Linode vs. DigitalOcean – performance benchmarks
jasonormand.com
jasonormand.com
I think it's funny how they can go from barring any information "offensive, abusive, inappropriate, malicious, or detrimental, including, but not limited to, those that: Are obscene, fraudulent, or discriminatory, including any containing profanity, or obscenities" to specifically permitting "adult websites that abide by state and federal law and regulation."
In any case good, luck keeping your blog, blog, comments, or whatever else you host with them clear of all profanity all of you who decide to try their services, because you've just walked into a convenient TOS violation. Damn! (Oops. TOS violation. Goodbye HN. Ha ha.)
As you have raised this issue, to make the situation clearer, the phrase referring to obscenity and profanity has been removed from our terms of service.
Thanks,
Etel
Good support counts for a lot, especially when the shit hits the fan, but resources/dollar count too, and it'd be nice to know exactly what kind of tradeoffs you're making, or if a hosting provider is simply below the curve.
It seems like sound testing and analysis of the things that are readily tested.
He doesn't even have a link to either provider. Doesn't look like he is looking for referral $ or anything other than sharing what he learned.
I did not say anything about the article being spammy. I said "hosting comparison companies". The article is quite obviously on someone's blog.
Compared to these sites, this article is chock-full of data. And I don't see a referral link anywhere. :)
I will be clear, I am not profit maximizing at all. The amount of companies trying to buy placement is staggering. It's very easy to understand why my competitors would do it. I could probably make 10x the money if I accepted their offers. But I don't plan on just doing web hosting forever, my goal is to scale the technology and review lots of things. If I sell out my brand now, who would ever use it later?
"one of these spammy hosting comparison sites..."
I am very open to new ideas and suggestions to improve it.
http://serverbear.com/9683/linode#view-uptime-beta
Plus benchmarks etc (user run, which means you get a decent dataset to compare by server spec/location).
http://serverbear.com/9683/linode#view-benchmarks http://serverbear.com/9806/digitalocean#view-benchmarks
We've actually considered ticket response times, a lot of hosts use WHMCS to manage their nodes - which we could make a plugin for. But a large amount of the more established hosts use a custom system.
Problem is, ticket response time isn't an indication of quality. Anyone could respond with a "We're looking into it" within 5 mins & not get a resolution for a few hours.
I didn't say it would be easy... you'd have to have a somewhat subjective metric of "did a person answer?" and "did they tell me something useful?". Maybe you could use Amazon's mechanical turk or something like that...
Can I give you a feature request? Can you track terms of service or somehow summarize them? I would love to filter/find hosts based on how liberal their terms are.
See also this comment/response from armored_mammal re: DigitalOcean's ToS:
That is a pretty good idea, I'll pop it into our roadmap. We've got a chunk of time allocated this month to work on some stuff & that's an easy quick win :)
+1 Linode.
While re-arranging the garage, I found some USB speakers and ran across the Cr-48, and set it up for some music in the garage. My daughter immediately wanted it back. She prefers Ubuntu with Unity to OS X and prefers the Cr-48 even though she agrees it isn't as snappy as the MacBook.
McDonald's isn't a quality product even though it is mass produced and a lot of people like it. Why isn't quality? Because analysts have questioned the quality of the meat in their products. Is meat? People who overeat the food might develop illnesses like heart disease and diabetes. That is a low quality product.
Microsoft Windows is a quality OS. A lot of people like it. A minority don't like. They have the top leaders of UI design and OS engineering working on the product.
If Windows is not a quality product? Then how do you define quality?
I'm considering switching to DigitalOcean just for the backups. Are these included in the cost?
One time I had to reset a VPS but in the end it was due to a mistake I made as root.
Generally I treat VPS like any other co-location hosting operated by a 3rd party. I make sure to have redundancy and backups managed seperately.
The only other situation I can imagine is site-wide issues, which you can do nothing to help with anyways so just sit back and wait for them to get their stuff together.
In other words, if you know your way around the shell and your server OS, the need for support becomes very insignificant.
On a few occasions, the machine has gone down because of some maintenance or some issue that comes up. I was not notified of this and had to submit a ticket myself, after which I was told that the techs had taken down my machine. After asking why I wasn't notified, I was given no direct response.
I've since migrated the service to Linode and holy shit, it's such an improvement. The support is great, but even more their admin interface is excellent. Gives me so much more info than I ever had before. It's even a little cheaper than the plan I had with Slicehost.
Just to throw some numbers out there: jsonip.com is a node-based app that supports upwards of 6-10 million requests per day. Lets just say that I couldn't be happier with Linode.
Note to Linode: Do not ever sell out to Rackspace or else I will be forced to go live in the woods somewhere. Thanks.
In the future, if it started capping out, I'd look at alternatives. But I'm not there yet, and I like to continue experimenting with node. =)
Use the DigitalOcean API from Ruby and Chef/knife to create/shutdown/destroy virtual servers ("droplets"):
Easy to spin up a machine with no usage and get great performance if there is nobody on it. That doesn't mean this data isn't correct of course but keep in mind the OP is making a comparison between 1 machine at Linode and 1 machine at Digital Ocean at a given point in time. Same equipment or different equipment at a later date could yield different results as the equipment gets filled up or depending on who else is using it. We had VPS's at Media Temple that would literally stall periodically for a few seconds. We then migrated to a different machine at a different data center and it hasn't happened since.
I will continue using linode for my servers that matter, and digital ocean for toys and experiments, just because it's so cheap.
Check Out Simplenode. ( I am not related to them in anyways )
I'm not saying that DigitalOcean suffers this problem but the cheaper a VPS provider is the greater chance of it being used by folks running bit torrent, warez downloads, etc and/or folks who think they need to smash the (shared) resources out of server to get their moneys worth.
Paying a few dollars extra for a Linode instance (and I have 100's of them, so it adds up) is worth the peace of mind of being unattractive to these kinds. IE 'bad neighbors" who will pee in the pool.
Support is also top notch (although I rarely need it).
Wouldn't this be more likely on a higher-priced VPS (like Linode) than a lower-priced VPS (like DigitalOcean)?
Which is funny because Linode has pretty bad uptime.
When I was on their Fremont DC it went down all the time.
And nice that you were alerted of the problem. I often found out hours AFTER their issue was resolved what was happening.
As for alerts, maybe they only automatically create tickets and alert users when there's a problem with a vm or the host it's on. Agree with you that doing the same for DC issues would be a good idea.
Disclaimer: Customer for ~4 years, VPS in all DCs but Tokyo.
They then try to eeek out every bit of performance they can from the accounts they "collect". It's a curious sub-culture.
People certainly have different uses than you (and your expensive 100's of hosts), that doesn't make all of them script kiddies and 12 year old hosts.
I respect the fact that the author tried hard and didn't bias the results, but I fail to see where he even repeated the benchmark over a number of days to attempt to eliminate variations in what the other guests were doing while the benchmark was running.
As I see it, the only way I would trust a benchmark of this nature was if it was somehow ran over multiple hosts and was able to sample a large fraction of the physical boxes over a large period of time. As you can image, that's impossible to do on S3 and any of the other larger VPS providers
Even if somehow someone did manage to pull off a benchmark like the above, I still wouldn't trust it because it would only be valid until the provider added "Joe's Cheap Video Encoding Service" as a client that's sharing my box. Or someone else who's running a benchmark just like you are?
With all of the above said: I'm sure that there is differences between the VPS providers in how they allocate CPU and IO resources and Digital Ocean may indeed be much better on average. I'm just not sure how you can measure it objectively to make sure they are or that they will continue to be so even 5 minutes in the future.
Which is a shame, because an objective metric that was trustworthy that compared all the different providers would be absolutely awesome.
With Linode, I know that I am getting excellent support. There's the peace of mind of knowing that when I submit a ticket, someone will read and respond in short order.
A $5 SSD VPS sounds great, but I automatically wonder what kind of hardware I will be on (does it compete with Linode's RAID 10 configuration?), and how long my support requests will take to be addressed, and whether or not the low prices are a sustainable business model.
I mean the best example is airlines. For the longest time the big national airlines claimed that they would laugh last since their support/customer care was /so/ much better than cheap no-thrills airlines.
But look at what happened! The no-thrills airlines have been stealing market share year upon year, people are booking whatever is cheapest in the price comparison all other things be damned.
The same is true of retail Vs. the internet. Retails shops claimed that they wouldn't lose market share because people loved the one on one customer service and face to face interaction, but clearly they were mistaken.
In fact the ONLY company I can think of who has high quality customer care/support AND is actually growing is Amazon. But they also often happen to be the cheapest.
Not sure about the US, but when buying electronics Amazon is usually somewhere in the middle, price-wise. They are almost never really the cheapest and if so, this is usually due to a Amazon Marketplace seller, not Amazon themselves.
Whereas in low-end VPS services, the labour cost for Great, Responsive Service will quickly begin to approach your capital outlay.
Amazon on the other hand, doesnt make any money - http://www.slate.com/blogs/moneybox/2013/01/29/amazon_q4_pro...
Talking about amazon as if they're a charity is not apt. Amazon makes a great deal of money and judiciously reinvests it.
The argument goes, no other company gets such a free pass from their shareholders. Amazon could be amazingly profitable right now; in real terms, shareholders are subsidizing their current strategy in ways that Apple or Microsoft or Walmart could never get away with.
This is not true in the hosting arena, at least.
Amazon is like everyone else in this industry; when they came out, they came out with very compelling prices.
Well, costs fall with moores law. Amazon prices (especially bandwidth prices) have not.
There are... a lot of new entrants to the market, and they /all/ have prices that are dramatically cheaper than amazon.com. hell, most of them make me look overpriced, and it wasn't so long ago that I was the unreasonably cheap option.
begin to approach? You need... dramatic scale to spend more money on hardware than on support, even when you have pretty minimal support. I spend rather more on labour than on hardware.
Also, to be fair to airlines, they have immense labour costs.
yeah. I also think that it's a good example of how some things? you can skimp on, while other things? not so much. I mean, if the airlines show that you can eliminate (or charge extra for) in-flight meals without disturbing anyone too much. You can even pack 'em in tighter (though, some people will pay extra for a little room.) - but yeah, you've still gotta keep the planes in the air.
Another interesting bit is that I'm not sure that it'd be cheaper to maintain airplanes to a lower standard, even without the customer backlash.
That's the thing; sometimes, the cheaper part is just as good- (for example, I think going supermicro is just as good as going dell, assuming the person assembling uses ESD protection.) - but other things? non-ecc ram, for instance, in my unscientific opinion, usually ends up being more expensive in terms of downtime and technician hours than ecc ram.
Of course, airlines are also almost all unionized; The hosting market is almost the opposite. Generally more is expected for less pay in the hosting market than of the same technical roles in other sectors. Traditionally, this means that many people start in the hosting market, then move up (certainly in terms of pay) into a corporate networking or corporate sysadmin role.
To be honest, the poor support was one of the reasons we moved away from Linode. We were in the Newark datacenter for about 8 months with 12 linode boxes. Had frequent issues with their load balancer and most of the time when we told them there was a problem, they asked us to prove it.
A few times we had extended outages due to "unscheduled maintenance"
Prior to running my business on Linode I had a single VPS with them for 2 years for personal stuff. I had zero problems... so YMMV. Overall, I'd still use them again, but not for mission critical stuff after that experience.
With that said, would you trust a $5/mo. VPS provider like DigitalOcean with "mission critical stuff"?
I still remember the old days when I switched from slicehost to linode, 30% (360MB vs 256MB instance) cheaper and later 50% (512MB vs 256MB)..
When you're comparing two businesses, you can safely assume that given the free market, most products with similar features converge to a similar price range. If a product offered by a company appears to be similar in features to another established product but does it at 1/4 the cost, it is completely valid to wonder what costs are being cut to achieve that price. Is it their infrastructure? Their support? Do they pay their team less? These are all valid questions, whether you are able to answer them given the available information or not.
You do realise that in the past user VPSs were rooted, Bitcoins stolen and Linode users had to find out from Reddit that their VPS was potentially hacked.
If you trust Linode (or any VPS really) with mission critical stuff then you are (being) an idiot. I am sorry but you are.
Edit: The idiot is a reference to behaviour not anything personal.
And frankly I can't understand how a VPS provider can act in this way and still have people acting like they have great support. As a customer I found out from Reddit before Linode. That's a pretty disgraceful effort.
1) I don't think your comments labelling those who disagree with you as "idiots" is constructive or in keeping with the spirit of HN.
2) "Mission critical" to one person may mean something else to someone else. Depending on their requirements, those who consider something to be "mission critical" may be willing to accept varying levels and/or guarantees of uptime or security.
Look at Cloudflare's handling of their incident. Then look at Linode's.
Probably was just a request for logs or other such evidence showing what went wrong so that they could diagnose and fix the problem? Might have been a misunderstanding
> A few times we had extended outages due to "unscheduled maintenance"
Hardware isn't magic.
It sucks that you didn't have a good experience, but this is highly anecdotal.
Here's an uptime report of our hosts on Linode from Pingdom: http://i.imgur.com/KggJZ.jpg
I handle customer support for us in a helpful and compassionate way. The support we were getting from linode sometimes felt like they didn't even read what we wrote to them.
I'm seeing 24 support threads listed in Gmail with them from 5/21 to 8/31 for various networking and uptime issues.
You can easily achieve good performance on some common desktop hardware (or some cheap home-built server) and some SSDs, but obviously, no one would host nothing more than a personal blog on it
More recently, their network has been the target of DDOS attacks, etc. I'm guessing this is partly due to their clientele. Also, all of their Las Vegas servers went down earlier today.
The lack of backup options are making me consider switching to DigitalOcean. We'll see if their reliability lives up to the hype.
"Investors are really looking for a couple of things... At DigitalOcean, we were able to differentiate ourselves by focusing on the mid-tier market, and catering to the needs of individual developers that were being completely ignored by the larger cloud hosting providers." http://www.forbes.com/sites/danreich/2012/09/19/startup-ceo-...
I think it would more fair though to compare a $20 Linode with a $20 offer from another provider.
It is also the recommended go to spot that is linked from a lot of "hack forums" for information on where to find VPS providers who don't have solid abuse handling in place yet.
Inappropriate content is also disallowed, what does that mean? It's not a legal term of art that I'm familiar with. Why does a sentence begin with "Transmit"? And please, consider using bulleted lists.
(I'm assuming lower is better for encoding time.)
I'm not sure quite what it was--whether malicious user behavior, a lack of separation in bridge-interface setup making network traffic on the box un-QoS-able, or maybe someone on the box I was on bearing the brunt of temporary, persistent DoS attacks--but I didn't care to roll the dice on another box from the same host, when I knew that this kind of thing "just doesn't happen" with the major providers.
Still, if there was a low-end host that specifically mentioned stable network throughput in their SLA, I'd love to try another one. $7/mo sure does buy a lot of horizontal scaling and High Availability, once you can ensure each node can hold onto the little bit of traffic you send it. :)
Linode was a great provider, but now I can manage my own VMs using Xen myself.
Background - my employer hosts on dedicated servers at Rackspace and we're considering using more of their cloud offering.
We're moving away as quickly as possible to Linode and AWS.
Like I said, Rackspace's model of hiring great people and giving premium service worked great in the old days pre-cloud, but now the people are poor quality as the company has grown, the underlying tech is horrible, and the billing and other services so bad that you can't justify the premium prices anymore.
Rackspace servers and service (in all forms) are considered harmful. Just go with AWS or Linode.
They need to call your over the phone in order to verify you are a human being and they need to know what you are going to do with their "cloud" - they called it "Onboarding"
I told them I just want to experiment with their apis and maybe they think I have no business value or I am a potential spammer, so they rejected my "Onboarding" request without any reason given. (So you know I am not commenting from the technical point of view since I have never access to their system)
On the contrary, AWS give you free tier access for a whole year, no question asked.
IMHO, don't waste time on Rackspace.
Also, their sorta-cloud-sorta-VPS setup for Rackspace Cloud seemed awkward to me. It made the control panels more confusing than they needed to be.
I really like openstack. As a python hacker, it's the bee's knees.
But Unlike AWS where you get all the other AWS features and services via API and a half decent price if you play their reserve pricing and bid game. Rackspace has NONE.
Pros: Great pricem, performance is excellent,UI is great for creating servers
Cons: I have had two periods of about 20 minutes of down time in the early mornings recently.
Digital Ocean is great hosting for developers. They are really doing it right.
Linode has been there through thick and thin for me. Sounds cheesy, but they report on and fix issues before I even notice them, I've only had to submit a ticket once (for something that was my fault), and their infrastructure is tight as a drum.
Linode is very, very reliable. Another thing that many people who work with larger applications than mine will tell you is that speed is not important, what's important is consistency. Linode's virtualization is set up to provide, above all, consistency (even on a loaded host). So sure your site might run 1.6778 times faster on DO, but when the host server is loaded your site's going to be crawling because the other neighbors are bursting their CPU into your space.
We'll see if DO's prices and performance stand the test of time. My guess is their support is extremely lacking in comparison. When everything's running perfectly, this is great. But when the chips are down can you really rely on them? I guess we'll see.
Linode would routinely not respond during major incidents (e.g. power outage at Fremont which was common for me) and when they did they were often not clear about what was happening. Their infrastructure is hardly tight as a drum (Google 'Linode bitcoin').
So maybe Linode's main support system is set up to favor Dallas, which if this is true is pretty shoddy. I'd be interested to hear more experiences from other people across different datacenters.
I wish this had been a $20 vs $20 plan comparison.
I have a Linode 1Gb (and another 512Mb) and I rarely go over 110% CPU utilization even when my server is busy (the max is 400%, 4 cores being used).
Also, price comparison, Linode charges an extra $5 (for 512Mb) and $10 (for 1 Gb) for Backups. DO seems to be giving them for free.
My yearly bill on Linode : $900
My yearly bill on DO : $180
I think I'm trying DO soon, especially I'll be in need for more servers.
Edit: I'm a little afraid from DO. They are simply too^8 cheap and it's scary. Will they be around for too long? What's their down time like? Will they lose my data (and backups)?...
Price wise, DO beats the hell out of anything. I can renew my smartphone yearly for free just by moving.
You lost me there. Is that slang for something?
By switching to DO (expecting they offer a comparable service), my purchasing power gets bigger and allow me to become richer (renew smartphone or buy some other stuff).
I'm just trying to show how competition, technology and efficiency makes us richer.
Otherwise 3-4 year failure is quite common.
I might consider moving off Linode for pricing alone.
Also, see: http://serverbear.com/9806/digitalocean for interesting benchmarks over time (with all the standard caveats, yada yada).
http://digitalocean.uservoice.com/forums/136585-digital-ocea...
Who is this "Terrence Andrew Davis" fellow who comments on pretty much every single post that reaches HN?
He sounds like a lunatic and religious fanatic constantly quoting bible verses and sermons. Pretty darn creepy to read at 4am.
If it disturbs you you can always turn off showdead (or if he made a new account, wait until it's banned as it soon will be.)
I had a handful of dyno's, a worker, and a few add-ons running my iPhone app's API within Heroku up until about a week ago where I transitioned everything onto a small collection of Linodes in the UK datacenter. We're launching in Europe so it made sense to bring everything closer. If you're comfortable and skilled with a terminal, you can get a lot more bang for your buck with Linode.
That being said, I am always interested in trying something new. For a $40 Linode you get 1GB of RAM, with DigitalOcean the same $40 gets you 2GB of RAM and they run on SSD's. I'm going to deploy a copy of my primary API server here to take it for a spin.
First impression thus far: fast deployment. From pressing the submit button to shelling in took literally less than a minute.
Unfortunately though the latency on the box seems a bit high. There is a bit of lag in the terminal, something I do not experience with Linode's UK datacenter.
I'm in Sweden on a 4G modem, here's a traceroute between me and both boxes. DigitalOcean is first, followed by Linode.
### DigitalOcean – Amsterdam Datacenter
traceroute to 192.81.221.48 (192.81.221.48), 64 hops max, 52 byte packets
1 tiny (192.168.0.1) 10.573 ms 1.658 ms 1.393 ms
2 10.9.14.115 (10.9.14.115) 33.218 ms 24.250 ms 59.223 ms
3 10.9.14.113 (10.9.14.113) 51.009 ms 35.878 ms 30.836 ms
4 90-229-25-24.link.se.telia.net (90.229.25.24) 33.153 ms 30.662 ms 34.999 ms
5 s-b2-link.telia.net (80.91.247.96) 33.962 ms
s-b2-link.telia.net (80.91.246.228) 74.544 ms
s-b2-link.telia.net (213.155.133.147) 77.395 ms
6 s-bb1-link.telia.net (80.91.246.148) 36.792 ms
s-bb2-link.telia.net (80.91.246.234) 29.966 ms
s-bb2-link.telia.net (213.155.133.142) 31.228 ms
7 kbn-bb2-link.telia.net (213.155.134.141) 151.952 ms
kbn-bb2-link.telia.net (213.248.65.29) 42.281 ms
kbn-bb2-link.telia.net (80.91.246.178) 123.130 ms
8 kbn-b3-link.telia.net (80.91.249.51) 39.032 ms 45.868 ms
kbn-b3-link.telia.net (80.91.245.159) 44.708 ms
9 te4-7.ccr01.cph01.atlas.cogentco.com (130.117.14.33) 67.971 ms 60.729 ms 96.055 ms
10 te0-7-0-0.ccr22.ham01.atlas.cogentco.com (154.54.62.117) 132.391 ms
te0-7-0-6.ccr21.ham01.atlas.cogentco.com (130.117.2.149) 97.568 ms
te0-7-0-0.ccr22.ham01.atlas.cogentco.com (154.54.62.117) 137.713 ms
11 te0-0-0-3.mpd21.ams03.atlas.cogentco.com (130.117.50.57) 101.141 ms 89.885 ms
te0-0-0-3.mpd22.ams03.atlas.cogentco.com (130.117.1.101) 147.596 ms
12 te2-1.mag01.ams03.atlas.cogentco.com (130.117.48.62) 104.005 ms
te1-1.mag01.ams03.atlas.cogentco.com (130.117.0.242) 52.704 ms 127.224 ms
13 nxs-internet.demarc.cogentco.com (149.6.128.6) 98.294 ms 130.686 ms 142.924 ms
14 192.81.221.48 (192.81.221.48) 148.148 ms 107.753 ms 134.880 ms
### Linode – London, UK Datacenter
traceroute to blitzen.dbld8.com (176.58.120.7), 64 hops max, 52 byte packets
1 tiny (192.168.0.1) 9.671 ms 2.313 ms 1.411 ms
2 10.9.14.115 (10.9.14.115) 32.881 ms 21.701 ms 34.152 ms
3 10.9.14.113 (10.9.14.113) 36.843 ms 21.986 ms 26.750 ms
4 90-229-25-24.link.se.telia.net (90.229.25.24) 26.421 ms 24.166 ms 26.982 ms
5 s-b2-link.telia.net (213.155.134.55) 27.415 ms
s-b2-link.telia.net (80.91.248.199) 33.347 ms 37.758 ms
6 s-bb2-link.telia.net (80.91.254.114) 34.993 ms
s-bb1-link.telia.net (80.91.254.58) 92.886 ms
s-bb1-link.telia.net (80.91.247.216) 38.297 ms
7 hbg-bb1-link.telia.net (213.155.135.236) 83.932 ms
hbg-bb1-link.telia.net (80.91.251.41) 36.076 ms
hbg-bb2-link.telia.net (80.91.247.147) 47.708 ms
8 ldn-bb2-link.telia.net (80.91.247.201) 67.937 ms
ldn-bb2-link.telia.net (213.155.132.247) 106.801 ms
ldn-bb2-link.telia.net (80.91.247.169) 63.880 ms
9 ldn-b3-link.telia.net (80.91.249.170) 54.534 ms
ldn-b3-link.telia.net (80.91.251.237) 101.259 ms
ldn-b3-link.telia.net (80.91.247.25) 64.316 ms
10 telecity-ic-150799-ldn-b3.c.telia.net (80.239.167.94) 59.564 ms 70.329 ms 64.835 ms
11 85.90.238.70 (85.90.238.70) 70.606 ms 68.463 ms 60.808 ms
12 212.111.33.234 (212.111.33.234) 69.237 ms 60.927 ms 70.726 ms
13 li522-7.members.linode.com (176.58.120.7) 65.147 ms 60.434 ms 62.646 ms
DO just need to improve on their Networking. ( But i guess you get what you paid for )
I don't think anyone can be a "hands down winner", because it varies based on the needs of the user. Businesses frequently use Cogent for bandwidth at the lowest price point. Their business model could even depend on it. Other businesses require low latency, service level agreements, etc.
And yes their Networking isn't that great.
I currently use one of the cheap servers from http://www.kimsufi.co.uk/ (9UKP/mo for 2Gb RAM, 500Gb space and 5Tb bandwidth (after 5Tb you are not cut off but bandwdth is throttled, though I've never come close to that to see how much effect it has)). Excellent if you don't need any hand holding (it is a completely self managed service) or much support (I've not needed to contact support so can't comment, but again "completely self manged"), or a very specific OS/kernel (they have their own builds, getting stock versions intalled is possible but a bit of a faf), you want a pile of space, and your tasks are not CPU intensive (the cheaper machines only have Atom or Celeron processors). Having your own dedicated drive makes I/O contention a moot point unlike with a VPS (though depending on how loaded DigitalOcean's host servers are they may still outperform this for some I/O patterns due to the use of SSDs).
Its KVM not VMware
I host with Ramnode.com, and I love their prices and SSDs. Not one issue yet except the odd DDoS attack. Used the server bear coupon code to get 31% off my services for life.
I wonder how it stacks up to DigitalOcean. I also have a linode server, but I'm thinking of cancelling because it is too expensive, but it is solid as a rock.
I'd like to be able to leave processes running that might have a lot of data mapped into their address spaces, but for much of the time that data can safely be paged out to disk and not used. When I come and use the machine, I'd like to have a large amount of RAM, but when I'm not using it, I don't want to pay for it.
We have all of this virtualization technology but I still have to make hard decisions about size ahead of time. There are lots of providers that will let me burst CPU, I'd like to burst RAM. For people building small apps that get occasional usage, having more flexibility on memory would be a huge boon.
With RAM prices being what they are, Linode should price their RAM more competitively and offer SSD for those who want it, at a reasonable price.
On the other hand they also have a data centre in Tokyo, which means noticeably better pings to Asia than DigitalOcean.
In any case, will give DigitalOcean a try for a couple of projects.
Might also want to try OVH and Hetzner for dedicated servers for about £50/month.
You'd need to provision and benchmark a bunch of Linodes to get meaningful numbers.
That's one of a dozen ways in which a VPS is not like a real server.
5 second benchmark which corresponds well to real-life disk performance:
sysbench --test=fileio --file-fsync-freq=1 --file-num=1 \
--file-total-size=16384 --file-test-mode=rndwr run --max-time=10 \
| grep "Requests/sec"It's not entirely rational, but if you ask around, I'd wager most people who use Linode love it.
Maybe it's because it used to be just caker doing everything, or the free disk / memory upgrades he kept handing out. I remember him asking about LASIK surgery in the (very helpful) forums; there's a family feeling to it. It's the anti-Rackspace. It's home.
So true. I moved a mail server I manage over to digital ocean recently because it needed more RAM and the price difference between a 4GB DO instance and a 4GB Linode one was just too much to swallow.
Very happy with DO so far but... I still have 7 1GB and 512MB vms with Linode and no plans to move any of them because I just can't bring myself to sever all ties with Linode :D
I've been a happy Linode customer as well. But being able to deploy silly side projects like this for just $5/mo is really cool. This is my 3rd DO VPS and they've all been great. I had a PostgreSQL problem on another one and my support requests were replied to really fast.
Alternatively, if anyone has successfully installed Windows on a VPS instance, please let me know :)
#linode London
Server Software: Apache/2.2.16
Server Hostname: localhost
Server Port: 80
Document Path: /
Document Length: 7483 bytes
Concurrency Level: 20
Time taken for tests: 33.140 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Total transferred: 8037000 bytes
HTML transferred: 7483000 bytes
Requests per second: 30.18 [#/sec] (mean)
Time per request: 662.791 [ms] (mean)
Time per request: 33.140 [ms] (mean, across all concurrent requests)
Transfer rate: 236.84 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 2.1 0 16
Processing: 118 658 1143.2 487 19006
Waiting: 113 655 1138.1 485 19006
Total: 118 659 1143.2 488 19006
Percentage of the requests served within a certain time (ms)
50% 488
66% 563
75% 616
80% 658
90% 779
95% 932
98% 3635
99% 6406
100% 19006 (longest request)
#DigitalOcean Amsterdam
Server Software: Apache/2.2.16
Server Hostname: localhost
Server Port: 80
Document Path: /
Document Length: 7483 bytes
Concurrency Level: 20
Time taken for tests: 118.087 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Total transferred: 8032000 bytes
HTML transferred: 7483000 bytes
Requests per second: 8.47 [#/sec] (mean)
Time per request: 2361.734 [ms] (mean)
Time per request: 118.087 [ms] (mean, across all concurrent requests)
Transfer rate: 66.42 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 1 4.0 0 33
Processing: 1256 2355 234.1 2284 3604
Waiting: 1256 2328 232.5 2264 3604
Total: 1256 2356 234.1 2284 3604
Percentage of the requests served within a certain time (ms)
50% 2284
66% 2328
75% 2412
80% 2500
90% 2666
95% 2768
98% 2877
99% 3204
100% 3604 (longest request)Or does that depend on implementation? I remember reading something about MRI only being able to utilize one core.
Both KVM and Xen use QEMU for their device models so the machine they emulate is very similar (at least in HVM mode). Xen and KVM use different paravirtual I/O drivers but everything I've ever seen indicates that I/O performance is better in KVM.
Xen PV guests are a mixed bag. On the one hand, they work on boxes without virtualization hardware (which are rare anymore) but on pre-Nehalem/Barcelona processors, they are much slower than using HVM (especially 64-bit).
I've never seen a good Xen HVM benchmark either comparatively speaking... Of course, I'm more than a little biased :-)
HN loves benchmark porn. Throw some dramatic charts up, show approximately zilch methodological control, voila: front page.
(Complete with sarky comments like this one.)
My personal question is: are DigitalOcean making a profit?
Let's suppose for a moment that I actually wanted to do a performance comparison for my own purposes (which means I actually care about the actual results). How would I do it right?
Yes. If you do the math they actually do break even and manage to squeeze some profit.
Benchmark does not prove thing for Webserver other than Disk I/O is greater with Digital ocean. If Disk I/O was so critical then no one would go with AWS since they have got the worst of all.
Benchmarks need to prove a thing in its application, unless it does its waste of time.
However, why in the world is this methodology relevant for the hosting service? Would any of you really do any x264 encoding on an outsourced server, or would you use your desktop?
I would much rather get apache running, and run the ab (Apache HTTP server benchmarking tool).
Because it is during security incidents that you really find out what type of company you are dealing with. With Cloudflare they informed users quickly, were transparent about what happened and took their licks. With Linode they didn't inform users and we still don't know what happened.
As a former customer I would never, ever use Linode again.
Linode is a toy and should never be used if you value the security of your data.
Substitute "Linode / data" for "Bitcoin / money" ? Sorry, couldn't resist.
Looking through your old comments and linking out a bit it seems the incident occurred as a result of a leaked password from an employee. Seems pretty simple, and I take them at their word in their statement (http://status.linode.com/2012/03/manager-security-incident.h...) that they've made improvements here. Only a month later Bruce Schneier praised some new password related features being set up (http://www.schneier.com/blog/archives/2012/04/password_secur...).
I'll agree they're obviously in no mood to make public comments about this, but I think that might be part of that I think they're more than happy to not be a destination host for users that want to store unencrypted bitcoin wallets on hosts. I'd certainly never store anything that translates into money on a public host.
We've heard you http://news.ycombinator.com/item?id=4840352 again http://news.ycombinator.com/item?id=4121734 and again http://news.ycombinator.com/item?id=4911934 and again http://news.ycombinator.com/item?id=4652662 and again http://news.ycombinator.com/item?id=4427802 and again http://news.ycombinator.com/item?id=4545458
In every Linode thread that pops up, no matter what the thread is actually about.
We get it. Move on.