This seems ridiculously cheap compared to something with comparable RAM/CPU on EC2.
Is there a catch?
They actually DO need to be able to scale servers up and down based on demand-usage, time of day, growth patterns, etc.
For them, the extra EC2 cost is negated by being able to spin up an extra 400 instances in the evenings, and turn them back down after everyone goes to sleep.
Under the dedicated server model they'd ALWAYS need to have an extra 400 dedicateds to handle that peak load (and in fact would need to have far more to handle additional spikes, projected growth, etc.) Those dedicated servers would sit idle, costing them money most of the time.
This is the difference between 'web-scale' and your typical startup's usage pattern.
Cost wise, your best scenario is usually going to be dedicated or colo + EC2 or similar for overflow / peak.
If you do just dedicated hosting you need to leave enough spare capacity that you feel comfortable handling the spikes for whatever the worst case provisioning time your host has.
It's still cheaper than EC2.
But if you do dedicated + ability to spin up EC2 to take peaks, you can go much closer to the wire with your dedicated hardware, and increase the cost gap to EC2 massively. You don't need as much spare capacity to handle peaks any more. You don't need as much spare capacity to handle failover.
It's rare for it to pay off to spin up EC2 instances to handle inter-day "normal" load changes, though - most sites don't have differences that are pronounced enough over short enough intervals for it to be worth it. If you do the dedicated + EC2 model, your EC2 instances needs to be up no more than 6-8 hours or so on average per day before it becomes cheaper to buy more dedicated capacity.
Sure, but a minor decision is who prints your business stationery or what kind of toilet paper to buy.
What hardware/networking platform to build your service on is not a minor decision. It will fundamentally affect the architecture of your entire system, and getting the decision wrong will potentially be expensive-with-a-capital-Failed-Company.
I pay about $3 per month for a very low-end xen VPS. Sure, I'm not running anything at all resource intensive on it.. but if I was maybe I'd spring for a higher-end VPS for $6, or if I really wanted to get crazy I'd find a really solid one for about $20 a month.
At $140-$190 a month, you'd better be getting a bunch of high-end dedicated servers and fantastic support, or you are getting seriously ripped off.
Hell, given the very modest needs of most small startups, having to shell out even $20 a month is a ripoff.
A $20 VPS is only for pet projects. For a startup it's worth it to invest in a/some dedicated server(s). Having some redundant firepower of dedicated servers so that you don't have to worry about the upgrades for a while presents a sweet spot between saving cost and scalability.
However, I remember that hosting in the US is surprisingly expensive for some reason, and has been for many years. Don't know why.
Recently i fired up a Support Request around 5am and got a response 6am when their support team picks up work. during the day you usually don't have to wait longer than one hour for a support response.
They have awesome hardware for reasonable prices... i migrated everything i own from ec2 to one EX4 box with 8 cores and 16 gig of ram for less than i payed at EC2...
i7 + 24GB RAM + 2TB disc + 100Mbps for €50
(they geo-restrict their plans, and I think only the French get unlimited bandwidth; UK people have to survive on "just" 15TB/month, for example)
I have had a box with them on and off (more on than off) for the last couple of years. I'm not doing anything critical there, but have also had no issues. Everything seems rock solid (which it should be -- Wikipedia says that "The company has six datacentres housing more than 100,000 machines.") You can reinstall from the web control panel, they rent KVMs, not sure what else to add.
The major things people tend to bring up in forum postings are:
Kimsufi is the "non-professional" brand, and hardware support is slower than the OVH parent. Nothing ridiculous, but you aren't going to get a bad drive replaced within a couple of hours either.
OVH has "low quality bandwidth". Of course, people making this complaint never quantify exactly what they mean by this. I just tested on my "24G" box, and see 30-40MB/s to cachefly, 90MB/s to Leaseweb in Amsterdam, 75MB/s to Linode in London, 3-4MB/s to Linode in Newark and Atlanta, 12MB/s to SoftLayer in Dallas, 10MB/s to SingleHop in Chicago, 11MB/s to Joe's Data Centre. (Of course these are large test files.)
If you're interested, there's no setup fee and no minimum contract length (check this for yourself, obviously), so I'd say go for it.
5Tb as a limit really shouldn't be a problem for what I would use it for.
Then again for what I currently do 10Mbit would not be much of an issue either to be honest, though I'll ask about the procedure to reset this if I do set a sudden bandwidth spike one month.
There is VAT on top of that listed price, but even with that it is still a good deal if the kit and company are up to scratch. Last time I looked there was an extra cost for paying monthly, but that seems to have been removed.
> You can reinstall from the web control panel
That sounds interesting for a dedicated box. Are they using some for of SAN rather than each machine having its own local drives (so they can "reimage" your machine just by dropping its volume on the SAN array and creating a fresh one from template)? Or is this less instant (more like: make the request via the control panel and in X hours a passing support tech will plug a USB stick in and reboot to reinstall)?
You can customise the partition layout, but otherwise get a standard set of packages for whichever distribution. There's a choice of the usual Linux suspects in 32- and 64-bit, FreeBSD, Open Solaris and Windows Server 2008.
The process is fully automatic and a CentOS install takes about 20 minutes.
Either you use Linux all day every day for some time, or hire someone to consult with you. That's my opinion anyway from years of consulting with people :)
With who?
one thing to watch out for with dedicated hosting providers is that the networks are sometimes not so great. you may get a great deal on hardware and tons of bandwidth, but if it drops packets all the time, it's not worth it.
make sure they segment you off onto a vlan (see http://jcs.org/mitm for why), make sure they are well peered, make sure they actually have staff at or near the datacenter they're running things from, and make sure they have the ability to block certain traffic from reaching you if you need them to (see http://jcs.org/sip for why) so it doesn't count towards your bandwidth total.
If you go this route you can just choose to run virtualization yourself with VMware ESXI or kvm and libvirt
I'm also uneasy with people that put up a page like this which basically makes it seem (unless you read the first sentence carefully) that you are in the secure hands of a much larger organization.
http://www.incero.com/datacenter-network
http://www.linkedin.com/company/incero-llc
Thoughts? (Note all blog posts are by "Gordon") I have less issues of course if the idea is redundancy.
Info in that post seems to indicate more employees:
Real office location in Austin, TX
Experienced owners; over 8 years experience in the web industry, over 50 years combined business experience
I haven't even gotten my incero server yet (just ordered it Friday evening); but I am impressed by their offering - such as IPMI/KVM, which is often not offered by fly-by-night hosters (and not offered by EC2 in the form of out of band access).
I also happen to be in a unique situation of not needing 100% uptime for my startup - a few minutes of downtime to switch to a hot failover is a non-issue; so if Incero proves competent to me, I'll still give them serious consideration for production hosting (again, with proper hot-failover backup elsewhere); despite potentially being a one-man-show.
Difficult to say which is the lesser of two evils:
* Being the customer of a small business, having to potentially wait for response
* Being one insignificant customer among millions when there's a datacenter meltdown at AWS, with no phone number to call, and no neck to strangle until you get back up -- at the end of a long list of small customers.
I only casually mentioned Incero here, vs. Voxel in many other replies, as I believe voxel is the more appropriate choice for most startups.
I'm reminded of this every time I log in to my now 12-year-old PayPal account. I see "1 open case" in my "Dispute Resolution Center" -- it's the case disputing my prepayment for the servers to this host, which PayPal said they would leave open forever, and if the principals of that company ever opened a new PayPal account they would have to resolve those disputes with all their former customers first.
Ultimately better to have a person or two that will pick up their cell phone when they are in the mall on a Saturday or out for dinner but are quite willing to tackle the problem when they arrive home. Not because they have to but because they care about what they do and have a conscience.
colocation was expensive and the hardware problems were all mine. i was pretty much tied to my local datacenter because i didn't want to ship a server around (which would be at least a day of downtime). pricing can be hard to compare because of power/space/bandwidth. if the equipment i colocated didn't have IPMI support, it could sometimes take up to a half hour to have a datacenter tech be able to put a remote console online when there were problems. at the end of it, i had a bunch of servers that were worthless on the resale market due to their age.
VPSes were never a serious option for the reasons stated in this article. it's impossible to track down performance problems when a dozen other VPS customers on the same server are taxing the CPUs and disks. i do use one that i pay $10/month for just to run a network monitor for some off-network perspective. they can be useful for single-task servers that don't need a lot of processing power like dns servers.
with dedicated servers, though, you can signup on a website and within a few hours have a complete server with modern CPUs, disks, and lots of memory assembled, tested, and connected to the internet with a remote console waiting for an o/s installation. when hardware goes bad, the server provider has lots of spare parts waiting around to be swapped in for free. and the best part of all, when you're ready to upgrade or move to a different provider, you just cancel the account and let the provider worry about what to do with the old hardware. i have a handful of these on various providers costing between $140-$190 a month for something like a core i5 ~2ghz with 8gb of ram, 2 big sata drives, and 100mbit ethernet with more than enough transfer every month.