in WA.
Luckily this controller firmware issue cropped up early enough that only a small handful of recently deployed customers had ~30m of downtime yesterday (Sunday 11th) for us to apply the fix. We're in the final hours of testing now, and if all looks good then an e-mail, discord thread, and a post here with the instructions for PA will be going up today. I appreciate all the patience, and you'll just have to trust me that not pushing last week was for a very good reason.
For me personally I am not in hurry, take all your time, I know how hard it is to get something done, sometimes I write code and I think it will be finished in one day, then 2 weeks later and I am still not done
I don't think anyone will be made when there is an ongoing communication, just my opinion 
Ohh I know that one
Usually hardware is a bit more predictable, but in this case it's a huge "IF" to move people over and pray (never a good idea, asking for trouble) or just prep some of our original older chassis and get those out to replace them. I don't think anyone is mad though, just an update. Prior to the firmware patch yesterday I still wasn't sure if it was fully solved since the issue was only coming up once units were loaded and running customer workloads, making it somewhat hard to reproduce.