I've actually had a pretty good luck with my $5 (base price $6) VPS I found on lowendbox. It has 512MB of RAM
Part of it is expectations. I know I'm not paying anything for this box, so I don't expect 100.0% uptime and the ability to run a high traffic service of which I really care about the availability.
I just need a place that can run cron scripts, regardless of whether my computer is on/online. And run screen so that it can download something that takes a few hours. And I do have a webserver on there, host some very unimportant things on there (like static versions of showoff slides from presentations I've given ... I'm guessing they get hit by humans once a month, though I guess I could check the logs), and occasionally I'll scp a big file into the web directory for someone to download.
The uptime isn't great. Pingdom sends me emails about it being down basically every month, but usually it's like 30 minutes. Uptime hovers right around the 99.9% that they guarantee. Sometimes pingdom monthly reports say it's a little less. I haven't bothered to try to invoke the SLA to get my month's $5 back, or whatever. For what I use it for, it works just fine.
If I were running a real website that I was invested in having people visit, I'd go for Linode/Slicehost/AWS/etc for sure. But this is the cheapest way to fullfil my needs for this box.
While there is certainly a huge amount of fanboyism surrounding buyvm, most of it comes from the stock scarcity.
I personally run 3 vps' with them (first three plans) and the performance has always been far more than acceptable.
I run an adult tube site that has exploded in popularity due to a fun domain name and being one of the first to post a leaked sex tape for a latin american actress. It can reach around 200 to 300 thousand uniques per month. Never had a load issue.
Of course, your mileage may vary. I could have ended up in a lousy node. Many people outright abuse their service for portscanning, spam sending and what not. Also, many people won't do this intentionally; they get infected through ancient wordpress installations.
My experience has been worse with Burst. Their IO is acceptable but their network is dismal. Ping times from aroud the world tend to confirm this.
Notes to take home:
ALWAYS make backups! If a provider goes deadool on you all of the sudden, you should always have fresh backups in a remote site. Honestly, there's no excuse these days. You can always get your money back if you file a dispute on Paypal. I wasn't affected by the collossal Hostrail crash, but many people were.
Lastly, always choose Xen over Openvz.
Edit: Forgot to add something very important: Avoid Hurricane Electric (he.net) like the plague. They are an absolute fuck up of a datacenter.
To be fair, that applies to any hosting, cheap or expensive, virtual or dedicated. I've only once had to restore data due to the fault of the host, but I've lost track of the number of times that I've had to get things back because of a user error. :)
I agree. I wouldn't completely write off those smaller servers and I think it serves as a very good entry point for learning Linux and server administration. If you're just starting out with Node or Rails its great.
I would recommend someone a Low End Box over a free Amazon Micro Instance any day, just because the CPU they limit you at is just bad. You start Apache or Nginx on those things or do a big yum install/apt-get and the terminal slows to a crawl.
Running a Production server off any low end box is definitely risky depending on who you're hosting with, but that goes without saying. I'd second BuyVM and Xen, I had a VPS with nordic and my cluster just died one day and was unrecoverable. They gave me a free small instance for half a year, but I opted to just switch instead.
My other recommendation is definitely Rackspace. Their low end service is only around 10 dollars a month, their chat support service is pretty good, and its fast and easy to scale. Database transactions on the cloud are still shit however. :( Really want a dedicated server if you're a larger or badly optimized site.
I agree with you, but would point out that CmdrTaco hosts his new site (that got front page on HN a few times) on a Micro on AWS.
He does a serious amount of caching (obviously) and offloaded all assets to S3, but it looks like it is still possible.
FWIW, I am still skeptical given the amount of complaints you see about the Micros on the EC2 forums regularly. He could have just gotten lucky I suppose (as far as his neighbors on the host).
Yeah, I'd believe it. In full disclosure I run my personal blog off an Amazon Micro instance. Its a Fedora instance running passenger, nginx, and rails. In production mode with assets cached it absolutely serves requests at a decent speed. Absolutely painful to run an integration test or development mode though. Probably just my fault for using Fedora and not the default "Amazon Linux". However, I setup a client with a small instance running RedHat and never experienced any latency.
Appreciate the heads up on Burst's network. I noticed their bandwidth for big chunks on dedicated machines is ridiculously cheap (40TB for like $100 [1]) but if the routes are inconsistent that doesn't help me a whole lot.
Also noticed while speed-testing the different nodes the CA connection (to Tucson, AZ) is phenomenally fast, but to PA about half the speed and to EU it drops quite a bit again.
When I do the speed tests for Linode using similar locations they all perform fairly well, same goes for AWS.
I must clarify that my experience was related to their regular VPS offerings.
They do offer a premium VPS (notice how its product descriptions mention "1GBPS port" and "Less than 50% load of budget VPS nodes", in contrast to just "XXX GB/MONTH" for the regular).
I didn't try their dedicated servers. They may have fatter pipes. I just had very poor latency, consistent with all their locations, being Florida the most stable for me.
If you do any sort of data crunching, you can get a good value for your buck. Certainly cheaper than AWS.
For my taste, Linode's latency has been pleasantly consistent.
95% of the time, it equals 142ms from Atlanta to South America (so far, the best I've found). I just wish they had a smaller offering.
I'll go back to watching the presidential election, have a nice day!
Well the VPS's are running on colo'd racks (and thus you're probably 2 or 3 steps of escalation further away from the DC operator).
Hurricane Electric in Fremont has had issues with power and previously with wide-scale IP block blacklisting due to widespread illegality (bot nets, etc) running on equipment in their Datacenter.
It might be fine now, I'm not sure, but there's a ton of DC's out there where this has NEVER been a problem, so I fail to see why I'd want to stick my assets there.
VPS hosting is the new shared hosting; so many applications and platforms exist that it's easy for anyone to get started operating a VPS hosting company. I could rent a dedicated server now for $200, install SolusVM and be a fully functional VPS host within a couple of hours and the majority of "budget" vps hosts are these. They don't have the funds to build a real business (like Linode), they just install some widely used software and off they go. They are limited by the software they use (whereas a company like Linode is limited by its engineers) so when things start to go wrong their response is to penalise the customers not make the engineers work harder and achieve more.
That said we run kvm on every box and run all customer and own apps in different VM managed (all managed by chef). So we're kind of running our own inhouse VPS-hosting-service.
We can achieve ultra-low-per costs per VM (as regular VPS providers would charce) but in the position to be the "root" customer of the box. We can decide which virtualization stack we want to use etc.
Does anyone have a good recommendation for a host with a nicely spread out global presence?
I am aware of the big players (AWS, Linode, Softlayer, RimuHosting) but they all seem to have shortcomings that don't make anyone particularly ideal and I'd love to avoid a combination of all if I could.
I am looking to minimize network latency and maximize disk space with memory and CPU being secondary concerns. More or less think CDN edge location requirements.
For example:
AWS - Speed to China and AUS/NZ from the Singapore region
is subpar. Speed to Japan from Singapore is much worse
compared to the Tokyo region directly. Managing many
images in many regions across the AWS stack can be time
consuming.
Linode - Their CA datacenter (hosted by HE) has a terrible
reputation and uptime. I love that they normalize their
bandwidth pricing across all data centers though. Disk
space is an issue for us (media distribution).
Softlayer - Looks solid, nothing for the AUS crowd or
Japan/Chinese users.
Voxel - Not cheap, again nothing for AUS/NZ crowd.
RimuHosting - Doesn't buy enough in EU/AUS to offer budget
bandwidth pricing, the London bandwidth is horrifically
expensive.
JoyentCloud - Single data center at this time.
These are all nit-picks, each of these services are great and I've hosted with them at different times and been very happy (other services like SliceHost and VPS.NET have been bad/very-bad - respectively) but I am curious if I am missing a single host that could meet most of these requirements that I just don't know about yet?
For example, I had never heard of BurstNET or BHost or A Small Orange until reading this thread... so I'm hoping I might be missing some fantastic host somewhere with medium/large VPSs or small/medium dedicated solutions.
Since you seem concerned (as I am) about AUS/NZ performance, it's worth repeating the unconfirmed gossip that Amazon have a closed tender out for a pair of Sydney data centers.
Ninefold are also in Sydney and doing a good marketing job, and I know several first-class engineers that have been involved in their setup and have made only positive remarks on their potential (with the usual caveats around Australian market scale being an issue).
I'm cautiously hopeful about Ninefold, but keep in mind they had some serious (multiple day) downtime a month or so back. I'm sure they'll get on top of their operations soon enough, hopefully before Amazon becomes the local gorilla-in-the-corner in the local market...
I wouldn't class Voxel.net as budget, but their low VPS is affordable, and based on my own experiences very reliable. They're very pro-active at alerting you to issues, even without their managed service.
The ultimate budget VPS is an Amazon instance. You can be relatively sure that it is not going to go under. You can run one free micro instance (http://aws.amazon.com/free/). You can scale as needed easily, both in terms of file system and CPU by simply rebooting your EBS to another instance type. You can have reliable, incremental snapshots. And finally, you have a great ecosystem of vendors providing AMIs for nearly all OS versions and popular web apps (like my own, http://bitnami.org/stacks)
Amazon isn't really budget. The free tier is nice, and the tiny instances are priced ok, but beyond that they're quite expensive for a VPs. The nice thing with AWS is the scaling and the supporting infrastructure, but you do pay for that compared to more basic VPS offerings.
Just a reminder to everyone that prices out the flat On-Demand rates of Amazon, a 3-year reserved instance price is 48% cheaper than OnDemand.
So if you know what your deployment needs are, take the On-Demand price and roughly divide it in half for your real cost-per-month (with a majority paid up front).
If you don't know what kind of deployment you need, then yes, running everything on-demand can be pricey compared to alternatives.
In addition to reserved instances, there are a couple of other ways you can achieve cost-savings for common situations. If you need to run background tasks (like video encoding, reports, etc) you can try running spot instances at a fraction of the cost of an on-demand instance. You run the risk that the instance is terminated, but you can always spawn a new one without affecting the main service, since basically you do not care if the task finishes a couple of minutes later. Another one is scaling up and down base on demand. If you run, say, a SugarCRM instance for your sales team, you can have it run in a Large instance during the week, 9-5pm, when it is going to be in full use by multiple people. Then, the rest of the time you can scale down to a small or micro instance. It is still going to be available, but at a smaller cost.
Well, if we are talking about bigger instances, then it is not "budget VPS" anymore either :) Still, depending on your needs (and specially if you combine with reserved instances, which can drive the cost 50%) it can be quite cost-effective
Yeah, but most people pushing more than a few tens of GB/month probably don't really need to save a few dollars over e.g. Linode.
Budget VPSs are the modern "shell account", their primary users are a combination of learners, early-stage developers, small-scale control-freaks, people who don't know any better, and of course the odd criminal enterprise.
Some of these will never grow much, the ones that do should normally have the cash for better accommodations and/or more EC2 bandwidth, and the criminals I don't really care about.
Or go one step further (depending on the design of your site) and switch to just S3 and CloudFront. I'm paying about $0.50/mo to host my sites since I converted to using static files.
Interesting that the author had trouble with BuyVM - my instance seems to be working out pretty well and i've recommended them to some friends. $15/year is incredible, i've had no trouble with support, and they just quietly sit there hosting rtorrent and nginx.
There has of course been some limited downtime.
EDIT: Uptime statistics hit 99.99% this month. http://buyvmstatus.com/info/34 On top of that, the Frantech Announcements feed is pretty good at explaining the state of their datacenter, maintenance, node migrations etc
Can anyone else comment on experiences with Linode (and equivalent alternatives)? I'm considering moving a group of servers off expensive dedicated hosting ($400/server) to a VPS service. I've looked at Amazon EC2 but they don't seem to have a sweet spot for my needs.
Are you sure that's the right move to make? Do you no longer need the resources of even a single dedicated server? At even 2-4GB RAM requirement, you would not be saving money buying a VPS over renting a server.
A Linode 4096, where you're sharing CPU and disk IO with all the other users on the physical server, is $159.95/mo.
For $159/mo you could rent a Xeon 3230 quad core, 4GB RAM, 2x250GB HD server from Softlayer with more than 4x the bandwidth allotment and a better data center with a larger support staff.
We deliver video content. We have a CDN that properly encodes and serves our content. All I need is to handle the HTTP requests for the "directory" of videos. A shared environment would probably be fine. All the 4GB servers seem to have most of their RAM dedicated to caching when I check what free -m gives me. Our physical disk space requirements are low. As I shift more of our content to the CDN, our bandwidth needs are decreasing as well (from 12TB/mo to 7TB/mo since I started optimizing).
I inherited this setup, so to me it looks like the servers are overpowered. Also, the customer service is terrible, which is the primary reason for wanting to leave them. I want a service where I can easily requisition and kit out new servers quickly in an "a la carte" fashion.
I have 3 servers running on EC2 already using micro instances. They're mostly to handle minor services (email list subscription/requeuing for all the web properties, payment processing postback handling, etc.) They've been up for nearly a year without any problem.
I've been leaning toward EC2, but I wanted to hear what else is out there.
All I need is to handle the HTTP requests for the "directory" of videos.
Have you considered just using Amazon S3, especially now they offer full 'httpd'-like functionality (no more bucket XML listings if you try to visit the root of the domain).
If you are running a CDN in front to stream the content, this sounds like the best solution to me. You could probably get away with Reduced Redundancy Storage too if you have backups.
I've run all my personal services (email, web sites, screen sessions etc.) off two Bytemark[1] virtual machines with 500MB of RAM each and the only problems I've experienced have been one or two 'high up' problems such as two core routers going down at the same time - i.e. nothing to do with contention. If you need more RAM, you just pay the extra monthly/annual fee for the next plan up and reboot the machine.
I pay £15/month per VM, which is probably $15-30/month each depending on where the exchange rate is.
I've been considering getting off of Google for my email hosting, but I remember there being a number of issues with virtual hosts getting their entire IP blocks blacklisted. Is this an issue anymore?
I've never had a problem, and I suspect it's more likely to be shared hosting which ends up getting blacklisted due to people installing insecure scripts.
If you use a half-decent provider who quickly investigates any unusual traffic on their network then I can't see it being a problem.
It depends on your application. Every provider has their own stack running in the back, their own policies, and a unique set of existing users. This translates to unique environments that are more or less suitable for certain needs.
My experience with EC2 is that you get more RAM for the buck when compared to Rackspace and others, but IO to disk and CPU is sub-par. As a result, I tend to prefer Rackspace who are also big players in the open source space.
Keep in mind, regardless of any advertising, there is no such thing as guaranteed performance in the cloud. Ever. Every provider oversubscribes ... they do not expect and cannot handle 100% utilization by their subscriber base. Behind the scene, a VPS is sharing all of its resources, always. If your provider does its job right, you'll never experience a drop in performance. Everything will be consistent. But there is no way for that provider to guarantee that. They can experience an unexpected jump in resource usage at any time.
With dedicated hosting, that's not a possibility. It's your hardware. You're the only one using it.
>> My experience with EC2 is that you get more RAM for the buck when compared to Rackspace and others, but IO to disk and CPU is sub-par. As a result, I tend to prefer Rackspace who are also big players in the open source space.
Seconding this. And I'd recommend a hybrid solution if it's possible. Dedicated db server with some cloud webheads for your front end. I really think that's the best bang for your buck. Has anyone tried the Amazon RDS or the RackSpace R2 servers? I'd really like to know if they provide any advantages over just straight up hosting your database on another cloud instance.
The ultimate hybrid is to find a budget VPS (or premium VPS) provider who is based in the same datacenter as your 'bare metal' servers. Granted VPS != cloud, but depending on your requirements you might get the same benefits.
The benefit here is that you can exchange data between servers/VPS's via the private LAN - low latency and potentially no bandwidth costs.
If you go with an Amazon based solution then you don't have this benefit due to the NAT-like way Amazon works and the fact they either own their own datacenters or at least don't allow any private routes between other servers in the same DC.
(I don't know whether every Amazon availability zone is in a wholly operated Amazon datacenter)
I'm very happy with them, but I cannot confirm with a 100% certainty that the experience will be the same for you.
May I ask what you are running in those servers?
Are your services bound to CPU, bandwidth or memory? All three?
I've had good times with 100tb.com, but the price is prohibitive for me (The dollar is not my "native" currency and I have to work my day job very hard to reach the 200usd it costs)
Currently they're bandwidth bound, but I'm dissolving that. In the next year I imagine that it'll move to being more memory bound. Then that will turn into being bound by I/O, but that can be solved with memory and more clever caching schemes. CPU should never be an issue unless I move the video encoding back in house, but I don't see that happening soon, and I would separate that into a its own server farm.
I've found Linode fantastic: fast connection, reliable service, reasonable disk speed (for a VPS), responsive support (rarely needed), decent control panel, and continually improving value. My servers are mainly in their London data centre.
Anyone care to follow this up with a brief article on the more expensive VPS providers? (such as A Small Orange) Are those of us on Linode, Rackspace Cloud, and Slicehost missing out at all?
Well as someone who has used quite a few VPS servers over the years from a plethora of providers I must say at this point in time Linode is probably the best bang for your buck with all the utilities and uptime it provides. The only thing about Linode is you must not order from the fremont datacenter and you have to subscribe to the status updates to realize what's going on.
The only thing that a more expensive provider will give you that linode does not is server management which can be useful but I don't really mind sysadmin at this point in time so I can't really review that.
+1 to linode - in the past I opted for cheap options, and suffered, but after experiencing the management tools and infrastructure linode has in place, I'll stick with them at the more premium price.
I ran with RootBSD.net for a year or so, and I thought they were awesome. Sharp guys and good support. I even migrated a client's site over to them, saving my client a bundle over unmanaged bare-metal hosting and earning myself some credit for the referral. At the time, I think $20/month was their smallest package, though they may have a $10/month now.
I've had a few Linux VPSs under $3/month (posted specials on lowendbox.com). I agree with the article: don't expect the world for $2/month, back up anything of value frequently, and enjoy the too-good-to-be-true ride while it lasts. The performance is hit-or-miss, you'll find your VPS rebooted more often than you'd like, but for dinking around with an idea, they're perfect.
I'd even be comfortable using these cheap services for backup mail relays or DNS servers. Just don't forget the backups.
Most seem to support PayPal subscription payments, which makes them easy to set up and forget.
Another +1 for Linode. I especially enjoy their DNS management which is included in the price (many providers don't offer/charge more for). They use each of their datacenter presences to redundantly host the DNS zones, so my domains have 5 name servers to fall back on across USA + Europe.
Shh, don't tell them, but given the'll host unlimited domain zone files I even use it to manage a few extra domains that have no zone entries pointing to a linode VPS. They don't seem to check/mind.
After lots of downtime across three of my VPS' at VPS.net i'm opting to migrate all of them to one box on the Rackspace Cloud. I'm not a server admin, nor do i want to be, so the attraction of the new 'Managed Cloud' is all that much stronger.
I have used Voxel.net in the past, and their technical support team are absolutely amazing. Always willing to go that extra mile. Uptime was also very good.
I had a very poor experience on the past with a cheap hosting provider. Actually it really does not matter if it is shared hosting or a VPS -- at the end of the day you get the service you paid for. Currently we use Rackspacecloud with very good results. Their control panel has a few glitches but my support experience was out of this world.
SolarVPS has servers in New Jersey and Los Angeles for $5-40 that I've been using for a years. Like Rackspace (4 cores), these small boxes also have up to 16 cores. Great for mysql replication, code repos, distributed data nodes.
I have been using a 15$ vps from www.tektonic.net since 2004 and have been happy. I run a few personal sites and sometimes test code. They have been down 1 to 2 times a year which IMHO is fine for the price.
Heh, the performance comparison linked to in that article is kinda worthless since it leaves out RapidXen, one of the largest bare bones VPS providers.
Part of it is expectations. I know I'm not paying anything for this box, so I don't expect 100.0% uptime and the ability to run a high traffic service of which I really care about the availability.
I just need a place that can run cron scripts, regardless of whether my computer is on/online. And run screen so that it can download something that takes a few hours. And I do have a webserver on there, host some very unimportant things on there (like static versions of showoff slides from presentations I've given ... I'm guessing they get hit by humans once a month, though I guess I could check the logs), and occasionally I'll scp a big file into the web directory for someone to download.
The uptime isn't great. Pingdom sends me emails about it being down basically every month, but usually it's like 30 minutes. Uptime hovers right around the 99.9% that they guarantee. Sometimes pingdom monthly reports say it's a little less. I haven't bothered to try to invoke the SLA to get my month's $5 back, or whatever. For what I use it for, it works just fine.
If I were running a real website that I was invested in having people visit, I'd go for Linode/Slicehost/AWS/etc for sure. But this is the cheapest way to fullfil my needs for this box.