Realize, of course, that "Free" usually means, "We'll let you use a lot for free, until you use too much, at which point we'll either (A) Rate Limit you, (B) Have a conversation about why you are abusing the Free Incoming ALlocation, (C) Charge you for your excessive incoming data (D) Suggest you take your business elsewhere.
There is no such thing as a free lunch, and it's inaccurate of Linode to suggest they can provide unlimited free inbound data.
The statement "This means you can upload an unlimited amount of data to your Linode without having to pay for any of the incoming data transfer." will not stand for all of their customers - the vast majority, yes, but don't create a business on the belief that you can do unlimited uploads into your linode VPS at full speed. You won't be able to in the long run.
I would have preferred they provide wildly generous inbound allocations (First 5 Terabytes inbound is free, or something like that) than make a claim they can't stand behind.
I'm a service provider, and we have similar "unlimited" offers, and I see where you're coming from. For us, there's the problem that many of our customers have no idea whether "5GB" or "10GB" or what-have-you is enough for them. We're willing to offer enough capacity that we're betting that they won't ever have a problem with it, and the shorthand for that is "unlimited". We're also gambling that if we do have one or a few customers that abuse the unlimitedness to such an extent that we have to talk to them about it, that we can have a reasonable conversation with them.
I think reasonableness is a big part of claims like this. Would a reasonable person expect to be able to download the internet into Linode? I don't think so, even though they're saying that it's unlimited.
If someone can think of a better way to tell most customers "don't worry about resources" without annoying a select few, I'm all ears.
It's pretty straightforward - find a high-water mark for customers in terms of their inbound data, that 99.9% fall under, say, for linode, 99.9% of their customers pull in less than 16 megabits/second sustained, which is 7.2 Gigabytes/Hour, 5 Terabytes/Month.
And then explain to your customers that 99.9% of them use less than 5 Terabytes/Month bandwidth, and that effective now, the first 5 Terabytes/month of bandwidth is included in the cost of their VPS, and for those 0.1% of customer who exceed that 5 Terabytes, they'll get a very reasonable rate of $0.01/gigabyte ($10/Terabyte) for inbound data in excess of the 5 Terabyte initial allocation.
Realize that linode is probably paying less than $4/megabit @ 95th percentile - but the vast majority of that is likely downloaded, so they have an asymmetric allocation that they are trying to utilize with this "unlimited inbound" offer.
I just want the HN audience, in particular, to realize that the statement ""This means you can upload an unlimited amount of data to your Linode without having to pay for any of the incoming data transfer."" should not be taken at face value if you are putting together a business plan.
In general - bandwidth is priced @95th percentile on the internet, and any time you run into the phrase "unlimited" you should take it with a grain of salt and get something in writing before you rely on it.
I'm not sure that will work — it certainly doesn't seem to be solving the problem that thaumaturgy brought up. Although it spells out a concrete limit, it does not give any kind of assurance that it's (probably) enough. Lots of companies use that kind of language to make a plan sound reasonable when it isn't. For example, I don't tether or anything and don't watch a lot of video, but I'm still way above the tiny data plan that AT&T says is enough for 95% of its customers. Like, are we counting customers who don't even have smartphones?
I'd rather just "unlimited (note: OK, you can't actually download the whole Internet)" or something like that if you want to get really picky. It conveys the right message while still providing needed information to slower readers.
I suppose the difference ultimately lies in the fact that were Linode to say that a certain level of usage was 95th or 99th percentile, they could be believed. With AT&T, you have to assume that everything they say about bandwidth limits is wrong, and probably a lie.
>Realize that linode is probably paying less than $4/megabit @ 95th percentile - but the vast majority of that is likely downloaded, so they have an asymmetric allocation that they are trying to utilize with this "unlimited inbound" offer.
At the he.net fremont 1 location, at least, you can buy a full 1000Mbps pipe and run it flat out for $700/month, if you get in on the special. (The old standard price was $1000 for the same thing.) At least at that location, you are paying a lot less than $4 per megabit. (from what I've seen, paying on the 95th percentile usually doesn't make sense, as you get radically lower prices if you buy by the gigabit.)
From what I've seen of other locations around silicon valley, $0.70 per megabit is a pretty low price, but $4 per megabit? if you are buying more than a gigabit, that's a relatively high price. (I mean, you can reasonably pay that much, but you are approaching 'premium' pricing, something that a large IT shop is much more likely to do than a VPS provider.)
so I'm getting quotes for bandwidth right now, and I have a correction: he.net and Cogent are the only two major players I was able to get in that range. There are smaller players in that range, too, but only two major players, so it's probably wrong to call that 'premium' pricing. None the less, he.net is under $1 per megabit, so someone in a he.net data center, very likely, is paying very little indeed for bandwidth.
If you tried to be a service provider advertising 'unlimited' in Australia yet provide limits, you would be slapped by the ACCC (government consumer watchdog) for misleading advertisement [0].
I do not know exactly the rationale behind ACCC's decision, but reasonableness to you means something else to buyers. For example, some people think Google is the Internet and think bandwidth is free.
[0] Telstra and Optus, the two biggest ISPs got slapped for advertising unlimited internet (until 10 GB, etc).
Can't you give the option of either a) a very low-speed, but truly unlimited connection, or b) a high-speed connection with per-GB costs after reaching a certain amount of data transfer?
This is how business between ordinary people is usually done.
Paper napkins are free at every restaurant I've ever been to, but that doesn't mean that they'll give me a truckload of them in one visit.
To me it's a little different than a local ISP advertising "unlimited internet" or wireless carrier advertising "unlimited data" and then putting artificial caps and throttles on it. That's typically a home's only data pipe or a mobile device is on a long contract.
But there's much more competition for virtual servers and we all know our server infrastructures can be migrated to a different provider at the drop of a hat. Oh wait. :-)
Restaurants rarely advertise "free unlimited napkins with any meal". They include them gratis, within reason.
Nobody can provide "free unlimited" anything. There are always limits. In Australia, these claims could run afoul of consumer protection laws, with good reason - anybody claiming to sell "free unlimited" anything is lying. OK, Linode probably means well. But before you know it, every company will match them, and you won't be able to figure out which "free unlimited" plan suits your purpose because all the limits are hidden behind a "within reason" clause in the fine print.
"Free unlimited" offers turn the whole thing into what Scott Adams calls a "Confusopoly", in which companies stop competing on value by hiding what value they actually offer.
Something similar has occurred with LCD monitors, all companies advertise stats that are grossly misleading like "dynamic contrast ratio" and "grey to grey response times", but sound similar to the legitimate measurements like contrast and response times.
These fake numbers are much more impressive sounding than the "real" ones (1:100000 dynamic contrast ratio (DCR), vs something like 1:10000 regular contrast ratio (CR), when infact the monitor with a 1:10000 CR could be greatly superior to the 1:1000000 DCR monitor, which may easily only have a 1:1000 CR), so companies that provide the real numbers are penalized in the marketplace, so instead start to only publish the misleading numbers. This has made it all but impossible to meaningfully compare screens on manufacturer provided stats.
Information asymmetry plays a powerful role in marketing, and can result in a race to the bottom where consumers cannot compare things in real terms before purchase, so providers have no incentive to deliver anything of quality (extra costs, no meaningful way to market the extra quality as an option). This sort of behavior is also extremely common in residential ISPs ("unlimited" internet always sounds better to the average consumer than some kind of published limit).
Yeah, at one time the US FTC cared enough to enforce honest reporting of CRT sizes back in the old days. Apparently they don't care about LCD stats though.
All they say is that no matter how much you use you will never have to pay for it (at least that's how I read it). Obviously they can't guarantee that they will be able to offer you an unlimited amount of inbound bandwidth (there is no such thing!). Do note that they make no guarantees about the speed of your inbound connection either.
The same goes for other things, of course. You can't necessarily get unlimited outbound bandwidth and you can't buy an unlimited number of servers from them and so on... Sure, those limits (both quality and availability) aren't stated explicitly, but I still think it's fair to say that "whatever amount of inbound traffic you use we won't charge you for it". In extreme cases they will need to do something, but there are other options. There's a certain value in keeping things simple and flexible.
I do agree, though, that hiding (important) limits is generally not a good thing to do.
There might truly be no limit. Let's assume that Linode has to buy its bandwidth symmetrically. If someone can confirm whether that is true or false, it would greatly be appreciated. With that assumption in hand, I think we can add the assumption that a web hosting company mostly does outbound traffic - people downloading things from their servers, not the other way around. They might have so much spare inbound capacity that it really wouldn't matter.
Plus, there's a small "what goes up must come down" logic to it. So, you get a Linode and you start pushing as much inbound traffic to it as you can. How much can you push to it for any purpose you can see as useful (useful being defined as something other than testing whether Linode will consider you abusive at some point)? The largest Linode (costing $800/mo) only comes with 800GB of storage. At that point, you need to push some of that data outgoing or delete it before you can really take more inbound data. You could run a service that was 24-hour YouTube - a site where a video uploaded would only be available for 24 hours and then deleted. Of course, the point of that site would be for people to watch it which would be outbound traffic.
The reason they might not be stating a limit is that there might not be a limit. Rather, I should say that you're going to be limited by storage. There's no point in taking the data if you have no place to put it and I can't really think of pulling in data that one would want to store for such short periods of time that it would put a strain on their network. This isn't rhetorical: how would you see yourself or someone taking inbound traffic on a Linode that wouldn't similarly see that traffic go outbound or need to be stored?
> This isn't rhetorical: how would you see yourself or someone taking inbound traffic on a Linode that wouldn't similarly see that traffic go outbound or need to be stored?
In general I think your argument is pretty strong, but I _can_ think of a use-case: if you were analyzing the data, and reporting very brief aggregate results.
Say, have swarms of web browser extensions uploading all the domain names their users are requesting pages from, and then sending to a single user, once per hour, the list of the domains, and how many users requested each.
Edit: oh, look, a few comments lower on the page, other users came up with the same example (search "telemetry")
What's the downside to the customer of just spelling out the actual limits?
Lots of customers have no idea, and most of the time it doesn't matter. It's a friction that just isn't needed for most customers.
Abusive uploading behavior isn't a line, it's a grey area. If you are paying for 30 servers, then the line is in a different place to if you are paying for one.
Yeah, the optimal place for the system as a whole is when it "just works" for everybody. If you put limits and overages in the contract, the the customer may get penalized (or at least subconsciously feel the need to restrict his consumption of your service).
This is probably a reasonable assumption coming from the world of ISP "all-you-can-eat" packages where they're saying "unlimited" but in reality it's not, because unlimited starts affecting other customer so it's "unlimited (T&C's apply)"
This is a completely different scenario though:
* Transferring stuff to linode isn't generally an end in and of itself. There are far fewer use cases for it. At the end fo the day you probably need to get data back out of it.
* The vast majority of cloud hosting and hosting providers in general's bandwidth is outbound. People download from the web. These hosting companies provide a lot of the web's main infrastructure. You're probably looking at a ratio of 20:1 or 30:1 here.
* Linode doesn't pay for uploads and downloads with it's carriers / ISPs. It pays for circuits capable of synchronous upload and download speeds. Using the ratios above, if they, for example, have 300Mbps of traffic outbound, then they have a lot of spare inbound capacity.
* Abuse here is difficult. Downloading movies, music etc. are consumed at home. It's difficult to imagine so many usage scenarios where huge data volumes would be consumed so easily on a linode server without having to be downloaded again. Maybe someone can come up with some creative suggestions here though? :)
* Bandwidth allocation and contention is very different within a hosting providers access network than an ISPs access network. ISP last miles, local DSLAMs, AP's etc. are huge points of contention. Inside the data center, I imagine most hosting providers "last mile" are 100Mb or Gig/E here straight to the servers with gig or 10gig/e uplinks from access switches.
The basic crux really is though, they have shedloads of free inbound bandwidth and why not give it away for another USP.
Creative use of unlimited BW - downloading the internet for data-mining purposes. You can't store it (as Linode's disks are pricy), but who cares? You can just download it again if you need to re-do your calculations.
Totally agree with you on all your points - but "shedloads" is different from "unlimited."
All I'm really trying to do is catch the attention of the HN'er who read the article on "Unlimited Upload" to linode, came up with a great idea (for example) for a remote telemetry application that required little back end processing, would be installed on millions of systems, and would upload 100 mbit/second aggregate to the linode environment.
If I managed to get that person's attention, and cause them to consider the possibility that the "unlimited" wasn't really "Unlimited" - then I'll have accomplished what I set out to do.
I think we had this discussion before about EC2. As long as Linode's total inbound is less than total outbound (which is very likely due to their focus on hosting rather than backup), it is free to them and thus they can make it free to their customers. Also, customers who upload a lot of data tend to store or process it, which Linode can make money on.
Is there a way to take advantage of it though? They charge for both out bound traffic and storage. So unless your just pumping data in to a remote /dev/null they are going to make it up eventually either because you are storing it or because you are pushing it back out. I guess it would make a pretty nice back up server though, only pay for bandwidth when you need to restore.
The preferred terminology is "unmetered" which typically means that you're sharing something like a 100Mbps pipe with other users and you accept the inherent bandwidth and latency issues that comes with that kind of setup.
If you want guaranteed bandwidth that costs extra and will definitely be metered, either for a fixed pipe size or per GB.
We'll let you use a lot for free, until you use too much, at which point we'll either (A) Rate Limit you
They throttle already. I had an app on Linode that required large uploads (around 1MB) but had to move to a different provider because of the throttling. The throttling kicks in during a single upload.
That said I would choose Linode in an instant for different apps that didn't require uploads because my experience with them was excellent and their network appears fast (possibly because of the throttling).
I think they can stand behind this claim. The largest plan they sell comes with 800GB of storage. Your OS installation will probably take 10GB and you're effectively left with 790GB to store your data in.
Unless you're uploading and deleting 790GB each and every day I don't see why they can't afford it. Considering that the most people will buy Linode 512 (20GB of storage including the OS) the numbers we are talking about are rather small.
This must be in response to Rackspace Cloud offering free incoming data due to their forced migration of Slicehost's customers.
Rackspace Cloud has a bit of a different audience to Slicehost's existing customers (I'm one of them).
If I were in Linode's position, I'd be offering migration coupons to existing Slicehost customers as an incentive to move to them instead. It's an opportunity wide open.
The way Slicehost has been communicating with their customers recently regarding the forced migration has been enough incentive. It's almost like Rackspace don't want the business the way they've been conducting themselves.
I've got a slicehost machine out there, and their communication with the cutover to rackspace has been awful. I'm still not sure if it's going to be mostly transparent, needs dns coordination, or if I'm going to have to rebuild my vm on the cloudwhatever. To be fair, it's not a common configuration (debian lenny/32bit,384mb), but everything I see just makes me think that it's going to be more of a pain than I want to have to deal with.
Linode looks good, are there any red flags to worry about?
Seriously, this forced march from Slicehost to Rackspace Cloud has been an absolute trainwreck and I've been meaning to move all my services away to Linode as soon as I can find the time. Maybe during this long weekend.
I don't see what's so hard about unlimited meaning unlimited. Australia has the right idea here. If you advertise no limits, and then attempt to enforce limits, you rightfully deserve the smackdown you'd get.
Moral of the story: Don't say you can provide something you can't.
This is great! I've been using my Linode as a VPN for when I'm working in coffee shops, so I'm very happy that half of my traffic is not counted any longer.
I know a place that does unlimited bandwidth. You're limited to 10Mbit/sec throughput though, so your theoretical maximum usage per month is only a little over 3TB.
I've been using Linode for over a year, and I must say their service is the best I've seen. Their customer support is also off the charts brilliant. I think this is truly wonderful.
I think the more important news for many is that the overage for outbound traffic is the save price as pre-paid plan, so you could lower your plan (e.g. going from Linode 1024 with 400GB to Linode 512 with 200GB). Serving static pages shouldn't cost much memory and processing power anyway.
We recently moved to Linode from AWS (after the recent AWScalypse)
However, the problem we currently face with linode is that the Disk spaces are puny and "attaching" additional storage is very costly as compared to AWS. There is no reduced pricing for Yearly contract either.
Love their sub-minute support times. Can't beat dat!
Prepaying for a year saves you 10%, prepaying for 2 years saves 15. If you already have an account, just open a ticket and you'll have the discount in minutes.
And, even if you prepay, you can still cancel and get a full refund. Even if you know you'll only use the server for 1 month, prepaying for 2 years is a smart move!
Keep in mind that if you prepay for 2 years, use the server, and cancel before the 2 years ends, you're refunded the amount paid minus the normal monthly charge times the number of months you used the service. The discount gets wiped away, at least it did for me.
My understanding is that Linodes aren't really "for" storage. IIRC it's often more cost-effective to just upgrade to a larger Linode than to add additional storage.
I'd recommend storing bulk data on S3, but as you said, you're wary of AWS :)
EC2 has had some issues in the past, but did those issues affect S3?
Even if you're on linode, you might find sending large binaries to S3 is a cost-efficient way to increase your storage size. Plus, you have a built in CDN when you need it.
There is no such thing as a free lunch, and it's inaccurate of Linode to suggest they can provide unlimited free inbound data.
The statement "This means you can upload an unlimited amount of data to your Linode without having to pay for any of the incoming data transfer." will not stand for all of their customers - the vast majority, yes, but don't create a business on the belief that you can do unlimited uploads into your linode VPS at full speed. You won't be able to in the long run.
I would have preferred they provide wildly generous inbound allocations (First 5 Terabytes inbound is free, or something like that) than make a claim they can't stand behind.