I switched from Webfaction to Linode about six months ago, because the WF server I was on kept experiencing hard drive failures leading to an unacceptable amount of downtime. Once is tolerable, but they should have replaced the drive then. They didn't, and two months later it failed again. Again they tried to repair the filesystem. It failed two days later again. When I challenged them on this, they admitted they knew the drive was bad, and that it had no redundancy (seriously?!!!), but blamed the guys managing the hardware (which is evidently NOT Webfaction). Not to mention a ridiculously slow control panel at random times.
Linode has been a dream-come-true. I pay only a few bucks more a month than I did for my Tier-4 plan at Webfaction. Plus, I'm back to managing my own server and no longer have to deal with all the Apache port-forwarding weirdness, local profile software installs, etc.
I've used other VPS hosting companies in the past, but Linode blows them all out of the water. Their VPS infrastructure is very well engineered, and fast. Nobody can touch their price/performance ratio. EDIT: BTW - I am excluding OpenVZ-based VPS's which come with a number of issues that I don't care to deal with.
Same here; Linode has one my heart for three very important reasons, that will mean a lot to anybody looking for small-scale hosting:
1. Proper provisioning. They aren't the cheapest option in the market, but they actually provide what you're paying for. I've never had CPU, disk, or memory contention issues on my Linode because they packed too many instances on a host.
2. Great peering. Their network guys know what they're doing, and it shows in the low latency I've seen on my server hosted there, even coming from Japan.
3. Great staff. I wanted to set up my DNS with a wildcard MX record for subdomains, so that I could route mail without having to manually add MX records for every new account. The RFC for DNS allows this, but the Linode DNS management interface didn't... until I contacted them and asked for it. Initially, they gave the 'sorry, but we don't support that' line, but after I pointed out the relevant bit in the RFC, the feature was added within two days. I have never had a hosting provider respond like that.
Interesting experience. I've got three or four sites on Webfaction. So far, ssh on a shared server plus the admin panel has been a sweet spot for me between configurability and not having to manage my own server. However, I've noticed an increase in downtime periods over the past few months and I'm starting to have concerns about the reliability of the service.
I don't know about nobody, but it seems very good. I downsized one of my machines from a dedicated server at Softlayer costing $234 a month to a ~$40 a month Linode and.. it's doing fine! (Though it's not like I was maxing out the dedi.)
I now have several Linodes, a couple in London, all the rest in Atlanta (which, by my research, seems to be the best US data center for them).
Disclaimer: Linode advertise on my Web site now, but I pay for my hosting and chose them independently - no deals :-)
This is a coincidence. I had a funny experience with Linode just today!
I´ve been in South America for a few months, and I found out that my bank card (visa) expired a little over a month ago.
"Hmm," I thought. "I sure hope that my linode is all right. But it´s been less than two months. They will turn it off, but that is all."
So I checked my email. Part of the joy of my style of travelling is not checking email for a month at a stretch.
They deleted my linode. It´s completely gone.
I feel a little sick ... it was just for personal use (for now), and I have all of the git repos on my local machine as well, but I´m not yet to the point where everything I do on the server is in a repo somewhere. Poof, hundreds of hours of server mangling, gone forever.
Literally, if you forget to check your email, they delete it in 20 days. If your bank card expires while you´re out of touch ... for some strange reason ... anyway, instead of warehousing your data for even a few months, they kill you.
I like linode. I just feel sick right now.
On the one hand, I´m an idiot.
On the other hand, 20 days?!?!?! When you know how precious a customers´data is in this context, and it´s purely a question of temporary storage, why?!?
I need a drink. Oh, yeah, Go Linode!
edit: okay, I´m over it now. Leaving the original, melodramatic writing because, darn it linode, someone from Slicehost posted that Slicehost did it differently, and my reaction is relevant to your business.
I'm sorry to hear you had a bad experience- I know how frustrating it can be to have unexpected loss, particularly here, where it feels like it could have been prevented.
I'm not associated with Linode; I don't even host anything there, but I can see things from their perspective on this.
With a service like this, you'll sometimes see people who never cancel a service, they just stop paying for it. Maybe their card expires and they decide not to renew, maybe their business closed, or any number of other things.
If I were in their position, I'd certainly reach out to you- I'd warn you it was ending soon, and I'd remind you to renew. For one, it'd be the decent thing to do, and for another I'd want to get the renewel revenue! As the expression goes, 'It's easier to keep a customer than gain a new one.'
That said, if the time came and you hadn't renewed, I'd have a decision to make.
I want to keep you as a customer, but from my perspective, I have no indication you want to stay as a client- You haven't sent me any email about it, and you're not paying me anymore.
I'd probably archive off the data for a while, and shut it down.
My mental thought process would be akin to "I'll turn it off, and keep the files around for a few weeks. If he's using the site, then he'll notice when it gets shut down, and pay to stay with us. If not, Good luck in the future!"
From my perspective, it looks like that's exactly what Linode did. They gave you 20 days; That's nearly 3 weeks of the server being off, without you even writing them an email to say "Hey, Noticed it's off and I'm not paying you, could you please give me a few more weeks?"
I understand it's very frustrating for you, but I really can't blame Linode at all.
The other thing I'd touch on briefly is that you really should have backups.
I don't want to harp on it, because I know it must feel like kicking you when you're down, but servers fail..
What if Linode's datacenter caught fire? What if they were bought out by a company who wanted to convert them all to Windows servers? ;) Who knows. Things happen. Having the service turned off for nonpayment is among the smallest of problems.
If you were running a nightly backup job at this point, you'd be very frustrated that you had to copy everything back, but you'd have not lost anything.
In any event, I'm sorry that things didn't work out, and hope that your experiences with Linode (Or whomever you use to replace them) are more positve going forward.
How long should they keep a backup of every user's virtual machines, after they stop paying? 20 days is reasonable, a month would have been better (If only since it's more predictable) but they really didn't do anything unreasonable.
Keep in mind, even if it's only $2/month (After redundant backus, etc), if you have had 10,000 customers, that really adds up! Keeping files forever "Just in case" isn't a sustainable business strategy.
How many of the customers stop paying? Probably not many. How much disk space are these customers using? Probably not much. How much disk space do they have sitting on their servers unused? Probably a whole lot.
I doubt it. Most people who would be good customers for them would realize the complaint was far fetched. They don't need customers who whine about things that far off the gamut.
People have suggested that 37signals would suffer from their attitudes towards implementing new features or crap like Gantt charts in BaseCamp, but I don't see them suffering for it, despite the ongoing rants and raves by "high maintenance" potential customers.
For me at least the problem is the attitude toward customer data. I would definitely try to be a good customer but there are all kinds of reasons why I might miss a payment. Even if the chance of that happening is 1% it would still be a reason to reconsider competitors even though Linodes specs are better.
I lost my credit card and got a new one, so the charges from slicehost were bouncing because they had my old number. They sent me several e-mails about it but I wasn't checking the e-mail account they were sending it to, so this went on for about 2-3 months past due. Eventually I logged onto the e-mail account and found out about it -- I hadn't even noticed anything was wrong because they kept my VM running. I just logged on, payed the $60 or whatever it was in back payments and that was it.
(I don't know whether this is their normal policy or not, so don't use this post as an excuse to try it.)
Heh, I stopped looking at this thread almost immediately after "getting over" the initial sting.
But I have to say, I still feel that this kind of behavior would be a sane policy.
No one´s saying you should hold onto data forever. But having a dedicated pool of dead storage that´s capable of storing the data for a few months of non-payment, or even longer -- that seems like a very good idea.
The running VMs are so many, many times more expensive than a block of dead storage of equivalent size. Probably orders of magnitude more expensive.
Then when you consider the business benefit to having the softer-and-fuzzier policy ... well, there´s always the "avoid 'high maintenance' customers" theory.
It's unfortunate, but I don't think linode was in the wrong here. probably more than half of my customers who cancel do so by not paying any more. If you want a 'free two months' at the end of your contract, well, they are going to have to start charging more up front.
Maybe another option would be to charge the customer an (optinal) up-front deposit and say "when we shut you down, we'll upload this to S3 with the following credentials" or something. That'd be some programming work, but it'd result in better outcomes. hmm.
Or implement the deletion policy like a LRU cache would. Have a fixed size storage for customers who haven't paid and just delete the image of the oldest customer in storage when the disk fills up. Send emails to the oldest 10 customers in that disk that if they don't respond, their data will be deleted.
For the record: The Planet will cut your server, wipe it, and reassign after 5 days of non-payment. That's standard operating procedure at every server host I've ever used. I'm honestly surprised that Linode gives you 20 days; that's very generous in my experience.
They didn't give him 20 days. He checked it after 20 days. Who knows the grace period for an expired card? I think that deleting the data is a bit harsh for a late payment. How many accounts really get to that point of late payment to where it would be costly for the company to just shut off access to the data until the plan is paid for. A customer oriented policy would be having a grace period of 3 months where the data is still there but they don't delete it. That's good customer service in the exceptional case of late payment/expired card.
Actually, I don't think it's unreasonable to expect a hosting company to store some data for free. Not in the sense that I'm entitled to it as a customer, but in the sense that it might be in their best interests: if the data is deleted, that's a pretty good guarantee that the client won't reactivate; if the data is saved and it's trivial to reactivate, some percentage of customers will do so.
It really depends on the number of people they get doing this. If there are loads of accounts that are just left to rot, then it might not be worth it doing this. If there are lots of cases like the original poster, then it might be.
They could have shut off the host, but kept the image file. I just can't see 16GB of storage space worth that much to a hosting company. It's likely that most of that space is empty space anyway. $110 for 2TB on newegg -- that would store 120 customers who may have accidentally have expired credit cards. It's not that expensive for Linode and the word-of-mouth value is probably worth so much more.
Maybe it would be worthwhile to distinguish between customers who had actively cancelled their accounts and those who had passively stopped paying? With the former, there is the expectation that you will not retain their data. As the originator of this thread shows, when customers' credit cards expire, there is sometimes the opposite expectation.
There´s an easy backup option now. A year ago, when I first signed up, the only things I could see were the options to use some of my storage space to make a backup image -- which you have to think of at the get-go, and make your image half or less the size of your storage.
I think all my monit and nginx stuff -- the per-project stuff -- is fine, as I keep in in the project repo. It´s just ... other stuff. Lots and lots of other stuff.
In addition to what wwortiz said, you do not have to create an image using less than half your allotted storage from the get go. So long as you can ultimately get it down to under half the size by deleting files, etc, later, you can downsize the image. In other words, get an 8gb image down to 4gb of actual data and you can resize the thing down to 4gb from the linode CP.
I'm not sure if you are more upset at the loss of the particular server setup or of the data. It sounds like the former. In that case take this as an opportunity to create an infrastructure for your servers where you can easily bring up new servers as you need them. Instead of logging into a new box and installing and configuring software, write a deployment script. That way you can migrate to/from Linode quickly.
My friendly neighborhood sysadmin says that Puppet is a good one. I've been using Fabric for some of these tasks. Also, just a shell script would do it depending on how sophisticated you want to get. I suggest getting a cheap VPS (search lowendbox.com, no affiliation) and practicing there. Or just set up a "fake" server using VirtualBox or some such.
A reasonable thing to do would be for them to call if your credit card expires to see if you accidentally let it slip without updating the info. If it was any other situation (e.g. the card was declined but still valid date-wise), I think how they handled it would be perfect.
And why not?? This would be the biggest reason I wouldn't go with Linode. Their Linode 512 instance only consumes 16GB of disk space, why couldn't they just compress the data and store it just in case the customer returns? 16GB is not a lot of money these days... a 1TB hard drive is around $60 and they could store 60 customers on one of these drives. The drive doesn't even need to be running for archival purposes.
That is very curious I've had a linode in dallas for a long time (2 or more years) and the only downtime that I have experienced on my server is from those few (I think it has been less than 10 times and they weren't more than a few hours) times everyones dallas linode went down.
I have one in Dallas, too. Nothing curious about it, take a look at http://linode.typepad.com/
Perhaps my requirements or expectations are higher than your own. The downtime hasn't been excessive, but sometimes, it is down for an hour before I see anything on their status update blog, and meanwhile, I'll have received a few notes from site users and advertisers.
There are three incidents listed in the past two weeks of packet loss, routing issues, and so on for 1-3 hours. There were other issues in May, I don't know where those are listed now.
Yes, it's only going down for a couple of hours at a time. Considering I'm using the server for images for a medium traffic website - an hour or two is VERY noticeable to my customers, and as noted, it adds up to more than the downtime on my Slicehost account (St. Louis).
Slicehost, much like Apple, doesn't want customers who primarily care about how much server space they are getting for the money.
Slicehost sells themselves on having superior support. (Now, I don't know if Slicehost support is /actually/ that much better than Linode, but if it is, the price difference could be a reasonable price to pay for the customers who need the most help.)
That's true. On the other hand, Linode is competitive on price and service. Every ticket I've submitted has gotten a response within a couple hours. That's pretty good considering how much I'm paying. Honestly, I don't know why someone would get a vps from a different company.
I'm happy you are happy, and am not trying to convert you to Slicehost. That said, a bit on my thought process: I've been with them for years, switching costs are non-trivial in terms of my time (and switching could break stuff), and I have absolute and total confidence in them to treat me well.
I just checked my email. In the last year, I've had three incidents with Slicehost, all connected to issues experienced when migrating VPSes. All were resolved in minutes.
Not to knock Linode, or Rackspace Cloud, or EC2, or whatever, but none of them have ever been in the foxhole with me at 3 AM when the server was down and got it back up before customers noticed.
I haven't checked for a while, but when I compared Slicehost and Linode pricing, they were about the same. I stayed on Slicehost for that reason. Slicehost has been great to me, and I like their IRC channel. I see no reason to switch. I guess I'll go compare again soon and if there's a big difference I'll be obliged to switch (very poor), but I haven't had any trouble with Slicehost.
Excellent comparison, and jibes with my own experience switching from Slicehost to Linode -- noticeably improved performance in CPU and IO. I've started recommending Linode, and now that they've upped the RAM, I will recommend them even more.
...despite Rackspace's well-publicized downtime, which has prompted prominent sites to move away from them?
One of my local competitors recently, proudly, informed me that they were about to launch a new hosting service supported by Rackspace. I smiled, and congratulated them, and started planning the marketing campaign I'd like to run when Rackspace goes down and takes my competitors' clients with them.
> Rackspace's downtime just makes more noise because lots of well-known sites are hosted there.
Not as many as there used to be.
I don't mean to be a jerk, but it's one thing when my site goes down because I did something dumb, and a whole 'nother thing when a site goes down, these days, because a data center is having a problem. I just don't understand why someone would assume that downtime was inevitable and not choose to minimize that risk as much as possible.
It's true that price isn't everything, and to some degree, web hosting is a funny market: it's really easy/visible to compete on price/features, but you only notice the service when things go awry (at least for someone like me who doesn't particularly need hand holding), and then you really want to be dealing with people who know their stuff and take their jobs seriously.
That said, I've never got the impression Linode was giving me anything less in terms of service than similar outfits.
Actually, HN did not switch over (at least, that comment didn't say it switched over). It says the main YC site WWW.ycombinator.com (and NOT news.ycombinator.com) switched, and that news.yc is serving just static content from slicehost.
One thing I personally like about Linode is that they continue to contribute code, sponsor conferences, and otherwise support the open source community. Of course there is a certain amount of self interest involved, but not every VPS hosting company gives back like this.
I first found out about Linode by participating in Rails Rumble, which they were providing hosting for, and switched to them from Slicehost after last year's competition. They offered 3mo free to any competitors who wanted to sign up. I sent in a support ticket and it was dealt with in literally 15 minutes. I've been extremely happy with them ever since.
Both Slicehost and Linode have top notch service. I switched because of price/ram and 32bit support. I found that the overhead on a 64bit machine on low ram instances is just not worth it for the kind of stuff I'm doing now.
We've done some performance and price comparisons of different providers including Linode and Slicehost (a.k.a. RackspaceCloud). In that price range there are a lot of options you could consider with cloud providers. Although cloud is generally (but not always) multi-tenant, many provide dedicated CPU cores so CPU performance is very predictable. One advantage to cloud also, is that oftentimes the OS runs of an external SAN so if the physical host machine does fail, it can be immediately migrated to another host with very little downtime. There are other benefits like backup images and instant upgrading.
As someone who sells VPSs for a living, well, I'd want a pretty significant discount to take a VPS over a dedicated server of the same specs. The big problem with VPSs is I/O. If you have a 1TB disk and you split it so that two servers are using it, each server, sure, gets 500gb, but each server gets /way less/ than 1/2 the I/O performance. When you have more than one server hitting the same disk, all your sequential transfers become random transfers, and random transfers, especially on the SATA that nearly all VPS providers use, suck.
That depends on what % of your cpu you're using, and how sensitive you are to disk bandwidth. Look at what fraction of the host machine you get ("How many Linodes share a host", http://www.linode.com/faq.cfm ), do you use more than that fraction of the resources that would be shared (disk I/O, CPU time)? How much is having a nice web interface for rebooting and such worth?
> Linodes of the same plan are grouped together onto a host. We adjust the number of slots on each host according to its resources and hardware specification. On average, a Linode 512 host has 40 Linodes on it. A Linode 768 host has on average 30. Linode 1024 host: 20 Linodes; Linode 1536 host: 15; Linode 2048 host: 10; Linode 4096: 5.
Does this mean that a 512 host has 40*512MB = 20GB of physical RAM? If it is less than that, then where does the 512 figure come from?
I believe the current Xen hypervisor uses about 128MB now, but yeah — I sincerely doubt it's out of strict necessity, but out of wanting to ensure not to overload each machine with too many nodes. They wouldn't bump down to 22GB because I assume their mobos are dual-channeled.
Or, they're fibbing a bit by using that "on average" language, and they actually have more like 42 or 44 nodes, but wanted the nice round numbers for the marketing copy.
Edit: Oh, I didn't check the username. I imagine you'd know better than I.
Depends on how process hungry your server is, if you constantly maximizing quad core xeon server, then its not a direct comparison in processing. But if you are using processing speed of a dual core, then I think you should move.
They might give you four "virtual cores" but its not a direct comparison to a xeon server. In everything else, memory space and transfer, you are better off moving.
On a well run VPS, CPU is rarely the problem. Often it's better than a dedicated, because it's a much bigger box, and you can often burst to use more cores. (of course, if everyone else is using a lot of cpu, then CPU can start to suck, too... but my experience has been that CPU is usually not a problem.)
Yeah, I'm glad they have the linode.typepad.com status blog - when Dallas is down, I can't connect to members.linode.com or status.linode.com to see what is going on! I know DOS attacks aren't their fault, but we've had a few more interruptions recently compared to usual.
I was on Slicehost, and I really struggled with whether to change to Linode. It's not that I was having any problems with Slicehost. I loved it. I just kept looking at that extra RAM on Linode, and all those data centers.
When they got the UK data center, that clinched it. I had been hoping to someday expand to UK, so when Linode did that, I changed over that day.
Interesting. Lots of disk space and bandwidth, too. Looks like their acceptable use policy is a bit strict and baroque, tough. I'm also guessing the admin interface doesn't give nearly as much control as linode's (no other VPS providers I'm familiar with do).
this lack of a 'bulk discount' is something I've given a lot of thought to. I do give a bulk discount that can seem pretty dramatic, in part because I assume the same amount of support on a $8 VPS as on a $68 VPS. -
What I've found is that this is not what many customers assume. Many customers assume that if they buy a larger, more expensive VPS from you, that they will get a higher level of support, even if ram/cpu/disk is deeply discounted on that large VPS vs a small VPS. So from that point of view, adding in a 'bulk discount' doesn't make any sense, because you end up spending just as much money supporting one large (and very demanding) customer as many tiny, undemanding customers, so from that point of view, Linode is doing the exact right thing.
Now, I've tried to tackle this by being more explicit about the "you are paying $4/month for support." bit. I'll be opening orders for large customers again later this month, I hope, and we'll see how well that works.
Another possibility is that Linode sees support costs as fixed... in many ways, they are. You need to acquire a good team, train them in customer service, and then have enough of them that at least one is on line at any time (I'm not quite to that last part yet, which is part of why I need to insure that my prices fly beneath Linode's.)
You need to have enough people to handle peaks, and support is very peaky, so unless you queue it up (causing latency and customer dissatisfaction) much of the time, many of your support people are going to be idle (or working on non-support stuff)
even so, the greater your customer's support demands are, the higher your peaks will be and the more support people you need, so I think the former is far more likely than the latter.
>You get 512MB for $19.95 and 1024MB for $39.95. If you buy 2x512MB, don't you get twice as much CPU too?
If their setup is the way I think it is, your 1024MB instance will have twice the cpu 'weight' as each 512MB instances, so, /in the case of contention/ the 1024MB instance will have twice as much cpu time as the 512.
Of course, my experience has been that cpu usage rarely goes above 50% on most VPS servers, so it makes less difference.
Their FAQ has a section on how many instances are on each host, which is 40 for 0.5GB and 20 for 1GB etc (ie, 20GB of customer instances per host, regardless of instance size). Which would mean that when you go for twice as much ram, you also get twice as much of everything else. The front page lists disk space and bandwidth, and these also scale linearly with ram size.
Linode is pretty much a pay-per-use service in the guise of a VPS host. You use twice as much resources, you pay twice as much. You stop using your VPS for a day, your account is credited back for that day. Overage charges are very similar to regular charges. And so on.
No, CPU seems to be an exception. Even on the smallest plan, you can often burst all the way to 400% (4 cores). So it depends on what other customers are doing, and if everyone else were idle, you'd get the same CPU in both plans. But the host machine is at least an 8-core beast, so one person bursting to 400% doesn't affect other customers as much as it would in other hosting platforms. Also, the main bottleneck in a VPS setup is RAM and I/O. Personally, I've never had a CPU shortage in the two years I've been with Linode ;)
In terms of iPhone apps, linode easily beats slicehost. Nice graphs for CPU, network, etc. And last i looked, slicehost required you to make a URL with you API key and mail it to yourself, which is a huge PITA.
I wrote the Slicehost app. Would love to add graphs, but that data's not available yet in the API. As for the email link, that's because the app came out before the iPhone had a copy/paste feature. You don't need that anymore with copy/paste.
Do you know why slicehost doesn't just make their private API key a link in the slice manager? I'm also curious what extra level of security the API key buys. If I lose my phone people can still mess with my slices.
I thought the iphone app was a slicehost product - but are you saying you just developed the app as a customer? That's mighty generous...
Linode is still second to Chunkhost in terms of value at the low-end, free-beta and beta tester discounts aside. Chunkhost has had 512mb for a year now, and still offers 20% more disk space. In my experience Linode's network speeds to be much less consistent and, at their peaks, considerably slower.
Yes, all this probably change when Chunkhost goes stable.
I like chunkhost and have a vps there, but the admin interface is still incredibly rudimentary. You can't really do anything other than reboot, create, destroy and reimage. Still, I'm a happy chunkhost customer, I think they have great potential and recommend them to anyone who is looking for a simple, cheap VPS.
Their traffic allotment is fairly generous compared to comparable services. Sure, if you compare it to something like 1and1 with 'unlimited' for $12, maybe not, but you don't get the same control and reliability for something like that. If you compare it to how much it would cost to serve the same requests on Amazon S3, it's much cheaper.
(None of this takes into account semiyearly/yearly pricing discounts.)
The difference is that they sell their boxes with prices proportional to memory so that doubling resources costs double (25.6MB RAM per dollar), whereas at prgmr, you charge a flat $4-per-guest fee and then charge linearly for space on top of that (64 MB per dollar).
Bonus fact: If you both were to offer a 173 MB size, the costs would be almost the same (~$6.70/mo).
It's funny, because my choice of pricing with my current support costs, and the fact that nearly all my advertising is fixed-cost 'brand building' vs. per-customer cost referrals or click ads, actually makes the smaller domains /much/ more profitable than the larger domains, which yes, probably means I've mispriced something. But the smaller domains do have a much larger 'margin for error' - if I sell a block of them to spammers, or to people expecting a full-time sysadmin, they can get expensive fast.