Vultr introduces $2.50/month plan
vultr.com
vultr.com
https://www.webstack.de/blog/e/cloud-hosting-provider-compar...
Edit: I've submitted it as well: https://news.ycombinator.com/item?id=13798023
Your numbers match the numbers on vpsbenchmarks.com [1], so it probably is as bad as your test suggests.
[1] https://www.vpsbenchmarks.com/trials/ovh_performance_trial_1...
What was your experience since you are in Germany too?
I remember being on an independent ISP (VDSL reseller), and my neighbor down the street was on the incumbent Cable provider. When I did a traceroute, I discovered my packets were being routed across the country, south, across the country again, and back north, before getting to my neighbor.
Sweetest thing is hourly pricing. Sometimes I need Paris VPN for few hours. I have some scripts, so I can launch new instance with configured OpenVPN in 2-3 minutes, use it, then dispose and I would pay only few cents. Impressive, if you ask me. That's what cloud is for.
Network speeds are quite good, vultr has lots of features (want to install OpenBSD from your own ISO via web 2.0 KVM, no applet nonsense? Or Windows XP? No problem at all), and prices are just so low now. Also IPv6 works fine. I don't know about their support, I never had to contact them yet, but otherwise it's awesome service.
There are options to load your own distro image and init scripts but I have not tried them since most things I need in a transient instance can be installed by following a simple bash script calling apt.
I think, you can snapshot the entire system, then delete server and recreate it from that snapshot later.
But yes, if you're just stopping server, you'll pay for it.
My Scaleway box's specs:
- €2.99/month (currently $3.17 USD/month)
- 2 GB RAM
- 50 GB SSD storage
- 200 Mbits/sec of unmetered bandwidth
- data centers in Amsterdam & Paris
When I need to take hosting more seriously, I'll reevaluate. Until then, Scaleway is a damn good deal :)
Unfortunately Scaleway is Linux-only, they use some networked block storage that's not yet supported on FreeBSD apparently. (Needs to be mounted from inside the VM I assume? Damn "cloud" stuff.) Scaleway also used to have no IPv6, I think this might have been resolved though (but one address only IIRC? no /64?)
prgmr is my current choice for personal stuff — 1.25 GB is not as good as 2 GB, but they offer normal Xen VMs with full access — install anything you want. Also the good feeling of supporting a small team of sysadmins instead of a big corporation (esp. vs Amazon/Google, not so much with the other small hosters). The only downside is that they only have US servers, nothing in Europe.
This may only apply to their physical offerings, not to the VM-based ones, and you may be able to get it to run using some nasty hacks (possibly involving loading your kernel after booting theirs). Either way, there are no official FreeBSD images so you'd be on your own.
Their IPv6 support is indeed lacking (one address, and you lose it if you stop the server).
https://github.com/scaleway/image-proposals/issues/11
"FreeBSD on x86_64 (C2S, C2M, C2L, VC1S, VC1M, VC1L) needs an alternative way to store data or an NBD driver for FreeBSD, and more work on the boot system"
Yeah, that's kind of strange—what kind of SAN product doesn't even expose an iSCSI target for hypervisors to talk to?
I guess I'm interested in what other costs beyond the monthly keep-it-alive fee can come up with this tier of hosting
Currently DO does not charge for network overages. I guess they would suspend your node for going way over the prescribed limits.
I'm not affiliated with DO, but I just really like the service.
https://www.digitalocean.com/community/questions/extra-bandw...
Does anybody else?
Anyone have experience with them? I'd love to hear about it.
In my experience, all 3 are very reliable: zero downtime since I started using them 8 months ago. Vultr has the nicest dashboard: it uses the noVNC HTML5 client to give you access to the console, it supports 2FA (TOTP) for authentication, etc. They give you $50 of free credit to use in the first 60 days (or did, when I signed up). Plus they accept Bitcoin payments!
Only minor complaint I have is when you create a vps they don't give you its SSH host key fingerprint. You have to log in the console to get it...
I immediately switched. This way I wouldn't have to deal with the generally more messy and weaker[1] OS that is FreeBSD, nor would I have to use old pf syntax.
[1] "Theo De Raadt Says FreeBSD Is Just Catching Up On Security": https://tech.slashdot.org/story/13/12/16/0121213/theo-de-raa...
I used to recommend it to everybody, but don't any longer. I had two negative experiences that made me stop recommending it.
The first is frequent restarts. Very small sample size, but both me and a friend started experiencing restarts every month or so. I don't think I got any advance warning about the restarts, but I'm not positive.
The second is that networking between instances stopped working for me in the middle of the night shortly after Vultr did an upgrade. Support wasn't able to help (though they did try) and I couldn't figure out anything on my end. I had to switch everything to AWS to get the site back online before people woke up in the morning.
I was super happy with it before those two problems, though, and the people there seemed competent and professional. I was also using FreeBSD, so it may not be as well supported as Linux (which may have contributed to the networking problem).
We're still on AWS too. Currently spending about $50 / site which adds up quite quickly when you have 20+ sites. Our current breakup is something like (2 x t2.micro - one for website, one for cron created using Elastic Beanstalk, 1 x db.t2.micro for RDS + Elastic load balancer cost). I'm sure we can bring it down to $10 / site if we were to move to Vultr or Scaleway but it's a very small site to pay for the reliability AWS has provided to us all these years (apart the recent S3 outage). So guess we'll too just stick to AWS for now until.
1) The whole system is running on Amazon's Elastic Beanstalk (via Docker). Beanstalk provisions 3 instances for every site (1 Web, 1 Worker/cron, and 1 RDS) for every app with ELB on front.
2) It makes installing updates easy and automatic using deployment scripts which we have on every site.
3) Everything (web/cron/database) is configured to scale up and down automatically, i.e. new servers start as per demand. This may not be possible otherwise.
Works fine, I just moved a site from $5 to $2.50 without issues.
>Our most affordable way to try out cloud computing, this plan features 512MB RAM and 20GB SSD disk storage. This is perfect for test-driving your next application before upgrading to production level specifications.
The bandwidth overage is also pretty okay. ($10/TB in NA/EU, $25/TB in AS, and $50/TB in AU.)
The idea is, if the customer wants to come back and reactivate service they don't have to provide billing details again.
I don't do this though. When customers cancel my startup, I delete the card and cancel the subscription in Stripe but leave the customer.
This is why I am baffled as to why the CC companies eliminated the transient number technology, which was pefect for e-commerce vendors like this. At least with those, after 3-6-12 months the number expires and no damage done when the vendor is hacked. I assume it was dropped because this technology complicated data mining.
"Account must be funded by credit card or PayPal before making a Bitcoin deposit."
The dashboard is nice, the ram is a bit low (but the price point is killer), I wasn't able to deploy a lets encrypt cert without pooled memory. The service seems to be excellenct in the few days I've used it. It doesn't seem to have a "don't ask for 2FA on this device for n months" option but it's a small complaint.
You mean you had to enable swap to run the Let's Encrypt client?
Packaged versions of certbot (from your distribution's package manager) don't suffer from that problem and have fairly low memory requirements.
After the performance issue with my second instance, I wrote them off as not being good for anything serious. And after the file system issue, I decided they weren't useful even for anything casual. Hopefully things have changed, but don't assume that just because you've been up and running for a couple months with good performance that things won't fall apart soon.
Compose is nice for backups and replicas but they drop the connection like all the time and then the replicas start playing who wants to be master. Which brought downtime to my application at times. Also there expensive so I said f that and manage a Mongo instance myself. Since it was a small project that had to much overhead.
I moved from DO was primarily pricing and the lack of data centers.
Vultr's Miami DC is my jam since I'm in Tampa. Atlanta is nice too.
So far I really like Vultr. 2GB for $10!
That DC has great connectivity to South America, which is what I'm using it for.
They customer support worked great!
Minor rant: for whatever reason, you can't spin up VMs using this $2.50/month plan through their API.
Does anyone from Vultr know why? This would be very useful.
Let's say they have physical machines with 96 GB mem that hold 200 1 GB VMs each, in the hope that no one uses more than half their ram. Same for CPU or disk I/O or whatever resource.
Now that the entry level prices are dropping but the offered resources remains the same, people will spin up more VMs. Doesn't it become dog slow for everyone?
edit: The DO FAQ (https://www.digitalocean.com/help/pricing-and-billing/genera...) says:
"Am I charged while my Droplet is in a powered-off state?
Yes. Your diskspace, CPU, RAM, and IP address are all reserved while your Droplet is powered off. If you want to save a Droplet for future use take a snapshot of it and then destroy it. You'll be able to create a new Droplet from the snapshot image anytime to bring it back online."
As an aside; what does Vultr stand for? Is it "vulture"?