Atlantic.net launches $0.99 SSD-based VPS servers
techcrunch.com
techcrunch.com
Edit:
So their AUP[1] is full of a lot of crappy policies. Not allowed to use all the RAM you're assigned. Not supposed to use all the bandwidth you're assigned. Not supposed to use all the CPU you're assigned. No crons or background services -- what the fuck? No IRC, TOR, or p2p activity. They have an obscenity/morality clause. Cannot do any webdev that uses custom headers -- would even suggest that running wget or curl with a --user-agent flag is in violation.
Edit2:
And they require a valid address and phone number to even register. Screw off.
Please note that in order to protect the integrity of
our cloud, Atlantic.Net verifies phone and other
contact information prior to account activation. Please
be certain to provide a working phone number where you
can be reached in order to avoid delays in the account
creation process.
[1]: https://www.atlantic.net/support/acceptable-use-policy/
But that issue with background services or cron - for that price, that's exactly why I'd consider their service. May I know what might be the reason why they don't allow this?
Nonsense. There are plenty of cheap hosts that don't do that sort of validation.
>May I know what might be the reason why they don't allow this?
Either they copypasta'd a boilerplate AUP -- in which case they're inept. Or they don't know how to properly virtualize and manage those provisions -- in which case they're also inept.
And then their users run warez hosts and botnet slaves, which then get DDoSed, saturating their neighbors' links. No matter how well you virtualize and manage your provisions, if one user has caused your incoming 10G to be saturated on layer 2, there's not much you can do at layer 3 to QoS that.
Personally, the combination of "nearly free" and "personally identifiable such that there won't be people translating 'cheap' into 'good for one-off masks for illicit activity'" is a real selling point to me.
People who run botnet slaves large enough to get DDoSed by their fellow botnet masters are smart enough to own Voip numbers for verification. The verification keeps 12year olds off not really anyone else. But it also stops legitimate users from signing up.
well, you can blackhole the target. Essentially, you tell your upstream to drop all traffic to one of your /32s (the one being targeted) at it's upstream. It finishes the job for the attacker, which is sad, but so long as you are willing to lose the customer in question, it solves the problem for you.
For details of how to set this up if you are a he.net bandwidth customer, see:
http://www.he.net/adm/blackhole.html
nearly all other bandwidth providers provide similar facilities; there are a few other things you can do to make this sort of thing more robust.
But the point is that there are things you can do about incoming DDoS attacks... if you are willing to kill all traffic to the target IP address.
I mean, for most ISPs this isn't automated... in my case, my pager goes off; I log into my quagga box, and I start typing. So you still see downtime, and yeah, you want to avoid it. But there are things that can be done.
I get the impression they've reused this AUP from a shared hosting environment though, as banning CGI scripts, "chat rooms", and background services are not things practical to police on VPSes anyway but are commonly policed on shared hosting. Statements like "Any database stored on Atlantic.Net servers shall be limited in size to 50% of the total disk space allotted for a particular domain" seem to back up this hunch, since it's talking about space for a "domain" and not a VPS or server.
I think they need to improve their AUP as I'm not convinced that it's what they intend for VPS services, too much of it doesn't make sense for that sort of environment.
That subsection is a morality clause, regardless of their explanatory blurb.
His right what?
I'm only making this jab because, like another user pointed out, there's an unprofessional amount of grammatical and typographical errors in your copy.
I respect you guys, and one of my friends has a bunch of cages with Atlantic. But not everyone has that connection or the wherewithal.
You mean "as another user pointed out" :-)
You mean "What?" :-)
You mean "an unprofessional number of grammatical and typographical errors"
Terrible experience.
[0] http://www.gandi.net/static/contracts/en/g2/pdf/MSA-1.3-EN.p... (Note that this is the English copy of their rules, which is not binding. You are bound to the French version.)
The spirit of this is to avoid abuse on our cloud. If you don't feel comfortable providing us with contact information, we're probably not the place for you. I'm working on the AUP tomorrow, the cron/background services doesn't make sense or the RAM.
Why should I trust you with my information if you can't even handle your own?
What kind of abuse do you think it is preventing? How does it prevent abuse from anyone but the least motivated abuser? If I wanted to abuse your service, I could pay people $0.50 on mechanical turk to register accounts for me (indeed this is a fairly common low-skill-low-reward task on mturk).
I'm sorry but this plus the launch of your HIPAA hosting, which was also plagued by typos and wrong ToS, doesn't inspire any sort of confidence in your abilities at Atlantic.net.
LOL. You want to use 1 petabyte for cheap?
If using the service that Atlantic is offering is going to somehow bankrupt them, then perhaps they shouldn't offer it.
https://web.archive.org/web/20141001003619/http://www.pslhos...
Edit: Jesus the more I dig into this the more worried I become about what the hell kind of business you're running. Assuming you're on the only one to use the `mp99e99` handle with any regularity, I'm greatly concerned.
Thanks to everyone who pointed this out.
(I'm exploring the possibility of a new kind of service which requires each user to have their own private cloud server, with all the cloud servers talking to each other.)
I ran nginx as proxy + one web app in a DO instance. CPU usage always < 2% for the whole system and serving 20-40GB of mainly html,js, css, json per month.
Do you consider the web app as background services?
Is it ok to expand my next server instance to your solution per your biz model/AUP?
web app is acceptable. Sounds like a great use for one of our servers!
This has nothing to do with me being untrusthworthy and everything to do with _you_ being untrustworthy. Work on changing my perception of you, or don't.
OVH's CEO, Oles, doesn't care about their big clients and happily changes the terms of their contract (in less than legal ways) to screw them. I'd rather not do business with them.
Source? Especially when accusing them of doing something illegal?
Look around their forums 2-3 years ago. Many of their HG and HGXL customers had their server's 10Gbps link limited to 300mbit, their bandwidth go from unlimited to metered, internal bandwidth change, their bandwidth go from ok to shitty quality (OVH's "volume" network). Similarly, if you had an entire rack of 10Gbps servers (SO ~5-8 servers) they limited the entire rack to 1 or 2Gbps.
Thanks for the thumbs up on this.
Firstly, we are a 20-year old startup! Everyone laughed at us when we talked about doing the original Internet startup -- dialup Internet. Over time, we adapted and rode the booms and busts that went on in our industry, all the while staying cash flow positive and learning to GRIND.
We are thinking there is going to be a explosion in startups globally, that will lead to the next big things, not just in the web but material sciences, biotech, etc.
Our goal is to build out a global compute infrastructure and get it in the hands of tomorrows Einsteins.. wherever they are born. We want to see the future happen faster, and see more of it in our lifetimes. Thats not going to happen if we wait around for today's dominant players, who for whatever reason are moving extremely slow.
Thats our over-reaching goal. You can read about it here:
https://www.atlantic.net/blog/2014/09/30/atlantic-net-launch...
Doing a startup is really, really hard. We're trying to make it less so. We're also ready to provide an alternative narrative to AWS conquering the world -- but we have to be aggressive and go FAST!
So, bottom line is we're trying to do something new,take some risks,be bold, and take on the giants of the industry.
AWS's margin is our opportunity :)
Wish us luck! We're going for it!
what virtualization technology are you using? Is it openvz/virtuozo/container based? or is it kvm or xen hypervisor based?
Are your backups hosted on a different provider or on a different server?
Backups are on another server in the same datacenter.
You might want to look into moving backups to an offsite location. Datacenter administration 101, if there's a fire or an earthquake or any other natural disaster, you definitely don't want to lose all your data because it was all in the same place. Just a "different machine" doesn't cut it.
You know how often earthquakes happen? fires?
You should guard against the common things first. Sysadmin error is a way more common way to lose data than earthquakes or fires. hostile compromise is also a lot more common.
Having the data on another server (rather than in a snapshot) is a good first step because it protects you from RAID screwups, and sysadmin error in whatever snapshot layer you use.
Generally speaking a reasonable defense against compromise is a reasonable defense against sysadmin error... best practices (and I know of no vps provider, including myself, that actually adheres to these best practices.) are to set up your backups so that the production root can not overwrite or delete old backups. Ideally, no one employee has write access to both production and backups, that way no one person, even if their credentials are compromised by a hostile, or even if they become hostile, can wipe all your data. (note, this requires an off-site backup. Physical access is write access, but this really only protects you from an employee who is willing to risk jail time to hurt you, and while that happens, it's pretty rare compared to an employee's login credentials being compromised. I would setup protection so that no one employee can overwrite both production and backup remotely before spending the effort and money to haul all my backups to another location.)
But like I said, as far as I know, nobody actually does that (and its a difficult sort of thing to verify.) - for the low end VPS market? if they have backups on another server (rather than in a snapshot or something on the same server) they are doing okay.
But really... if you care? you should do your own backups. As a customer, you don't have any choice about letting your hosting provider have write-access to production. Make sure you have a backup somewhere that they don't have access.
Yes, let's save money from buying fire extinguishers, that doesn't happen very often anyway. I mean, for big companies with money it's okay to have safety procedures but for small companies we can expect them to ignore it, right? After all "nobody does that anyway" :)
Sarcasm aside, such accidents can wipe out your entire business and destroy your credibility, just because something doesn't happen often it doesn't mean you should ignore it. Plus, depending on the area you live in, earthquakes can be very common, but I digress.
All of this aside, I agree with the rest of what you said, but it doesn't invalidate the fact that if you want to have proper data safety practices you need offsite backups (I'm talking as a service provider, not as a user. Of course you want to backup your data yourself too, as a user)
And yes, all of us know that if we lose the data? we lose the customer, and worse, we lose the customer and they say (justifiable) bad things about us. If you lose all your data? yeah, you are out of business.
Even so, there's a huge difference between how a low-end hosting company is run and how an "enterprise" datacenter is run, and expecting enterprise reliability at low-end pricing is... not realistic.
If you think that your low-end VPS provider is doing everything possible to back up your data... you are likely to be disappointed. Hell, I don't have regular backups, or offer for-pay backups at all myself. I'm working on it for the next version of my management software, but for now, I'm very up-front with my customers that they need to back up their own stuff, and my architecture is such that most of the time, data loss is confined to one server, and yeah, if someone got in and wiped it all out, I would very clearly be bankrupt. but... I'm only saying this because yeah, in this market, you don't get "enterprise level" backups. If you want 'enterprise backups' you have to do it yourself, or pay 'enterprise money'
You can say that it's like not having a fire extenguisher, and I'm not going to argue with you, but I've worked in this sector for well over a decade, and yeah, that's just how things are done. A low-end hosting setup is going to be way different (and way cheaper) than an "enterprise" hosting setup.
and I do know that many of my competitors have gone out of business because both backup and production could be written to by the same user (do you remember HyperVM? it was a disaster for many in the industry.)
I don't know of any competitors that have gone out of business because of physical destruction of their datacenter.
just saying... off site backups are good... but I would get the security and 'defense in depth' setup squared away first, and nobody does, because it's not on the checklists.
more on the hypervm thing:
http://www.itwire.com/business-it-news/security/25559-hyperv...
http://www.theregister.co.uk/2009/06/09/lxlabs_funder_death/
The idea was that lots of people used this cpanel-style VM hosting software. It worked pretty okay, from what I hear. I never used it myself. Anyhow, there was a vulnerability, and some asshole decided to use that vulnerability to hack into a bunch of different providers, wipe the production data, and then to wipe the backups, too.
it happens.
What is going on in your ToS? Sure looks lifted from an older provider... And the morality clause with copyright violation of the ToS is rather funny, albeit sad.
Why should I pay _you_ money?
I guess they flag high risk and only call in those cases.
Unless you sort that out, you're stopping people from buying from you, and that my friend, is your biggest mistake.
My country doesn't have states, at least ones that have any kind of influence at any level.
I don't know if that's the case anymore. By 2007, dialup was dead and the company had pivoted to making most of its money from hosting. I left a few months prior, but still had friends working in the call center when they were all laid off. I don't know if they outsource their phone-based customer service now or if they set up a new call center in Orlando instead of Gainesville, but if it's the former, it's a shame.
It has, though, become more difficult to get additional IPs than before. Linode used to hand out one additional IP per VPS on demand, but now they require a justification for all additional IPs.
I agree, although consumer devices - especially mobiles - could be easily NATed by a provider, no? I'm not too savvy on the demands of mobile devices but I don't think incoming connections are a requirement for anything(?)
But it's notoriously difficult to set up NAT on a large scale without seriously impacting performance, so large ISPs in the developed world tend to avoid it.
It's worth noting, though, that smartphone users don't care whether their traffic is carried using IPv4 of IPv6, so you can slowly transition from one to the other without users noticing. 15% of T-Mobile's connections are IPv6: https://conference.apnic.net/data/37/464xlat-apricot-2014_13...
If you started only giving out IPv6 addresses to VPS customers, I don't think they would be as oblivious as the smartphone users.
There's a lot of cheap VPS hosts which only do protocol 6. For many purposes, that is just fine.
IPv6-only hosts I understand, but I can't imagine that you'd be able to get IPv4 addresses cheaply enough to maintain this pricing for long.
This low price makes me wish that RAM and CPU could be added to cloud servers on an as-needed basis, i.e. hot-swappable RAM. That would be much more useful than autoscaling since it could react to traffic very quickly. You'd need live-migration behind the scenes to make it work, of course.
Edit: nevermind, just saw their TOS
I meant the opposite. They have free reign to censor content.
More likely scenario is a tiny PHP site with the database hosted elsewhere?
(Wow, I didn't know you could just serve content straight off the storage. I need to re-examine AWS.. thanks :D)
They are doing no such thing.
At 99 cents per month, they are targeting people who want things for free, hoping that they can later up-sell, or convert them to a higher plan.
Don't know how well this is going to work in the hosting industry - usually the lower the cost, the worse the customer is.
I'd like to say maybe as someone on the 99 cent plan sees that they need more resources, they could click a button and get on the $9.99 plan, but that might be wishful thinking, and they are just opening themselves up to an enormous drain on resources.
Either way, if there is any indication of people signing up for this, all the other hosting companies are going to match this plan within 24-48 hours, just like they did when everyone switched to "unlimited" offerings. It's a race to the bottom.
Such a strategy could work even better with a VPS since it's harder to migrate away due to the more complex things people tend to do with them.
For instance crissic.net (happy customer) offer 256MB ram and either 20GB SSD or 50GB spinning-rust for 15USD/year.
The only news here is that Linode and DO don't think they can be profitable at these price points (and it's probably not possible when using Xen/KVM that can't easily be overprovisioned).
dd bs=1M count=1024 if=/dev/zero of=test conv=fdatasync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 2.7785 s, 386 MB/s
hdparm -tT /dev/disk/by-uuid/50543a27-2cb6-4d68-9ab4-e09493f946d5
/dev/disk/by-uuid/50543a27-2cb6-4d68-9ab4-e09493f946d5:
Timing cached reads: 14362 MB in 2.00 seconds = 7189.28 MB/sec
Timing buffered disk reads: 1770 MB in 3.00 seconds = 589.21 MB/sec
But, assuming they aren't horribly under-powered on any of the vectors that affect performance, this is a great deal. 256MB is just enough to run a reasonable web server on, even with a database-backed application. Two more servers for DNS, and one for SMTP, and you'd have a solid setup for a large variety of tasks for under 5 bucks, and you'd control all of your services and all of your data. You couldn't run a big community site with lots of users, but you could definitely run a blog (for potentially hundreds of thousands of visitors, depending on how you setup that blog) or simple site.
$ dd if=/dev/zero of=~/test.bin bs=1M count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 3.84267 s, 273 MB/s
root@hackerboosblog:~# dd if=/dev/zero of=~/test.bin bs=1M count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 3.11193 s, 337 MB/s [root@srv3 joe]# dd if=/dev/zero of=~/test.bin bs=1M count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 0.310283 s, 3.4 GB/s
[root@joe2 ~]# dd if=/dev/zero of=~/test.bin bs=1M count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 0.967834 s, 1.1 GB/s
Rackspace Cloud for relatively cheap compute once in a blue moon
For an instance you run only rarely the somewhat higher cost of EC2 instances compared to other providers should be fairly irrelevant. You'll find options from about 30 to 80 cents an hour.
I've seen that being done around here, since the hardware you're describing really is pretty close to a beefy desktop/home server which usually spend their lives idly spinning their fans.
However, they noticed the downtime problem and emailed me saying that a technician had been dispatched to look at the problem. Within 15 minutes, they fixed my GRUB configuration and had my server back up. And, they didn't charge me any for the work the technician did.
It was awesome, and they immediately won my long-time business!
E3 w/ 32GB ram or dual E5 with 384GB ram.
Disclaimer: I work for Rackspace
I gave up when they couldn't call me on my phone and requested a copy of my passport, driver's licence or a bank bill...
I'll just stick with DigitalOcean...
That fact actually blindsided me since I already gave info and couldn't just login and start up the server.
I wish I'd have read stuff carefully before blindlessy jumping into a deal.
ramnode has a roughly equivalent offering for $2/mo (8/q)
Really, all I've ever wanted is an IaaS provider that gives me VMs in the shape of Heroku's instance partitions: a little bit of guaranteed CPU/memory/IOPS per dollar, with cheap/free low-jitter bandwidth thrown in, made specifically for allowing network-server applications that scale horizontally.
My 64MB-of-memory IO-bound Erlang chat server doesn't need a 4GB instance with eight cores. But it has to take that instance, because that's what you pay for in order to get a good uplink.
That being said, some providers let you tweak everything independently. I know Gandi does [1], there are probably others.
(I don't use them, no idea what their cloud service is worth, I'm just showing the sliders)
1 core, 512MB, 20GB ssd = $19.84/month or $13.89/mo prepay compare to DO at $5/month
4 core, 8GB, 80GB ssd = $124/month or $86.88/mo prepay compare to DO at $80/month
Anywho.. multiple reliable hosting options always good :)
The interesting part (and why the pricing scheme that I chose moves in the opposite direction, e.g. you get more resources per dollar as you buy more) - is that while the cost of hardware is pretty linear, there is certainly a support and abuse cost associated with every account, so the lower-end accounts cost the provider more per unit ram to provide.
The trend to discount the smallest plans has got to be mostly 2.
It's an interesting strategy; if it works, it would imply that there is more sensitivity to price at the bottom of the market than there is at the top of the market.
https://www.arin.net/fees/fee_schedule.html#isps
also, at least until ARIN runs out, it's in your interest to qualify for as many IP addresses as you can, and VPSs are one way to legitimately "show efficient use" of your current IP addresses.
Maybe I'm just being a suspicious cynic, but do you think there could be value in a business trying to qualify for as many IP addresses as possible in the short term so that they could be capitalized on when/if the final IP address crunch occurs?
It's really not that hard to fake the "efficient usage" requirement, if that is really what you were after. If you are going through contortions to legitimately qualify for a block from ARIN rather than just lying, my assumption would be that you are doing so for honor, not money, otherwise you'd take the far more efficient route of lying.
(non-affiliated plug) I've been using http://RunAbove.com for something this large, and it's an absolute steal. They're backed by OVH.
Versus a similarly-specced DigitalOcean node for $240 (2.5x RunAbove the cost), we were seeing a performance increase of 35% on our high-volume API.
And that's on non-tuned Postgres/Rails4/Redis Docker containers.
The Registry database contains ONLY .COM, .NET, .EDU domains and Registrars. Domain Name: ATLANTIC.NET Registry Domain ID: 1092612_DOMAIN_NET-VRSN Registrar WHOIS Server: whois.tucows.com Registrar URL: http://tucowsdomains.com Updated Date: 2014-09-16 11:43:09 Creation Date: 1995-05-17 04:00:00
As a business owner, I wouldn't bet anything on that. For personal development stuff though maybe it's great.
It's easy to make programs that run in 1mb of ram and use 100% cpu for example. Ram costs less to be full than CPU does.
My experience with them has been positive. They have slightly higher specs and a good variety of locations.
Wasted my time. GFY Atlantic.net
Personally i find their policy very understandable. Obviously you won't be able to host a TOR node on this thing or push tons of traffic over it. That's obviously not what it was made for. This $1 deal is perfect for hosting simple, static sites or just playing around with the OS. But some people will complain about anything.
The only way to go below $1/mo is by using over-subscription. You don't get dedicated hardware for $1/mo.
With over-subscription, it's a mathematical fact that you cannot use all the processing power you've seemingly been given. The VPS provider could just give you only as much CPU power as you can use at any time (effectively 1/n the total power with n users) -- but that would be terribly inefficient (most of the CPU power would be unused at any given time).
This isn't for everyone, of course, but I don't see a reason to criticize this business model when it delivers a definite advantage: lower prices.
If you don't like the trade-off of a lower price at the cost of not being able to use as much CPU power or RAM as you like, you may choose a service with less or no over-subscription. But I don't see how that subtracts from what this service has to offer.
Besides with that TOS it should have been part of the headline.
I found them when I was looking for a provider that had Linux Arch images.
If you're outside of the United States, VPN on a tiny server is by far the cheapest way to access US-only content.