Seems to be a pretty good offer, even in comparison to digital-ocean (https://www.digitalocean.com/).
Has anyone hosted anything with them before? I'd like to know whether they're dependable.
Lol, I migrated from Linode to DigitalOcean one month ago.
The pricing of UbiquityServers seems very tempting, however I will calm down this time :)
Anyway, has somebody had a chance to try both DO & US?
I had a Linode 512 and switched to a 2x512 Dropplet setup and put my Database on one 512 and my Rails app server on the other.
My stuff runs as quickly if not quicker and I spend $10/mo less. In addition to this I also have a clear scaling path (e.g. If my DB grows I can up it to a 1gb. If my traffic grows I can either move the app server to a 1gb or maybe get a few more 512's and setup a load balancer.)
They have been. All servers are now 8 core and they cranked the bandwidth allowance up by about 10x if I remember. Apparently they are meant to be increasing RAM soon but I'm not sure how reliable that information is. However, as they've done everything else it does make sense.
RAM is the only thing that I feel Linode is lacking as of late.
My one complaint with DigitalOcean is that I had a dev server setup and I was snapshoting it and destroying it when not in use then recreating it from the snapshot when I needed it. A few weeks ago they had a failure of some sort and my snapshot was lost.
It took me a fairly long time to setup the dev environment so I was a bit pissed.
Have your setup for your dev environment under version control.
I do this with Linode stackscripts, and I couldn't live without it. Unfortunately DigitalOcean don't support stackscripts as of yet. And their support didn't want to tell me if they will ever support it.
I was dubious about switching back to DigitalOcean as I was worried about how cheap they were offering packages. I bought a DigitalOcean VPS that had slightly better specs than my Linode and ran some tests. Just a few like response time on the site they were both holding, I/O and a Redis stress test and Linode seemed to come out marginally on top.
I was an extremely satisfied Linode customer for several years. A few months ago I migrated to Digital Ocean to save money and for the SSD performance. I'm happy that I made the move. I get better performance for less money. Also, I have spun up and down a lot of different servers for testing purposes. It's fantastic to pay by the hour for servers so that testing a beefy server for a day or two costs very little.
If you want to compare to digital ocean they are having a coupon code on Twitter - Includes 512MB RAM, 20GB SSD Disk, & 1TB Transfer for $5/mo.! Use promo code "SSDTWEET" for a $10 credit.
Here's the #1 question I have for any of these hosts, and I never see it answered: how stable is network throughput?
I run what amounts to a fancy real-time chat service. Are my users guaranteed a slice of the pipe? Do you have the resources to make sure they won't lag horribly if my neighbor is getting DDoSed?
Keep in mind that there is a difference between "we actively make sure you'll have a good experience" and "you happened to get provisioned on a machine with good neighbors", even though the anecdotes will be identical.
Bingo. Thoughput of all kinds is the number one gotcha of low end budget hosts. Disk, network, etc. I work with tons of these kinds of hosts for clients and I'm just not at all surprised anymore when everything goes to shit a month after I sign up and suddenly the VPS gets transferred off that special "new customers" node for any number of bullshit reasons.
Bottom line: You get what you pay for. End of story.
Disclaimer: I work for UpCloud, also a cloud hosting company.
This is a great question. Something I've been thinking about a lot is also the benchmarking of redundancy and the protection of one's data.
It's easy to benchmark performance, but as Pirelli stated in their advertising: speed is nothing without control. I feel the same about online hosting providers.
Naturally I have no knowledge of US redundancy solutions, but it's one of the most critical yet toughest things to compare when choosing a new hosting provider.
Why does every host offer a plan that scales at the same rate in every column (CPU, RAM, Disk space, bytes/month)?
I can think of almost no application which requires lots of all of these. Why isn't there a host which allows you to scale your own requirements? Is it a technical thing, or just another form of "oversell and underprovision because most people won't use their plan"?
it's far easier to standardize (xsmall, small, medium, large, xlarge) than to allow for custom requirements for the purposes of communication. Most people don't necessarily know exactly how much they'll need, so its easy to just say "large".
Plus, in a virtual setup, optimizing allocations with custom requirements is a far trickier algorithmic problem.
It makes things easier for the provider, build 1000 identical boxes then sell by the slice. It means they don't have to build odd boxes to deal with odd demand, or try to pair up high memory use low compute use with a low memory use high compute use client nodes.
At that "4 core" (virtual) pricing I'd take dedicated from OVH instead [1]
The $6 pricepoint is likely to just get you in the door for an upsell.
Why would you subject yourself to unknown neighbors with unknown abusive habits on the server resources when you can just get a dedicated with SSD for less money and co-host your own projects with known behaviors (and without the overhead of a vps hypervisor).
You can also get 2 free servers a month from Amazon.com (1 Linux one windows). It's actually 750 hours each so could use more than one at a time if you turned them off afterword. http://aws.amazon.com/free/
Note: These free tiers are only available to new AWS customers, and are available for 12 months following your AWS sign-up date
Please don't present this as equivalent. AWS's free tier, amortized over the life of a five-year web service, isn't free, or even cheap.
Now, I did make pretty good use of AWS myself for that first year... and then it ran out, and suddenly AWS was much more expensive than Linode (and more recently, DigitalOcean.)
When all you want is "a thing to keep your Rails app running for its measly 10K hits a day" AWS loses hands-down. It's just not their use-case.
At 2c an hour * 365.2425 days * 24 hours a day /12 months * 4/5 = 11.70$ a month. But, if you want an always on connection for 1+ years you can get discounts dropping down to 8$ a month and if you link it to a high volume account you can get another 20% discount.
Where does the first 20% discount come from? (And I would completely disregard the second one; if you know someone who has a high volume account, you can probably get some spare slices just with a polite word.)
More importantly, though: you're assuming Micro instances. Micro instances are tolerable, but not really comparable to anything offered to paying customers at even the most downmarket VPS. The moment you want to burst the CPU of a Micro for any extended period (say, starting up Unicorn workers), your instance will become a chugging morass.
When I was on AWS, I had to go in and reboot my instance every other time I deployed an update to it, because the update process ran just long enough to prompt a 99%-CPU-throttle. At that point my server was basically wedged, since new HTTP requests were still trickling into the old app servers and queuing up (with the throttling, it now was too slow to handle them before more arrived), leaving the CPU unable to "relax" enough to get out of the throttled state even after the deploy was over.
> if you want an always on connection for 1+ years you can get discounts
With a down payment. You have to calculate the NPV of the cashflow, not just the aggregate sum; $10/mo is a lot easier to manage than $100 + $8/mo.
But disregaring that: even with a three-year term--costing $257 up-front, and committing you to payments in a similar way to a cell-phone contract--it looks like an Small will still cost $15.80/mo. Why put yourself in such a bind just for a not-that-good slice? (Network throughput is one of the few acceptable answers to that.)
This does all assume that your app that will live comfortably on any consumer workstation machine for the foreseeable future (its growth has already plateaued, in other words.) If there's any possibility of hockey-stick growth in your app's future, AWS becomes a lot more attractive.
You said "amortized over the life of a five-year web service" that's 4 years of cost + 1 year free.
As to Cost it's $100 for 3 years at .5cents an hour or 63$ for 1 year at .5cents an hour. I used the 63$ for 4 years vs 3 for 100 plus one at 63. If you really care about the time value of money your first year is free so it's better than than paying up front for that first year any way you do the numbers.
PS: I don't actually use AWS I was just defending them from the old 'there vary expensive' which I found to generally be false. Most of the time your time is worth far more than hosting costs especially when dealing with such small servers.
Co-lo for a single server is not practical if you are relying on the datacenter techs for repair/replacement. When it's their own hardware and the have hundreds (or thousands) of identical systems, much easier/cheaper.
> Ubiquity offers a 100% uptime SLA on our entire Cloud infrastructure
> In the unlikely event that one of the solid state hard disks fails on the Ubiquity Cloud there will be no data loss and no impact on the cloud instance's performance
First one is not necessarily a red flag. It doesn't mean that you get 100% uptime. It just means you get paid back if you don't get it. It's an service-level agreement, not a guarantee. (it also doesn't necessarily mean that uptime == your machine is responding... read the document carefully)
The second can be interpreted two ways... Do we start at any state or a perfect state? If you start will all good drives and one fails, it shouldn't affect anything assuming they're using raid. With raid mirroring there shouldn't be an impact on the performance either. However if you start with some number of disks already failing, of course there will be data loss. They could probably say that they're doing N+X redundancy instead to be clear.
I'm now confused about what "Cloud Hosting" is. I thought it was a "virtual" dedicated server. All that ubiquity provides is a control panel:
> Our managed cloud servers come with cPanel/WHM access. No root access is provided.
How is this different from regular hosting you can get elsewhere far cheaper? Is it just the promises of RAM and SSD?
59 comments
[ 0.23 ms ] story [ 123 ms ] threadI guess it depends on what you want for your fiver (5-6$). Ubiquity's deal has twice the RAM.
I'm suspicious regarding dependability, which I assume one should be in this price range irregardless.
Also I lost all my data and backups.
I had a Linode 512 and switched to a 2x512 Dropplet setup and put my Database on one 512 and my Rails app server on the other.
My stuff runs as quickly if not quicker and I spend $10/mo less. In addition to this I also have a clear scaling path (e.g. If my DB grows I can up it to a 1gb. If my traffic grows I can either move the app server to a 1gb or maybe get a few more 512's and setup a load balancer.)
RAM is the only thing that I feel Linode is lacking as of late.
It took me a fairly long time to setup the dev environment so I was a bit pissed.
I do this with Linode stackscripts, and I couldn't live without it. Unfortunately DigitalOcean don't support stackscripts as of yet. And their support didn't want to tell me if they will ever support it.
I was dubious about switching back to DigitalOcean as I was worried about how cheap they were offering packages. I bought a DigitalOcean VPS that had slightly better specs than my Linode and ran some tests. Just a few like response time on the site they were both holding, I/O and a Redis stress test and Linode seemed to come out marginally on top.
If you want to compare to digital ocean they are having a coupon code on Twitter - Includes 512MB RAM, 20GB SSD Disk, & 1TB Transfer for $5/mo.! Use promo code "SSDTWEET" for a $10 credit.
I run what amounts to a fancy real-time chat service. Are my users guaranteed a slice of the pipe? Do you have the resources to make sure they won't lag horribly if my neighbor is getting DDoSed?
Keep in mind that there is a difference between "we actively make sure you'll have a good experience" and "you happened to get provisioned on a machine with good neighbors", even though the anecdotes will be identical.
https://www.ubiquityservers.com/ddos-protection
Bottom line: You get what you pay for. End of story.
This is a great question. Something I've been thinking about a lot is also the benchmarking of redundancy and the protection of one's data.
It's easy to benchmark performance, but as Pirelli stated in their advertising: speed is nothing without control. I feel the same about online hosting providers.
Naturally I have no knowledge of US redundancy solutions, but it's one of the most critical yet toughest things to compare when choosing a new hosting provider.
Best of luck to US though!
I can think of almost no application which requires lots of all of these. Why isn't there a host which allows you to scale your own requirements? Is it a technical thing, or just another form of "oversell and underprovision because most people won't use their plan"?
Plus, in a virtual setup, optimizing allocations with custom requirements is a far trickier algorithmic problem.
So I wouldn't call it a technical thing, but more of a strategic decision in how you want to position your company + offering.
Is that billed by usage, or billed by throughput? I'd love to get a cheap rate on something like "guaranteed 10Gbps port; 100GB monthly transfer cap."
You have no idea how many neighbors you have on that vps.
And that "vcpu" is likely a hyperthreaded core, not physical, so it's a "half core" in reality.
At that "4 core" (virtual) pricing I'd take dedicated from OVH instead [1]The $6 pricepoint is likely to just get you in the door for an upsell.
Why would you subject yourself to unknown neighbors with unknown abusive habits on the server resources when you can just get a dedicated with SSD for less money and co-host your own projects with known behaviors (and without the overhead of a vps hypervisor).
[1] http://www.ovh.com/us/dedicated-servers/sp_32g_ssd.xml
Note: These free tiers are only available to new AWS customers, and are available for 12 months following your AWS sign-up date
Now, I did make pretty good use of AWS myself for that first year... and then it ran out, and suddenly AWS was much more expensive than Linode (and more recently, DigitalOcean.)
When all you want is "a thing to keep your Rails app running for its measly 10K hits a day" AWS loses hands-down. It's just not their use-case.
Which sounds cheap to me.
More importantly, though: you're assuming Micro instances. Micro instances are tolerable, but not really comparable to anything offered to paying customers at even the most downmarket VPS. The moment you want to burst the CPU of a Micro for any extended period (say, starting up Unicorn workers), your instance will become a chugging morass.
When I was on AWS, I had to go in and reboot my instance every other time I deployed an update to it, because the update process ran just long enough to prompt a 99%-CPU-throttle. At that point my server was basically wedged, since new HTTP requests were still trickling into the old app servers and queuing up (with the throttling, it now was too slow to handle them before more arrived), leaving the CPU unable to "relax" enough to get out of the throttled state even after the deploy was over.
> if you want an always on connection for 1+ years you can get discounts
With a down payment. You have to calculate the NPV of the cashflow, not just the aggregate sum; $10/mo is a lot easier to manage than $100 + $8/mo.
But disregaring that: even with a three-year term--costing $257 up-front, and committing you to payments in a similar way to a cell-phone contract--it looks like an Small will still cost $15.80/mo. Why put yourself in such a bind just for a not-that-good slice? (Network throughput is one of the few acceptable answers to that.)
This does all assume that your app that will live comfortably on any consumer workstation machine for the foreseeable future (its growth has already plateaued, in other words.) If there's any possibility of hockey-stick growth in your app's future, AWS becomes a lot more attractive.
As to Cost it's $100 for 3 years at .5cents an hour or 63$ for 1 year at .5cents an hour. I used the 63$ for 4 years vs 3 for 100 plus one at 63. If you really care about the time value of money your first year is free so it's better than than paying up front for that first year any way you do the numbers.
PS: I don't actually use AWS I was just defending them from the old 'there vary expensive' which I found to generally be false. Most of the time your time is worth far more than hosting costs especially when dealing with such small servers.
> Ubiquity offers a 100% uptime SLA on our entire Cloud infrastructure
> In the unlikely event that one of the solid state hard disks fails on the Ubiquity Cloud there will be no data loss and no impact on the cloud instance's performance
The second can be interpreted two ways... Do we start at any state or a perfect state? If you start will all good drives and one fails, it shouldn't affect anything assuming they're using raid. With raid mirroring there shouldn't be an impact on the performance either. However if you start with some number of disks already failing, of course there will be data loss. They could probably say that they're doing N+X redundancy instead to be clear.