too good to be true?

Posted February 24, 2013 29.4k views
I have been using Dreamhost ($10/month) for over 8 years. Still with them. Ten days back I sign-ed up for Linode-512 and I am quite impressed. I don't mind paying 20 bucks a month for VPS --- 24GB disk and 1TB transfer meets my needs but I prefer 1GB of RAM. Then I heard about DigitalOcean and signed up for it yesterday to check them out. While not as polished as Linode --- which is to be expected --- I found it quite good. I could do a side-by-side comparison of running nginx, php-fpm, https etc and the 1 core of DigitalOcean was not limiting. I have the option of trying the 1GB and 2GB plans (2 cores), pay for what I use, and still pay the same or less than what I would pay for Linode. So, what is the problem? The problem is that it is too good to be true. DigitalOcean must be losing money at these prices and what happens when the seed money runs out say in 6 or 8 months? $1.25 per week wouldn't probably pay even the power bill. Lots of money will be required to expand beyond NY and Amsterdam. Linode would typically give free upgrades every June and if they upgrade the RAM to 1 GB and make backups free then it would make Linode look quite good. Linode has been around for 8 years, they may charge more but the probability of them being around a year or more from now is almost 100%. My worry is that if DigitalOcean folds without warning a year from now and doesn't give option to copy your data it would be a problem. So we have to have frequent backup outside of DigitalOcean to account for this. Linode is putting forty 512M nodes on a server. Is DigitalOcean putting 160 nodes? Babu

These answers are provided by our Community. If you find them useful, show some love by clicking the heart. If you run into issues leave a comment, or add your own answer to help others.

18 answers

The cost of hardware drops when your a power buyer and I would presume DigitalOcean is in a position to receive pretty decent discounts on hardware given their position in the market in just 24 months. Having seed money and financial backing just deepens the discounts and resource availability.

Any company with a large enough deployment and continuous purchases from a supplier is bound to get specialized pricing which converts to lower costs per physical server deployment.

While a rather broad example, if one server costs an everyday buyer $12,000 to build, DO may be in a position to build the same for $6-$8,000 which is a rather substantial reduction in costs, thus they're able to recover their investment quicker, even with lower costs per month to the end-user.

If you take the number of Droplets that are said to be deployed right now, and if that is a current production number, and you just take the base cost of $5 a month x Production, that's a rather hefty $500k+ per month. Of course, that discounts anyone use anything other than the $5 a month VPS's, so I would say that number of much higher.

Of course, I'm sure Moisey and the rest of the team isn't going to divulge exact numbers, but it's not too hard to see that they are indeed making money. As long as they keep up what they're doing now and as long as Mosiey sticks around, I think you'll see DO in operation 6 years from now, giving them the 8 year advantage as well :-).
We're going to be celebrating our two year anniversary this coming June so while we haven't been around for 8 years, we have been at this for a bit of time.

While I can't comment on the specifics of our business I can tell you that we are profitable and we are looking to provide the best service possible to our customers and also to continuously improve on the product.

Aside from that I would say sign up, use the product, and let us know how we can improve.

Thanks for the prompt reply. Glad to hear that you are profitable.
I am using the product, and as I have said, I am quite happy with it.
I think you hint at the reason they can be cheaper: 1 core for the base package vs. 4 on Linode at 4x the price.

I'll admit, when I signed up last night, the core distinction wasn't entirely clear to me, but at the same time, I'm only looking at a small blog and a few Rails apps that I run for personal use right now. Even the $5 plan seems like it'll be fine for me, although I'll have to see the memory usage with both WordPress/php5-fpm and nginx/passenger running in 512 megs.

I would love to hear from Raiyu on whether only having a single core puts it at a significant disadvantage here.
I did a benchmark comparison of the 2GB plan before leaving Linode for here. It was nearly exactly 30% slower with 2 cores compared to Linode's 4. At 1/4 the price simply adding a second machine puts you ahead.
One issue came up when compiling on DO-512. The same compilation ran successfully on Linode. On DO, I got some cryptic error at link time.

I then created a snapshot, a 1GB droplet with that snapshot --- this is one of the cool things about DigitalOcean, pay for use, experiment etc. This is like a hybrid between Linode and EC2. The compilation succeeded. However, I was wondering why it worked on Linode with same memory. Then I realized that when I setup Linode I was given an option to choose a swap partition. On DO, there was no option. So I fired up the DO-512 and followed the instructions (dd etc) to create 512 MB swap and then I was able to compile.

Things that will be nice to have:
1) Provide for a default swap (same size as the memory) at droplet creation.
2) resize without losing the assigned IP address.
Resize between hypervisors has been a bit problematic and we've been making a lot of code updates to improve it and it would migrate the VM between hypervisors preserving the IPs.

We don't provide SWAP by default but we do have articles in our community on how to set it up, that was just a choice we made in favor of simplicity. But we can always revisit it if customer demand is high and get one setup off of a harddrive swap file.

In regards to the cores - it was just a decision based on pricing and resource allocation.
Swap can be easily created if you're a sysadmin. And in my view, digitalocean's servers are much faster than anyone else. The hard drives are faster than my dedicated server. The servers are quite cheap as well. Good job.
We also have a couple of articles in our community section that detail how to create and activate a swap file in both Ubuntu and CentOS.

I understand the economies of scale.
I will take Moisey at his word that they are profitable.

I guess that this business generally provides low margins and so you need huge scale. This volume purchasing power applies even more to larger more established players like Linode, Rackspace with longer track record of making payments --- assuming servers are purchased on credit. If DO is profitable at their current pricing structure then it means that Rackspace, Linode are enjoying much better margins. If DO grows at current rate then it will push Linode, Rackspace etc to lower their margins. I think we are already seeing that with Linode's announcement of linode-nextgen --- here there are not reducing their prices but 'providing more' by spending a million dollars to completely upgrading their network to as they say "deal with our scale; greatly improve throughput;decrease latency; and add redundancy at access/host layer". Linode has also increased the outgoing bandwidth limits by 10X --- instead of 200GB for their $20/month plan, you get 2TB, which is more in line with many VPS vendors.

Getting back to scale, DO needs to open at least 2 more locations --- Asia (Singapore or Japan) and West Coast of US --- and this will require significant expenditure for leasing building, payroll etc. Also this may not result in lot of new customers but existing customers will migrate to a closer location.

Finally, I think there is room for all these players to thrive and customers like us can get the best possible price and performance.
I am in the exact same position as the OP. I'm a very happy Linode customer and when I heard about DO, the prices seemed to be to good to be true. One of things that would really help to make me feel more comfortable with DO is some technical specs. Some of the questions I'm having trouble finding...

What are the specs and density of DO's current servers?

A fresh server seems to run great, but what happens when that server is loaded and my server is starved for CPU resources? Linode publishes it's density and EC2 uses cgroup style limits to guarantee resources.

How are the virtual disks provisioned?

Linode provides RAID-6 redundancy and EC2 has long publishes the S3's policy. While free backups are great, how do you handle data corruption and drive failure?

How is bandwidth allocated?

Again, my fresh droplet is blazing fast, but what guarantees will I have when the current node is maxed out?

DigitalOcean has a VERY compelling price point, but until I have some more information about the technical aspects of the environment, I could never move my core customers over from something like Linode. Until then I couldn't recommend DO for any serious workloads.

I hope that this information, along with a feature roadmap, will be made available publicly soon.
+1 for automatic swap creation, wouldn't it help people have some peace of mind?
Swap is bad, I wouldn't want it turned on by default. Makes me have to go turn it off...
For what its worth - DO is certainly offers the best price - feature ration compared to Linode! So far very impressed with what DO offers. Certainly there are quite a few features missing when you move from Linode like Swap IP, Load Balance, HA Setups etc

Check this out on Linode vs DigitalOcean Feature & Price Comparison with 8GB plans

By the way Linode is revising the infrastructure, Linode sizes, prices etc continuously to beat the competition with great service - way far from Do.
DigitalOcean is certainly on the winning side with good price, feature and performance! Checkout those image links ... !!!
One thing is missing in comparisons is use of fast SSD drives. If your application does lots of I/O, then this will make good difference in performance.
My biggest problem with DO is the limited selection of server locations. I get much better latency on my Linode than I do on the the closest DO server. The disk I/O is faster on DO, which is nice.
@shawnbadger: We have a few more regions on the roadmap, such as Brazil and Asia. No ETA yet, though :]
I have been with DO for about 2 months. In this time I have had no down time, easy access via a control panel and had an opportunity to scale up and scale down server resources as required. After using VPS packages from many different ISP's I think DO is a awesome concept. I will definitely stay.
Submit an Answer